My question is: Why isn't anyone else doing this specific analysis? You would think that governments, universities, etc. would be motivated to do this type of analysis. (Maybe they are, and I just don't read the right blogs to see those posts?)
As a statistician, I'm very skeptical of people who crenulate their zeugmas. Just because you have some asymptotic zuegmoyal theory doesn't mean that it applies well to finite samples. If you're going to be a Crenulist, at least verify that Epstein's theory holds in your data set!
> One thing I found helpful was to ask Person A what they thought of Person B's work, relay it to Person B, Person B would say something like "he's neglecting to consider that you need to fulminate the synecdoche", then I would relay that back to Person A, and after enough of this I would get a meaningful sense of where they disagreed (or occasionally one of them would just admit they had made a mistake).
This is *EXACTLY* the model I've found myself stumbling into working on my Georgism follow-up. I've got all the Georgists on one side telling me everything checks out, and then a bunch of skeptics on the other, and all of them have impeccable credentials and PhD's in econ or tons of professional experience, and I'm just a guy who read a book and wrote about it on the internet that one time.
So far the above method seems to work. Speaking of:
@Erusian, if you're around I'd like to get in touch off-site so I can send you some of the best arguments about the practicalities of land assessment so I can start the claim->response->counter-claim->counter-response cycle, if you're up for it. Email me at lars dot doucet at gmail dot com if you're game.
Many many years from now the verdict will be that lockdown effectiveness was not worth the negative externalities inherent to the lockdown, regardless of the marginal benefit some lockdown statistics have shown. I know that no one will, but feel free to screencap this and throw it back in my face at some long distant future date, so long as you acknowledge my ability to do the same.
I think there is a substantial difference between someone who leans liberal or leans conservative and someone who is an ideologue. If someone like Tyler Cowen - who leans libertarian/conservative - comes out in favor of lockdowns that voice carries more weight with me. Alternatively Freddie DeBoer's writings on human ability, testing and education in general carry more weight with me as he is from the very far left. But in both cases they are calling them as they see them. They are not insisting on finding facts that fit their existing ideology. Or if they are, when the facts don't support their ideology they are open to saying, "My theories don't work in this instance."
As far as whether a conservative prognostician arguing against lockdowns is more "marked" than a liberal one arguing in favor, I don't think it's necessarily determinative, but I do think it's relevant that under ordinary circumstances, nobody favors lockdowns at all. It's not that lockdowns in general are a liberal position which you'd expect liberal people to be in support of unless the evidence weighed really heavily against it, it's only due to extraordinary circumstances that anyone supports lockdowns.
This is a distinctly America-centric perspective, but I think you could tell a similar story about other countries, where conservatives are biased against taking the pandemic seriously, not because this is an inherently conservative position, but because an important figure, Trump in America's case, took a stand on downplaying the pandemic. If Bob Woodward can be taken at his word on the subject, Trump's efforts at downplaying it don't even reflect his own best understanding of the pandemic at the time, since he thought having people take it less seriously was better for the economy, and likely him by extension. If we had had a different Republican politician in office, someone like George W. Bush perhaps, then taking the pandmic seriously might never have become a tribal signifier in the first place.
The upshot seems to be that human societies are complicated, social science is insanely hard, politics is even harder, and even the smartest and most honest politician has to rely on a combination of judgment, intuition, and common sense to implement their policies. That's why open debate and epistemic humility are important.
Yeah, most of the time people support their own side's ideological positions. Unless you know everything about crenulating zeugmas you're going to have to either A) if orthodox, discount because orthodox ideologues are sometimes sloppy due to most of their peers being relatively uncritical towards the work, or B) if heterodox, discount because it could be a clever arguer trying to sell you a box that you do not want.
Of course, you could be dealing with a clever arguer also in case A, so maybe you should discount orthodox arguments slightly more anyway. I have no idea, this problem stumps me. I feel you.
At least this was a good reason to post that cute bat again. More covid posts, please.
One issue I have about emotional costs is the comparison point. It's a global pandemic, an intrinsically "bad" event, and life is going to be less "good" than before. There's going to be long term trauma about how poorly it was perceived to be handled and how ill prepared we were for it. Some of the measurements of quality of life are good at showing the impact of the pandemic, but not the cost effectiveness of lockdowns.
> One thing I found helpful was to ask Person A what they thought of Person B's work, relay it to Person B, Person B would say something like "he's neglecting to consider that you need to fulminate the synecdoche", then I would relay that back to Person A, and after enough of this I would get a meaningful sense of where they disagreed (or occasionally one of them would just admit they had made a mistake).
It sounds like you kind of rediscovered the Delphi method? RAND came up with a consensus-building framework specifically for opposed experts back in the 60's and it seems to work *really* well. I wish all political and economic arguments were given that treatment. Original article here:
To be honest, I found this post much more interesting than the post it is about.
I have been trying to get this across to everyone who will listen, for over a year now, and I haven't been successful but I can't stop because it's extremely important to my continued survival:
The question "do lockdowns work?" is incoherent and unanswerable, because both "lockdown" and "work" are and will always be underspecified.
You can't aggregate over millions of people, summarize death counts or economic costs or whatever, generate an average, and use that number to judge things. As you point out above, this question is much too multifaceted to do that
But I'll make a stronger claim: you can't even do this in principle. One might be tempted to say "aha, well, maybe there's 50 dimensions but we can analyze all of them and come to conclusions" but you can't! Because people have varied lives and, no matter what conclusion you will come to, somebody somewhere will say "if we did it your way, I would not be alive right now. Eff you" and there is absolutely nothing you can say to that person to change their mind.
There is no way in principle to settle this argument. It is a bona fide conflict, and the way it gets settled is the guy who gets to decide, decides, and then says "tough if you don't like it". That is in fact what our society did last year, and I didn't like it.
All of the mainstream, public scientific discussion about the pandemic is bullshit. It just is. It is not the case that a whole bunch of honest, concerned, and curious scientists all decided to dive in and see what happens. I know we here like to think that's the case, because we here actually do do that. But the rest of the world does not. The rest of the world does the science they're paid to do. If it gives the results that the people in power want, then they get published and publicized. If it does not, they do not. As a result, you end up with a bunch of 'science' supporting what the people in power have already decided they want to do, and the people in power use that 'science' to wash their hands of any responsibility for people being mad at them.
You can see this happen in real time if you pay attention. This is a relatively trivial example, but it is very accessible: Dr Fauci's overnight 180 on masks, and everyone acting like this didn't happen (but was totally fine if it did). Prior to his flip, Fauci and the rest of the medico-political establishment told us that The Science Says Don't Use Masks, and the news dutifully reported that. Then, overnight, Fauci says that The Science Says Use Masks, and the news dutifully reported on that. When questioned on it, all they say is "we're just following the science, and the science evolved". But we know that's a lie. Why? Because Fauci said it was a lie under oath to congress.
In reality, the actual fact of the matter, which I expect most of you here either knew or just assumed, was political. They were afraid of a run on masks, which would put medical practitioners at risk, and so they told the public not to wear masks. Then, when they were satisfied that the medical practitioners were well supplied, like magic, 'the science changed'. In reality, the government gave out an arbitrary order ("do not buy masks") in order to achieve a domestic policy goal ("don't let hospitals run out of PPE"), and instead of just saying that, they lied to our faces and bastardized 'science'.
In fact (I'm not gonna go look this up, so apologies if my memory is wrong), I think that was the conclusion you came to in your original mask post. The public health advice prior to covid was "do not wear masks", and the reasons for this was a) people wear them wrong and it wastes them; and b) that takes them away from people who actually need them.
What is the point of my work-procrastinating rambling? My point is that fundamentally you cannot use 'science' to determine if 'lockdowns work'. The overwhelming majority of the 'science' on this is science in nothing but name. If you do _actual_ science to try and get an _actual_ answer, everyone will ignore that answer with some variation of "yeah but no because thing I care about". You can ask extraordinarily specific questions like "did imposing a 8pm curfew reduce the number of deaths by 500 or more". Even that isn't really a great question, because imposing a curfew doesn't mean people comply with it, and so it equivocates the question of "does adhering to a curfew work" with the question of "does imposing a curfew _policy_ work" which, despite appearances, are two very different questions.
Maybe, if we constrain ourselves the platonic world of abstract thought, that cannot interact with the outside world in any way, we could come up with The Correct Answer to the question of lockdowns. But that's not particularly useful. Implied in the question of "do lockdowns work", is the hidden question "should we implement lockdowns". But as is abundantly clear by now, the decision process we use to implement lockdowns is completely disconnected from the actual answer to the question of whether or not they work. The whole question is poorly defined, and the answer doesn't exist
TL;DR version: lockdowns: good/bad? - is a waste of time that we could be spending on a more important question: how to persuade people to demand a better overall pandemic response (and prevention) in the future. Your post about it was great and fascinating and I loved reading it, but I will straight up paypal you money today if you produce one as good about "finding the best way to demand better next time."
Long version:
I think this is the key point: "But the smartest people I talked to kept - is "derailing" the right word? - derailing onto more interesting and important pull-the-rope-sideways plans. "
And:
"none of that turned out to matter at all compared to some question about stoned driving vs. substituting-for-drunk-driving, which nobody started out caring about."
Because that's exactly the truth - lockdowns yes/no is a very interesting (and extremely difficult) question to answer. But really, for the question of "how should we fight pandemics" it doesn't matter that much - there are other interventions that will have a massively larger effect (and in fact will make doing the research to determine if lockdowns work really hard, as you discovered) that this is like trying to figure out if you should amputate the left or right leg of the guy who got run over by the wheat thresher. The real answer, for society, is taking precautions to avoid the "got run over" scenario, or "let a pandemic get so bad you are debating lockdowns" scenario. The various "Titanic deck chair rearrangement" plans don't move the needle very much.
In fact, given the pandemic's ability to inspire Americans to polarize even the most mundane things, I think that the true answer to your question: "lockdowns? yes/no?" is actually "I think you should stop asking this question." I say this is as a libertarian, first amendment maximalist who has indeed been following this issue and happily debating it with others for a year and a half: this is a rabbit hole that doesn't have... whatever rewards rabbit holes are supposed to have at the bottom of them.
I don't mean to diminish the importance or quality of your work (or anyone else your reference), but for the goal of having the public know what good pandemic response is, this just doesn't help. (I intend no moral or judgement criticism here AT ALL) It is so hard to isolate, so amenable to data cheating, mistakes and incredibly confounding errors (maybe lockdowns work in areas where people believe they work and therefore cooperate, but don't in places where they hate that kind of thing?) that time spent worrying about it just convinces the average person that the question DOES matter, and that other questions (that really do matter more) don't. That is, those people who tried to derail you are exactly right: every minute that you (and all the researchers you cited) spent trying to figure out the truth of this question was time that could have been spent mobilizing public demand for the truly effective pandemic solutions (which surprisingly seem to be far less controversial, and easier to prove they work in practice), for a net benefit.
By any objective, scientific - and indeed, conventional-wisdom-approved measure, there are lots of things (again, many uncontroversial and agreed upon) we could have done better on the pandemic. Politicians and bureaucrats from both parties (or even unaligned) flubbed it to previously unimaginable levels (I used to think the CDC was great!). So the fact that we have no mass movement to Do Better Next Time (and stunningly little promise that any changes have been made, or that anyone at all was held accountable for failures) is shocking, and trying to fix that seems like a much more fruitful use of (anyone's) time.
If someone could figure that out, I would love it. I would love it so much that I would send them money right now to work on it.
It's interesting to see the question of "how much more do you trust someone arguing against their typical ideological bias" in a very different context than I'm used to seeing it in. I work in a corporate business environment, and the analogy my boss likes to use is "PETA vs. the butcher", in that those are two people which you would expect to have very different opinions on the sanctity of animal life.
So yeah, assigning additional credibility towards people arguing against their typical positions is meaningful - but exactly how much additional credibility to give something, and how much more seriously to take the argument, is a very interesting question. On first glance it's the sort of thing that seems unavoidably subjective; every individual arguing against their ideological bent is going to have a varying degree of attachment to their "tribe" and therefore the emotional or social costs of believing things that go against their prior attachments - and thus the amount of additional 'virtue' we would assign them for 'having done science correctly' - would similarly seem to vary on a case by case basis.
Do you award extra 'virtue points' if the head of the NRA argues for gun control in a specific scenario, vs someone who's just a member, vs someone who's just Republican?
What about the opposite scenario, where the head of the NRA argues for total lack of gun control? My statistics background says "this is the null hypothesis, there's no additional information to be gleaned here" but is there ever a point where someone is so ideologically committed to certain positions that even otherwise good-seeming arguments that they are making should be discounted or taken with a grain of salt? Maybe if they're trying to sell you something?
An interesting question that I'm going to be thinking about!
1. What you describe doing in representing A to B and then B to A corresponds pretty closely to the job of paidhi in C.J.Cherryh's (very good) sf Foreigner series. The paidhi first acts as A's agent in representing his position to B, then switches roles and acts as B's agent in representing his position to A.
2. Your experience trying to figure out the net effect of either lockdowns or Marijuana legalization is similar to mine looking at the effect of climate change. There are a whole lot of possible effects, positive and negative, and what one mostly sees is a calculation based on a subset biased towards the result the person calculating wants or expects. Nordhaus gives a lot of weight to very speculative calculations on low probability high cost effects of climate change but, so far as I can tell, never considers low probability high cost effects of preventing climate change — even though the obvious one, the end of the current interglacial, is a climate effect that has happened multiple times over the last few million years.
There is one way of including emotional effects in the calculation — leaving individuals free to decide for themselves whether to quarantine. They have the right incentive with regard to emotional costs, since they are the ones bearing them. It's an imperfect solution because they have too weak an incentive to take precautions — if they get Covid, much of the cost goes to them but much to other people to whom they might pass it. And all of the cost if they already have it and don't take precautions goes to others.
My family in fact quarantined more tightly than the rules require until we were vaccinated.
On your problem with fancy statistics ... . I tried to follow the controversy over the Lott and Mustard paper on the effect of concealed carry, eventually stopped when the statistical arguments got above my level. Like you I have taught low level statistics and, while my mathematical background is much stronger than yours, my statistical background probably is not.
"Apparently pro-lockdown academics who get too close to the public spotlight have been getting harassed by lockdown opponents, and this is a known problem that pro-lockdown academics are well aware of. I was depressed to hear that, though in retrospect it makes sense."
I'm very new to this blog. I love the smart writing. But man do I wonder sometimes what alternate universe the author lives in to not know this stuff has been going on for a year and a half now. Federal, state, county and city health directors have received death threats, with armed protestors sometimes showing up at their house. The health director instantly becomes the most reviled person in their region. Many have quit over it.
Do you really believe that the mathematician with his zeugmas, crenelated or not, started off with a totally unbiased mind, a tabula-rasa, and let the math lead him to a conclusion? Or do you believe that he started off with some emotion, instinct-driven position and found the data and the math to justify it?
You can’t perform any meaningful post-facto analysis of the pluses and minuses of lockdowns because you can’t perform a test with one America locked down and one not locked down. Comparing states is a worthless exercise – the social, economic and cultural differences between, say, California and Alabama, are simply too great. The researchers who have tried to do that just prove my contention: they begin with a belief and try to find the data to support it, ignoring the fatal limitations of the exercise.
The same is true with data for marijuana usage: we know with some degree of confidence the typical health and behavioral impacts of marijuana usage, but we cannot possibly forecast the net benefits\harms caused to American society at large, and all those who pretend we can and that they have the right methodology, have started out intending to prove their instinctive belief (even when they’re adamant they haven’t) and produced nothing of value.
So, what do we do when decisions have to be made, like about lockdowns? Well, we start off with what we know for sure:
Covid is an infectious disease spread by contact. Covid is potentially fatal or highly debilitating. We have very limited facilities for treating patients with serious cases of Covid. People will respond to images of hospitals denying admission and people dying at home in great distress by self-locking down. Such ad-hoc quarantining will be chaotic and lack legal protections, and will cause great economic damage.
We don’t need Zeugmas to be sure of any of this. All we need to do is to confidently state the common-sense facts, draw the obvious conclusions and act accordingly. For the most part, that, thankfully, is what we did.
One description of why the unknown unknowns can be high impact and unpredictable:
Suppose the US had absolutely no lockdowns. This would probably make the population dislike the incumbent Republican government more, which would probably lead to the Democrats win with a huge majority in the election. Whether this is good or bad is up to your interpretation, but it could plausibly have an impact much larger than the marginal impact of the lockdowns.
On the other side, suppose the US had stronger lockdowns with more bipartisan support from the federal government. This might lead to the incumbent Republican party staying in power in the election, especially if the stricter lockdowns had a large impact on keeping case counts down. Similarly, this could also plausibly have an impact much larger than the marginal impact of the lockdowns.
Of course, the assumptions behind both of those scenarios could be incorrect, but they're at least *plausible*, making the whole situation impossible to evaluate to a binary yes/no.
"If a professor comes up with some study that shows guns decrease crime, and that professor is a gun owner, goes hunting every day, and donates to conservative organizations, is that suspicious? What if another professor shows that guns increase crime, and this professor has never owned a gun in her life, hates hunting, and donates to liberal organizations? It feels like "enthusiastic gun owner" is more of a "marked" group than "non-gun-owner", but that's just a coincidence - if we had been in a world where academia leaned conservative and most professors owned guns, it would be the opposite. But should we really discount the fact that the pro-gun study professor has an NRA bumper sticker on his car? I’m still not sure how to think about this."
This strikes me as the fundamental crisis both of rationality and our current moment. My sense is that most people have already surrendered to the view that political or class alignment overshadows all evidence of unbiased analysis or opinion. And its not like there aren't plenty of real examples of this! But if we can't find any way past this problem we are probably screwed.
When there is enough uncertainty, model outcomes reflect the biases of the author a surprising amount of the time. I'd do a study on this but I'm biased that this is clearly true.
Here is some advice, if you don't like questionable statistical manipulation then never look closely at some areas of climate science. This is particularly true for the ones that make it through the media's new and alarming results filter.
A lot of statistical manipulation can be a sign of a very difficult problem, it can also be a sign that the less complex models weren't giving answers the author liked. For some problems where there were literally new statistical methods invented for analysis it would be nice to see a progression of different model outcomes and how the results were affected. Sometimes raw data is just crap, and you can't polish it no matter how hard you try.
I'm very surprised and sorry that <i>"it just felt too weird and transgressive to focus on something authorities weren't even talking about."</i> That's one of the biggest reasons I started following your writing, Scott - you've been willing to focus on weird but important aspects of situations, even when authorities aren't talking about them!
The toughest thing about this topic for me is the fact that for many conservative places (like Indiana where I live), "lockdown" was largely hypothetical! Yes, stuff closed for a bit, but for most of COVID you could go most places and largely wouldn't even be asked to wear a mask.
How do you study the effects of a mask "mandate" (read: suggestion) that more than half the population simply ignored? Surveys seem like a very thin way of learning things like "a quarter of the people at the grocery store don't wear masks" and "lots of people never stopped getting together at all."
I could tell that paper was exhausting to write, and thank you for going through with it, even if the results weren't quite what you wanted in terms of clarity.
Quick back of an envelope gives me that if the whole world was as prosperous as W. Europe (where I am) that would save ~3m child deaths/ annum
3m/ annum is roughly the C19 death rate over the last year. I have no idea what would have happened if we'd done nothing and carried on. If we'd doubled the death rate but the economy had carried on growing normally we'd be at break even.
The correlation between wealth (GDP/capita) and longevity, both in life expectancy and child mortality, is overwhelming. I saw some talk about this in 'phase 2' of the conversation last year but it seems to have been forgotten
"Other times I just went for maximally stupid models and saw what happened"
I think you're probably being too harsh on this as a good judgment method, at least in this case. Lockdowns are, in general, very 'maximal' policies; they are not "adding a penny tax to drinks at the bar" or "restrict advertisements for tourism" which are the much more typical policies you might see a government introduce. They are a great cudgel of a policy.
As you state, this is helpful in bounding things, but given the political interests, and the immense difficulty of teasing out impacts, it should bound people to a great degree. If the broad policy of preventing interactions fails to produce broad results, we should be quite suspicious of very careful, complex models. Given the replication crisis and similar scandals, I think a skeptical instinct in these cases would prove a better judge overall.
If one was studying lockdowns, why choose red vs blue states? Why not instead choose US vs China? Because that's an emotional ouchie for many.
I would argue that the targeted comparison reveals a lot about the motives of an author. Even red and blue states aren't that much different. They each have large percentages of the opposition who behave according to their tribal tendencies regardless of government edicts. I'm no fan of China but the real question to me is whether hard core China-style lockdowns are what is needed once widespread transmission is underway, not these flip a red/blue coin and micro-analyze the results.
I live in Florida and we have been bombarded by the national media with partisan accusations constantly. No state mask mandates! Curiously my county was under mandatory mask orders for over a year, as was almost every metropolitan area in Florida. Florida left the decision making to the local areas. Florida-Man-Moron. Aaaaaghh, people on beaches! It was all so tiring and wasteful to read.
Nobody won against covid. Tabulating the losses to 50.1/49.9 is not very interesting. Open a county level national map, set it to cumulative cases/deaths per population and scroll around. The timing was different, but the cumulative results were pretty uniform. There was a big dry forest and it burned down.
"One thing some people fairly called me out on was asymmetric political bias adjustment."
Yeah, I think you are correcting in the wrong direction.
Who gets better at arguing and defending their points: Someone who is surrounded by people that agree with them, or someone who is in a hostile ideological environment?
I think if you want to be an openly conservative or libertarian academic you have to be *better* than your peers within your field. Otherwise you'll get washed out. Just like anytime you see a racial minority succeeding in a field with discrimination, they are almost always above average in that field because they are operating with a handicap.
Scott: Thanks for the effort you devoted to the article. I do not think we are near a final accounting on this issue. And, I doubt that we will ever achieve real clarity because some information that would be very good to have was never gathered, such as: We did not do random tests of healthy individuals to determine what the real rate of infection was; There were nowhere near enough autopsies performed to be able to get an accurate picture of dying with COVID vs dying from COVID.
Further we do not have a good grip on many collateral issues:
"Study: Hospitalizations for eating disorders spike among adolescents during COVID" by University of Michigan
"The number of adolescents admitted to the hospital for severe illness from eating disorders has increased significantly during the COVID-19 pandemic, new research suggests. At one center, the number of hospital admissions among adolescents with eating disorders more than doubled during the first 12 months of the pandemic, according to the study that appears in a pre-publication of Pediatrics. The 125 hospitalizations among patients ages 10-23 at Michigan Medicine in those 12 months reflect a significant increase over previous years, as admissions related to eating disorders during the same timeframe between 2017 and 2019 averaged 56 per year."
"Emergency Department Visits for Suspected Suicide Attempts Among Persons Aged 12–25 Years Before and During the COVID-19 Pandemic — United States, January 2019–May 2021" Weekly / June 18, 2021 / 70(24);888–894 On June 11, 2021, this report was posted online as an MMWR Early Release.
"In May 2020, during the COVID-19 pandemic, ED visits for suspected suicide attempts began to increase among adolescents aged 12–17 years, especially girls. During February 21–March 20, 2021, suspected suicide attempt ED visits were 50.6% higher among girls aged 12–17 years than during the same period in 2019; among boys aged 12–17 years, suspected suicide attempt ED visits increased 3.7%."
Re. Bias adjustment: I'd say it's valid, in that anyone who breaks from the consensus needs to be more harshly evaluated, even if that consensus might be political in nature; Like how market skeptic economists get raked over the coals
> For example, one person brought up that I (as a younger person) might be underestimating the emotional distress older people feel about COVID, because they (unlike me) are at serious risk of death.
Isn't it the opposite? For me (and my young friends), and lots of people around 25 (in France), the conclusion is that we sacrificed a big chunk of our mental health and youth to protect the physical health of old people.
If you have a good memory, you can play blindfolded chess against 10 GMs and score 50%. Just pair them up and let them play each other with you as a conduit. Derren Brown does this on TV with a twist (he takes 11 and defeats 1 to end up with a winning record).
I don't think "emotional damage" is the right term. It wrongfully implies that the damage of being unable to do normal activities like going to a bar is somehow different from health damage. It's not. The reason we dislike being ill isn't because it gives us some abstract "bad health" state. We dislike it because it stops us doing normal things. That's what health actually means to most people. Ability to do normal activities is a critical component of QALY - a disability that has no harm beyond making you unable to go to work, unable to do recreation, and unable to see your extended family would be considered a devastating health impact, not just be called "emotional damage".
Thanks for writing this one and the earlier one, Scott. There is a lot of clear thinking still to be done about the global Covid issue and moving it forward is important. Also nice technique for getting researchers to talk to each other.
This got too g-d-n long but I think I had a point.
Part of the challenge of analyzing lockdowns in the United States, and some commenters have mentioned it somewhat - I am not sure anywhere in the US "really" did a "real" lockdown. There were narrowings, yes, there were some stoppages - school! work! but the combination of society, government and policy delivered something other than "lockdown." Some call it closures, I'm sure there are other names. Some smaller locales (e.g. Indian reservation) had curfews and "only residents allowed in" policies but this was not widespread.
So, in the US, lockdown it was not, and so the dramatic mental references that attach to that word are not as pertinent. Economic damage, business damage, livelihood damage, educational damage, relationship damage, all to varying extents. We don't need the word "lockdown" to communicate that there were bad consequences. Some of the bad consequences derived from the partial nature of the restrictions; we didn't shut enough down for long enough to make enough difference, and it's possible that fewer shut-days would have been required to achieve the same outcome or better, if the shut-days and open-days were distributed differently.
I remember late Feb/early March 2020 because I had just left a job and moved to a different state. I had enough money to get by while finishing the move and for the first time in a long time I was not freaking out about 100 things. So when I'd get up in the morning and hear birds and there were no contrails and little traffic and then we'd go to the park and walk around trying to be socially distanced, it was like a gift. Which felt dirty in a way, because people were dying, suffering, struggling, but for my specific situation it felt like, ok, this is what it takes to lift the miserable rat race lifestyle off of us and let's make sure we never accept it back. For many people that lift did not occur at all but for some it did. Air quality improved dramatically where I was. My suspicion is that people who could work from home, had some conscience for their fellow man and were able to act on that by conducting isolation, and tended to left/center-left politics, may have felt that lift more than other groups. And so now for some people the word "lockdown" means "that lift." When the people with the means to do so undertook more of a "welfare of humanity" project than perhaps ever before, and it was one uniquely suited to the "personal is political" stance; making personal choices for the benefit of humanity, writ very large. It hit a lot of buttons.
That began to sour a few weeks in though when people were still going out in public but not as often and some of the social graces eroded; people who had gotten very good at pretending not to fear each other now showed furtive glances in stores, plus, what if someone coughed on me? And it had never been good for people without financial security, work-from-home, decent daycare. Then the "welfare of humanity" project shifted to a laser focus on the experience of these essential workers, who needed to be helped. Then the welfare of humanity project, which was already coming apart, became aligned with social justice in the United States and became even more "welfare of humanity." But by now "lockdown" meant "welfare of humanity" for enough people that discussing it was difficult. "Lockdown" did not mean "welfare of humanity" for everyone by any stretch; even with what we had in the US which was not lockdown, even that did not mean welfare of humanity project.
So we have to face that "welfare of humanity" project did cause suffering, did cause harm, and there are a few levels to that. The tradeoff level is tempting because it gets over with faster. But we skip the "what is harm" and "what is welfare of humanity" and "what was lockdown" definitions at our peril.
An element of the harm which I don't see mentioned much is an increase in paranoia in multiple quarters. Paranoia of government, of international politics, of other people. Maybe sublimated fear of Covid itself, maybe just looking at the policy challenges and disagreements and struggle. The "wow, they're incompetent" reaction hit people differently and I think depended somewhat on how much an individual was invested in "welfare of humanity," Some people already thought the government was incompetent and were maybe saddened but not surprised. Others found themselves in contexts where paranoia/criticism was embraced, for the first time they saw behind the Oz curtain. That had some low-information aspects though but I think it partly spawned the Jan 6 capitol situation, and this a-ha moment was pretty widespread. It became "a-ha" versus "welfare of humanity" and there we linger at the social level. Covid-era has caused wakeup calls in different ways for different people; some who had some trust in society prior to Covid looked at "welfare of humanity" and "shutdowns" and realized they were at the mercy of people they didn't trust, and reacted to that.
I think "welfare of humanity" jumped the shark in early summer 2020 but I have not yet mapped exactly how. The "a-ha" has not always reflected reality and has moved further away in fact, but to me it is less hollow, it still has fuel, and so the rabid mistrust will continue. But "welfare of humanity" euphoria causes hangover, but since it also coincides with a left goal, or many left goals, it cannot be put down either. In July 2021 now it is paranoia versus hangover, with some liberal policy accomplishments.
> mostly (completely false) claims that lockdowns cause more suicides, lockdowns devastate businesses, etc.
As far as I can tell the first is (probably) not true (though we don't really know, who knows if the toll takes a few years to set in, Tyler Cowen thinks emotional connections that were cut during the pandemic won't recover), whereas the second is probably not entirely false, at least as a first approximation?
>And it's so stupid - emotional damages! People being annoyed that they can't go to the bar
A few sentences later you write how weird it is that the libertarian people are writing angry op-eds, where did you think this anger came from? I mean "emotional damages" seems like such an obvious thing; this is obviously what people are also emotional about when they hate lockdowns; why else would they be emotional? No offense, but how is this in any way surprising?
Great post, well-written, fascinating, as always — but I'm sorry, I've been completely sidetracked by something you mentioned.
There's a chance cannabis decreases IQ? By how much is it posited to do this, and is it permanent? Because I smoked a lot of marijuana back in the day, and my IQ is the only good thing about me*, and I would like to know how much dumber I possibly made myself.
*Actually, I'm kind, handsome, charming, and humble, too, but I won't mention that stuff.
I think this is a great post simply for admitting the complexity of the situation. The only thing I've found odd about this whole thing is that nobody – yourself included as I recall – contrasted the response to this pandemic to the response to the Spanish flu a hundred years ago. Hundreds of thousands of people died in America alone, and in a very COVID-like way, yet as far as I know there were never any involuntary restrictions anywhere, people just went "well that sucks and is terrifying" and went about their normal day until the flu became nothing more than... the flu. (I've seen an Everett True cartoon where Everett canes the living shit out of a guy for not voluntarily wearing a mask, but that's about the peak for stringency of enforcement to my limited understanding.) However, I'm far from an expert on this, and I've been hoping in vain all this time for someone who does know to do a carfeul compare-and-contrast on this.
"it's so dumb that it couldn't lie even if it wanted to"
I really love meta posts like this. You have a really interesting way of thinking about things. I don't always agree with your specific opinions but I am always fascinated by how you think about them.
History always ends up oversimplifying stuff, because actual history is far too complex to document in detail. I expect we will eventually converge on just using total deaths per million within jurisdictions over a fixed period (not yet over) as the sole metric of success.
Thank you for both the original and meta analyses!
Re the bias question: the expert class is almost by definition devoid of poor people, young people, and uneducated people. So it doesn't seem conspiratorial to suggest that the harm done by lockdowns to those groups of people is probably underestimated in expert analysis.
Wouldn't it be strange if that were _not_ the case? This seems like a more important issue to me than specific ideological biases, like "that guy writes in Mother Jones" or "that lady owns a lot of guns."
I don't think there's been an adequate accounting of the many ways that lockdowns greatly improved a lot of people's lives, partly because it's taboo to talk about a lot of people having a nice time while others are dying, and partly because even if it was asked about, people are reluctant to admit it because of said taboo.
Yet SOOO many women I know will secretly say that 2020 was one of the best years of their life, that the lockdown was a huge relief, and that it vastly improved their physical and mental health. A lot of guys do too, but definitely way more women. Granted, I am talking about middle class people who didn't lose their jobs. They just suddenly got an extra 15 hours of their life back each week, which was previously spent in unpaid but required labor (commute, etc.). And now because of the lockdowns, they have far greater bargaining power and flexibility at work. I understand the impact was very stratified based on demographics, but I think there is a major underestimation of all the positives, because it's just not very nice to talk about how a bunch of relatively secure middle class suburban people suddenly had more enjoyable lives, while older people were scared and essential workers were having a terrible time.
I don't know if there will ever be good research on this so long as it's taboo to admit, and someone who lost a family member or job is going to scream at you and tell you what an enormous jerk you are.
The junior associates at my firm are all getting a $15k raise and the ability to work flexibility from home or reduced hours -- things that were unthinkable pre-pandemic -- just to induce them not to leave. For them, it looks to me like the whole thing net positive, and not by a small amount.
I was initially appalled by the lockdowns. I could not believe that this country is so senior-focused that we were going to sacrifice the livelihoods of young and poor people on the alter of keeping a bunch of near-death retired people safe, when they were the ones who could most easily stay home. However, I'll admit that 2020 was a fantastic year for me personally, and continues to have positive impacts on my life. Most people I know say the same.
Also, are people really as polarized on Covid as some of these comments are making it seem? Not in my experience. There is a hysterical small percent on either tail end who are get very upset, but the vast majority of people are not and have never been like that. Most people complied with the orders without much grumbling, but also didn't particularly care whether others followed them or not. Out of about 300 familial, professional, and friendship associations I have that span the entire political spectrum, I knew maybe 7 or 8 who took adamant and angry stances on one side of the other. Most everyone else was sort of mildly but not overly concerned and rather live and let live about it. Every right wing Republican I know got the vaccine, and a large portion of the left-wing people would admit privately that they thought the whole thing was overblown and they weren't particularly concerned about catching Covid (and most of them broke the lockdown rules anyway). I think the supposed polarization is a bit of a internet creation. You can ask someone a yes or no question on a survey, but are they measuring the strength of the response and how strongly one feels about their answer?
I was also surprised that the two Astral Codex surveys on Covid asked questions solely about negative impacts on the respondent's life and not a single question about positive impacts.
I really enjoyed this meta process reflection on doing that long piece. I hope you do this kind of thing again, the process reflection, not necessarily the long piece, though that was good too.
It's an important insight that when you get closer to something, you find the naive concerns about it are irrelevant to what's actually going on. I find this is true across a wide range of topics and disciplines. Consequently I highly recommend getting close to things as soon as possible.
For example, if you think you might want to be a lawyer, go shadow three different ones for a day right away, rather than spending extra time studying for your SATs so you can get into a pre-law program, so you can get into law school, so you can get an internship, so you can get hired 20 years from now and then finally discover that you hate something about the job that nobody ever mentions. (John Taylor Gatto was big on this, and recommended "a thousand different apprenticeships" starting in elementary school as part of his education reform platform)
Or consider the debates about ethics in self driving cars; should they swerve to hit the aged nun or continue off the cliff? Great for journalist thinkpieces, but completely irrelevant to the ways that self driving cars are actually programmed.
Heinlein recommends getting close to lots of things: "A human being should be able to change a diaper, plan an invasion, butcher a hog, conn a ship, design a building, write a sonnet, balance accounts, build a wall, set a bone, comfort the dying, take orders, give orders, cooperate, act alone, solve equations, analyse a new problem, pitch manure, program a computer, cook a tasty meal, fight efficiently, die gallantly. Specialization is for insects."
But I find it's also important within narrow specialties; I've seen people argue about a flaw in a piece of hardware in endless meetings without ever having laid eyes on one. When I started carrying one around to all the meetings, the problem was understood much faster. Or when I've gone off daydreaming for days about a potentially difficult part of a problem, only later actually making an attempt at a solution and finding the hard parts are elsewhere.
I'd say the effect is roughly logarithmic; huge gains when you go from zero encounters to one, then diminishing returns as you spend more time with a thing and become an expert in it. But it's amazing how long people will go before getting those first few encounters.
Re: the CS/AI/stats thing - I've tended to notice that something similar applies in psephology (study of elections). Nate Silver was primarily a statistician, not a political scientist.
Over here in Australia, our best-known psephologist (employed by our public broadcaster, the ABC to analyse redistricting and provide election coverage) was primarily a computer scientist who went into election research on a whim:
Another well known election analyst in Aus was primarily an invertebrate ecologist (Kevin Bonham).
One who used to write a lot about psephology and polls (now moved onto COVID-19 stuff) graduated in econometrics (Bryan Palmer, pseudonym Mark Graph).
One who used to have an active pseph blog but is now mostly on Twitter (Pollytics/Possum Comitatus, aka Scott Steel) was primarily an economist.
I know at least one pseudonymous pseph who was primarily a philosopher (unable to reveal which, it's in confidence).
Of the remaining well known ones, I think three are primarily political scientists (Peter Brent, Ben Raue and Malcolm Mackerras). So that's a 4-3 in favour of non-polisci to polisci (or 5-3, if you count the pseudonymous one) amongst Australian psephologists.
I think there's probably something - maybe the uncertainty, maybe the politics, maybe the status - about fields like election analysis and live pandemic epidemiology which attracts people from other fields to come and contribute.
My feeling is that, if the validity of a study hinges upon the right way to "crenulate the zeugma", then it is very likely that the study is too tentative and/or the effect is too weak, and should not be used as the basis for public policy.
>Whenever I talked to Aussies or New Zealanders, they just really wanted to stress that they had ascended beyond such primitive mortal concerns, defeated the virus their own way, and were somewhat annoyed that the rest of us were squabbling about the relative merits of our inferior plans rather than focusing on how great they were. Sorry, guys.
We Aussies boned up the initial phase as badly as anyone else (refusing to close all borders long after it had spread beyond the PRC), we were just lucky it came here late. An amount of the rest I am unable to quantify is downstream of that.
Re the emotional damage, I find it striking that you're overlooking the biggest source of emotional anti-utilities on the pro-lockdown side. The pain of loosing a loved one years before they would otherwise have died.
I don’t get your comment about Aus/NZ. The key tool we used once cases got too many to contact trace was lockdowns. We used them repeatedly and they worked every single time. A few times we tried to avoid them and things got out of control. See Melbourne last winter and Sydney now. It seems very solid evidence that lock downs work extremely well.
Isn't it fascinating how conservatives, usually the ones more concerned about hygiene and infections, threw all of that overboard without blinking in this particular case, when the normal tendency would have been unusually useful to stay with?
I'm surprised nobody has touched what I thought was the most interesting part of your post - the bit where you find two opposite impossible-to-measure outcomes completely overdetermine your complex model and any work you do refining anything other than these parameters is completely swallowed up. I built a similar model to you for lockdown and found the same result you did - your assumption around the emotional impact of lockdown dwarfed everything else you might assume about the disease, including deaths and the emotional impact of deaths (actually I found that if it seriously and permanently harmed productivity / capita by knocking people permanently into an unemployment trap then this was plausibly of the same magnitude(ish) as the emotional effects of lockdown if you assumed weak emotional effects, but emotional effects were still a big deal)
At the risk of raising yet another highly controversial topic, I tried to do what you did for lockdown with Brexit. I included maybe 20 parameters in my model, and then discovered that the 'will it increase or decrease economic growth over ten years relative to no Brexit' was literally the only parameter that made any difference whatsoever; for any reasonably set of preferences a difference in the answer to this question of 0.01% either way was greater than literally any possible combination of outcomes to the other 19 parameters.
Therefore I don't think this is an isolated example, and it seems likely that most big political decisions have this character (on a sample of n=2, but a couple of other rationality tropes are similar - x-risk, cryonics, EA and so on). I really think this must be an understudied problem in decision making, because otherwise it would be all anyone talked about when making decisions.
For the record, I thought it was clear your conclusion was that lockdowns weren't worth the emotional cost, but I guess putting it in bold, TLDRing, and adding a witty partisan headline would have helped.
"This question was too multi-dimensional. As in, you could calculate everything right according to some model, and then someone else could say "but actually none of that matters, the real issue is X", and you would have a hard time proving it wasn't."
Scott, I hope you will never let something like this discourage you from treating a subject as multi-dimensional. It's easy to underestimate how different the utility functions of different people are. Some people will not care much if they (or others) die from Covid, but find the risks of LongCovid terrifying. It is quite tempting to say "they don't understand what they are talking about" in such cases, but I think that we should really not give into that temptation.
Giving a multi-dimensional treatment helps all readers, not just those who happen to have the same utility function as yourself. I remember your post "moral cost of chicken vs. beef". It convinced me that we have a strong moral obligation to replace beef with chicken. This was exactly the opposite conclusion from yours! (I had this opinion before reading your post and already acted on it, but your post made me much more certain about it.) This is just because we have different value functions.
I am not discouraging you from giving your opinion and conclusions. But I would be careful leaving things out because they seem not so relevant. (But obviously, there are limits to this.)
On your liberal vs. libertarian question, dont you think that in Iran the far majority of psychologists would make arguments for the positive psychological effects of Niqab? If then the one secular dude, who writes about Russell and Ataturk all the time, all of a sudden is the only guy to find evidence to the contrary, is that suspect?
"(...) To keep all populations outside the minority elite under an almost permanent state of oppression and make them love their servitude, (...) the main tools used to achieve that were some vaccines which altered the functions of the brain and the medicines that the State forced the population to consume. In Well’s opinion, this was not a conspiracy, but rather a ‘one-world brain’ which would function as a police of the mind."
"that's just a coincidence - if we had been in a world where academia leaned conservative"
There is a widespread liberal argument that this isn't a coincidence, but that academia leans liberal because liberal positions are more often factually correct and conservative positions are more often factually wrong.
As a fully-fledged member of the Blue Tribe, I can't possibly come to any neutral judgment on this, but it does seem to be worth consideration.
Re: discounting a conservative/libertarian's study against lockdowns and not discounting a presumed-liberals study in favour of lockdowns - it's not as simple as just "study aligns with beliefs". There are a number of factors all pointing in the same direction.
It seems obvious on the face of it that *of course lockdowns work to some degree*, and the question is just how much. It would actually be pretty bizarre if lockdowns didn't attenuate the transmission of an airborne agent. It would be like finding a study that flossing didn't prevent tooth decay - it seems pretty obvious that it must, so we will naturally be skeptical of the result.
There is a general consensus that lockdowns work, so they have to provide enough evidence to show they're right and everyone else is wrong.
It's also true that the "conservative/libertarian" position on many science-related issues is patently false/absurd. Climate change isn't real, for instance.
So it's not one pointer like it seems, it's a few which conspire. It's "study aligns with prior beliefs" (possible bias to find desired result), it's "group is known for lying about science" (low prior credibility for members), "result contradicts obvious likelihood" (higher burden of proof), and "study at odds with general scientific consensus" (higher burden of proof).
The liberal, pro-lockdown study in the hypothetical case of a researcher who's equally biased only has to contend with "study aligns with prior beliefs", and in fact any doubt is *attenuated* by "group is known for trusting science", "result in line with obvious likelihood", and "study concurs with general consensus".
Of course, the study can be good or bad quite apart from all of that, but I think that's the explanation for why "conservative/libertarian finds evidence lockdowns are bad" seems on the face of it worse than the liberal pro-lockdown study without reading either. It applies more or less equally to the pro-gun/anti-gun position, or a number of other issues split along red tribe/blue tribe lines.
If it was extremely difficult for a well-connected, intelligent, and thoughtful person such as yourself to analyze the pros and cons with the benefit of a full year of hindsight, imagine how hard it was for the public health authorities, and government, to do so last spring and summer.
That is, I think two additional conclusions you might reasonably draw are:
(1) It's really REALLY hard to get public health measures bang-on correct, when you are looking forward and dealing with an infectious virus, the human immune system, and dynamically responsive human populations in an ideologically, economically, and intellectually diverse population of 300 million. Perhaps "didn't utterly screw it up" is about the highest standard of success we can reasonably expect, and we should critique our public health authorities and government appropriately*.
(2) Maybe a lot of governing of 300 million people with a view towards eliminating bad luck and improving everyone's lot in subtle and difficult to measure ways -- like their emotional well-being! -- is incredibly difficult to get right, and our default expectations about that should be deeply skeptical, no matter how persuasive the hypotheses are -- and we should craft our political institutions with an eye to that skepticism.
---------------
* It isn't cost-free to have expectations for government that are much higher than it can reasonably achieve, because when it routinely falls far below expectations, that leads to broad cynicism about even the things it might be able to do well *and* it leads politicians to start lying a great deal more, on the grounds that it might actually turn out to be easier to obfuscate the data to avoid the lie's exposure than it is to get genuine results done that are good enough to win re-election. I think we see both these trends in politics today, and I think both can at least in part be traced back to unreasonably high expectations among voters about what government can actually be reasonably expecte to get done and get done right.
I wonder why emails were more helpful than comments on the post. I have a couple of hypotheses.
1. Commenting may be a form of status signaling, in which you align yourself with one side and bash the other. Hence, it is less prone to neutrality than an email.
2. Emails are generally longer and more detailed than comments, and hence can be more helpful/insightful.
3. Researchers in the field, who on average would have more informed and accurate things to say, may be reluctant to discuss their stands on a public forum, and hence emailed Scott directly.
Although 3 seems to be the most important reason, I wonder if 1 and 2 played a part at all.
I wonder if the point about "300 million people going to the bar" is valid here.
Imagine that there are 1000 billion creatures in the universe. If the pain of the death of one person is is roughly equal to the inconvenience caused by stopping 300 million people from going to the bar, then this calculus can be extrapolated to "death of one person= slight ticklish sensation for a second to 1000 billion people". Can we really say that the two are morally equal?
I think pain is mostly personal, and cannot be "added up" across persons.
I doubt that pro-lockdown authors are facing a lot of heckling and anti-lockdown authors are facing little. Rather, concern about a deadly viral pandemic and concern about how strangers will treat you on the street/internet are probably correlated. Both have ambiguous and unclear yet occasionally fatal or life-ruining danger, for which the main defense is to stay inside.
Also, the question of whether a conservative producing conservative science or a liberal producing liberal science, etc, can be trusted, I suspect you have to ask this on a case-by-case basis, and I don't think that any one ideology is more trustworthy. Find a few scientists who seem to be particularly honest, credible, diligent, listen to what they say and ignore everyone else.
For me, the huge issue with lockdown is not emotional damage, but something related : it's establishing a precedent for a central power to monitor and control personal behaviors and limit essential freedoms to achieve difficult to measure societal benefits... This is a very bad precedent, like anti terrorist measures but even worse. The west looks a lot like China did 5 years ago, when all progressive media were telling how Orwellian it was. The same media now preach safety above everything else without worrying much about citizen tracking and restrictions...we are on a very bad slope here, and we sleep faster and faster regardless of the actual epidemy strength... Which make the epidemy looks more like a convenient excuse...
About math and models, more specifically predictive models:
I think worth mentioning that math is not the hard part in predictive models. You make your hypotheses, you set your parameters, you do the math, you read the results. The math here is often simple. Or maybe it is not, if you want a beautiful solution as closed-form expressions; but we do not need that to read the results, we only need an accurate enough numeric value. The models are usually a set of differential equations or some kind of discrete dynamic system: our everyday computers can solve most of them numerically in the matter of seconds.
The hard part, really, are the hypotheses and parameters. About parameters, John von Neumann put it this way: “With four parameters I can fit an elephant, and with five I can make him wiggle his trunk.”
But the real stopper is hypotheses. Models have a boatload of them, most of them implicit.
For example, when people were about computing the herd immunity threshold: the main parameter was the average number of people anybody may contaminate. Key word: average. Introverts will meet fewer people, with a lower risk of contamination, the opposite for extroverts. Does it average out? No: common sens tells us that extroverts will be contaminated early, then immune, hence slowing the propagation. And models designed to test that common sense confirm it: random graphs where 90% of the nodes have two links and 10% have 12 have a lower herd immunity threshold than graphs where all nodes have three links.
Furthermore, some epidemiologists knew that. There were articles about it published, even before the current crisis IIRC. But when you look at the models, the average is the only distribution parameters. That translates into the implicit hypothesis “everyone's as much an extrovert as everyone else”. Which does not match reality. But since it is implicit, it is rarely discussed.
The important thing to remember about predictive models is that they can tell us what can possibly happen, but they are too many uncertainties for them to tell us what will happen.
If you can read French, or stomach an automatic translation (it should work reasonably well for that kind of text), the following blog article, and the related articles, tells it better than me:
It connects with the Euler remark: when Euler says “𝑒^𝑖π+1=0 therefore God”, he really is saying “𝑒^𝑖π+1=0 AND beauty implies God, therefore God”. We can call bullshit on the implicit hypothesis, which is not math.
I can argue that the Nazis caused the Manhattan project, which caused civil nuclear power earlier, which might allow us to barely escape the worst of climate change. So, Nazis good? Certainly not, it was not on purpose.
The conclusion I take is that History is chaotic, and worrying about hypothetical indirect consequences too far removed is a waste of time. Cannabis may decrease intelligence? Maybe, but people more relaxed may hit their kids less, making the next generation smarter.
The lockdowns, by changing our relation to work and welfare, may have opened the road to Universal Basic Income and Universal Healthcare, and caused the end of poverty before 2040. Or they may have strengthened the choke-hold of tech corporations on society and led us to the standard cyberpunk dystopia.
I do not have a strong argument for it, but my intuition tells me that we need to assume that the indirect chaotic effects average to 0, because otherwise we cannot do anything. It is not very satisfying; but if psychohistory was easy, it would have been invented already.
You certainly remember that after 9/11, authorities have pushed for security theater, widening the powers of the executive branch and law enforcement, and it happened again after each episode of terrorism. For example, the Crypto Wars were re-kindled by the need to decrypt a terrorist's phone.
The phenomenon is simple: power attracts people who want power, if they can get more of it they will, and events that scare the population are a good occasion to consolidate power.
As you pointed out, lockdowns in Europe were probably more strict than in the United States. To implement them, governments had to have emergency laws voted. When the emergency subsided, some of these laws ended, and some of them have been kept and made permanent.
One of the concern I and many other people had about the lockdowns were exactly this: they being the excuse to weaken legal protections for fundamental rights, with a good measure of ratchet effect: once the population of an area had been locked down under police control to save lives from a pandemic, and society has not collapsed, we can be locked down under police control again for other reasons, say a terrorist manhunt, or protests against an unpopular reform of tax pensions.
It was especially visible in France, where the implementation of the lockdown were quite stupid, to the point that German newspapers called us Authoritarian Absurdistan. Parks and forests were forbidden. To go to groceries, we had to fill and sign a form for ourselves; police chiefs have bragged how many people they have fined at the very doors of convenience stores. And whoever complained about it were swiftly called covid denialists and compared to Trumpist anti-maskers.
The weakening of fundamental freedoms and checks-and-balances, both in law and in mentalities, is a long-term consequence of the lockdowns that I fear might feel deeply.
We are carrying out a real-time experiment on this in Australia as we speak. Both Melbourne and Sydney are having lockdowns because of delta outbreaks. Melbourne lockdown was very quick and strict, Sydney hesitated. As a result, Melbourne looks very close to proving conclusively that it is possible for a 5 million people city to control delta and (hopefully, with a little luck) bring it down to 0. Sydney at this point seems to be losing control.
The difference between the 2 cities is the ideology of the state governments, Labor in Melbourne, right-wing Liberal in Sydney.
The utility arithmetic of emotional consequences doesn't sit right with me...
I don't think you can assign a negligible negative utility to some experience, then just naively multiply that utility by the number of people having that experience to create some huge negative utility.
For instance: If a genie would let me save one person's life on the condition that a hundred people had to stub their toes, I would take that deal in a heartbeat. My decision doesn't change if it's a million people. Or a billion. Or a trillion.
Somehow, my intuition is that if a utility is small enough, it doesn't simply add up over several people.
Similarly, I *don't care* if a billion people are sad about not getting to go to night clubs for a year, if that's the price to pay for saving lives. For the same reason that I would happily sacrifice a billion servings of caviar to stop covid. "Oh no I don't get to eat caviar" times a billion. Get over yourself you spoiled brat. Night clubs are a luxury, and you should be happy to give up that luxury for the opportunity to save lives.
Maybe write a post about the emotional damage subject specifically? You're right that it would be transgressive. Anyone who brings it up is in for some shaming as being selfishly concerned with their own comfort "while people are dying!"
These kinds of calculations always subtly ignore the value of simply not being forced to do something. You don't have to be a raging libertarian to realize that "mere" personal freedom has some real moral value. Not being forced to do something is an inherent good.
That's why I think it's much better to do trade-off-type thought experiments to judge the rationality of lockdowns. Find some phenomena, X, that roughly matches the distribution of mortality risk as covid (plus some padding for the "unknowns", if you want), and then 1) see how much effort people currently expend avoiding such risks and then compare it to relative hardship of lockdowns, and/or 2) ask people if they'd vote to lockdown to avoid the risks of X. (Of course, when polling you have to avoid mentioning covid, lest you taint it with politics.)
I feel like your analysis had an asymmetry, where you acknowledge that R reduced mainly because of voluntary behavior changes rather than lockdowns, but then attribute all the costs to lockdowns when most of the costs were also caused by voluntary behavior rather than government policy.
I feel dumber for having read this. Sweden, Afganistan, other places that simply ignored covid, provide irrefutable proof that covid while a real cold/flu like disease, us no more dangerous than flu. There are no excess deaths, or the effect is very small. Death rates are roughly same as 2017, 2018
Any kind of statistics other than overall amount of people dying, are meaningless. US gov got initially freaked out by COVID, and threw tons if money at it. Where there is free money, there is grifting.
Situation: 10 grandmas walk into ER, presenting fever, headache and coughing. These are symptoms of both flu, and covid. But treating COVID patient milks insurance for 100k, while treating flu patient hardly milks 20k. Do you think our hospitals are going to classify these 10 grandmas with flu, or with covid? "But but but the PCR test show covid". Do you undetstand how PCR test works? Its exponential magnification. With 45 cycles you will detect covid on a ham sandwich with 50% probability. With 50 cycles every test on any substance will come out as covid.
My question is: Why isn't anyone else doing this specific analysis? You would think that governments, universities, etc. would be motivated to do this type of analysis. (Maybe they are, and I just don't read the right blogs to see those posts?)
As a statistician, I'm very skeptical of people who crenulate their zeugmas. Just because you have some asymptotic zuegmoyal theory doesn't mean that it applies well to finite samples. If you're going to be a Crenulist, at least verify that Epstein's theory holds in your data set!
So I guess you could say it’s complicated.
Yeah, I know. I’m a chronic smart ass
It was an enjoyable read though. You would have had to make it a lot more boring and a lot longer to get away with tl;dr
> One thing I found helpful was to ask Person A what they thought of Person B's work, relay it to Person B, Person B would say something like "he's neglecting to consider that you need to fulminate the synecdoche", then I would relay that back to Person A, and after enough of this I would get a meaningful sense of where they disagreed (or occasionally one of them would just admit they had made a mistake).
This is *EXACTLY* the model I've found myself stumbling into working on my Georgism follow-up. I've got all the Georgists on one side telling me everything checks out, and then a bunch of skeptics on the other, and all of them have impeccable credentials and PhD's in econ or tons of professional experience, and I'm just a guy who read a book and wrote about it on the internet that one time.
So far the above method seems to work. Speaking of:
@Erusian, if you're around I'd like to get in touch off-site so I can send you some of the best arguments about the practicalities of land assessment so I can start the claim->response->counter-claim->counter-response cycle, if you're up for it. Email me at lars dot doucet at gmail dot com if you're game.
Many many years from now the verdict will be that lockdown effectiveness was not worth the negative externalities inherent to the lockdown, regardless of the marginal benefit some lockdown statistics have shown. I know that no one will, but feel free to screencap this and throw it back in my face at some long distant future date, so long as you acknowledge my ability to do the same.
I think there is a substantial difference between someone who leans liberal or leans conservative and someone who is an ideologue. If someone like Tyler Cowen - who leans libertarian/conservative - comes out in favor of lockdowns that voice carries more weight with me. Alternatively Freddie DeBoer's writings on human ability, testing and education in general carry more weight with me as he is from the very far left. But in both cases they are calling them as they see them. They are not insisting on finding facts that fit their existing ideology. Or if they are, when the facts don't support their ideology they are open to saying, "My theories don't work in this instance."
As far as whether a conservative prognostician arguing against lockdowns is more "marked" than a liberal one arguing in favor, I don't think it's necessarily determinative, but I do think it's relevant that under ordinary circumstances, nobody favors lockdowns at all. It's not that lockdowns in general are a liberal position which you'd expect liberal people to be in support of unless the evidence weighed really heavily against it, it's only due to extraordinary circumstances that anyone supports lockdowns.
This is a distinctly America-centric perspective, but I think you could tell a similar story about other countries, where conservatives are biased against taking the pandemic seriously, not because this is an inherently conservative position, but because an important figure, Trump in America's case, took a stand on downplaying the pandemic. If Bob Woodward can be taken at his word on the subject, Trump's efforts at downplaying it don't even reflect his own best understanding of the pandemic at the time, since he thought having people take it less seriously was better for the economy, and likely him by extension. If we had had a different Republican politician in office, someone like George W. Bush perhaps, then taking the pandmic seriously might never have become a tribal signifier in the first place.
The upshot seems to be that human societies are complicated, social science is insanely hard, politics is even harder, and even the smartest and most honest politician has to rely on a combination of judgment, intuition, and common sense to implement their policies. That's why open debate and epistemic humility are important.
Yeah, most of the time people support their own side's ideological positions. Unless you know everything about crenulating zeugmas you're going to have to either A) if orthodox, discount because orthodox ideologues are sometimes sloppy due to most of their peers being relatively uncritical towards the work, or B) if heterodox, discount because it could be a clever arguer trying to sell you a box that you do not want.
Of course, you could be dealing with a clever arguer also in case A, so maybe you should discount orthodox arguments slightly more anyway. I have no idea, this problem stumps me. I feel you.
At least this was a good reason to post that cute bat again. More covid posts, please.
One issue I have about emotional costs is the comparison point. It's a global pandemic, an intrinsically "bad" event, and life is going to be less "good" than before. There's going to be long term trauma about how poorly it was perceived to be handled and how ill prepared we were for it. Some of the measurements of quality of life are good at showing the impact of the pandemic, but not the cost effectiveness of lockdowns.
> One thing I found helpful was to ask Person A what they thought of Person B's work, relay it to Person B, Person B would say something like "he's neglecting to consider that you need to fulminate the synecdoche", then I would relay that back to Person A, and after enough of this I would get a meaningful sense of where they disagreed (or occasionally one of them would just admit they had made a mistake).
It sounds like you kind of rediscovered the Delphi method? RAND came up with a consensus-building framework specifically for opposed experts back in the 60's and it seems to work *really* well. I wish all political and economic arguments were given that treatment. Original article here:
https://www.rand.org/pubs/papers/P3558.html
To be honest, I found this post much more interesting than the post it is about.
I have been trying to get this across to everyone who will listen, for over a year now, and I haven't been successful but I can't stop because it's extremely important to my continued survival:
The question "do lockdowns work?" is incoherent and unanswerable, because both "lockdown" and "work" are and will always be underspecified.
You can't aggregate over millions of people, summarize death counts or economic costs or whatever, generate an average, and use that number to judge things. As you point out above, this question is much too multifaceted to do that
But I'll make a stronger claim: you can't even do this in principle. One might be tempted to say "aha, well, maybe there's 50 dimensions but we can analyze all of them and come to conclusions" but you can't! Because people have varied lives and, no matter what conclusion you will come to, somebody somewhere will say "if we did it your way, I would not be alive right now. Eff you" and there is absolutely nothing you can say to that person to change their mind.
There is no way in principle to settle this argument. It is a bona fide conflict, and the way it gets settled is the guy who gets to decide, decides, and then says "tough if you don't like it". That is in fact what our society did last year, and I didn't like it.
All of the mainstream, public scientific discussion about the pandemic is bullshit. It just is. It is not the case that a whole bunch of honest, concerned, and curious scientists all decided to dive in and see what happens. I know we here like to think that's the case, because we here actually do do that. But the rest of the world does not. The rest of the world does the science they're paid to do. If it gives the results that the people in power want, then they get published and publicized. If it does not, they do not. As a result, you end up with a bunch of 'science' supporting what the people in power have already decided they want to do, and the people in power use that 'science' to wash their hands of any responsibility for people being mad at them.
You can see this happen in real time if you pay attention. This is a relatively trivial example, but it is very accessible: Dr Fauci's overnight 180 on masks, and everyone acting like this didn't happen (but was totally fine if it did). Prior to his flip, Fauci and the rest of the medico-political establishment told us that The Science Says Don't Use Masks, and the news dutifully reported that. Then, overnight, Fauci says that The Science Says Use Masks, and the news dutifully reported on that. When questioned on it, all they say is "we're just following the science, and the science evolved". But we know that's a lie. Why? Because Fauci said it was a lie under oath to congress.
In reality, the actual fact of the matter, which I expect most of you here either knew or just assumed, was political. They were afraid of a run on masks, which would put medical practitioners at risk, and so they told the public not to wear masks. Then, when they were satisfied that the medical practitioners were well supplied, like magic, 'the science changed'. In reality, the government gave out an arbitrary order ("do not buy masks") in order to achieve a domestic policy goal ("don't let hospitals run out of PPE"), and instead of just saying that, they lied to our faces and bastardized 'science'.
In fact (I'm not gonna go look this up, so apologies if my memory is wrong), I think that was the conclusion you came to in your original mask post. The public health advice prior to covid was "do not wear masks", and the reasons for this was a) people wear them wrong and it wastes them; and b) that takes them away from people who actually need them.
What is the point of my work-procrastinating rambling? My point is that fundamentally you cannot use 'science' to determine if 'lockdowns work'. The overwhelming majority of the 'science' on this is science in nothing but name. If you do _actual_ science to try and get an _actual_ answer, everyone will ignore that answer with some variation of "yeah but no because thing I care about". You can ask extraordinarily specific questions like "did imposing a 8pm curfew reduce the number of deaths by 500 or more". Even that isn't really a great question, because imposing a curfew doesn't mean people comply with it, and so it equivocates the question of "does adhering to a curfew work" with the question of "does imposing a curfew _policy_ work" which, despite appearances, are two very different questions.
Maybe, if we constrain ourselves the platonic world of abstract thought, that cannot interact with the outside world in any way, we could come up with The Correct Answer to the question of lockdowns. But that's not particularly useful. Implied in the question of "do lockdowns work", is the hidden question "should we implement lockdowns". But as is abundantly clear by now, the decision process we use to implement lockdowns is completely disconnected from the actual answer to the question of whether or not they work. The whole question is poorly defined, and the answer doesn't exist
TL;DR version: lockdowns: good/bad? - is a waste of time that we could be spending on a more important question: how to persuade people to demand a better overall pandemic response (and prevention) in the future. Your post about it was great and fascinating and I loved reading it, but I will straight up paypal you money today if you produce one as good about "finding the best way to demand better next time."
Long version:
I think this is the key point: "But the smartest people I talked to kept - is "derailing" the right word? - derailing onto more interesting and important pull-the-rope-sideways plans. "
And:
"none of that turned out to matter at all compared to some question about stoned driving vs. substituting-for-drunk-driving, which nobody started out caring about."
Because that's exactly the truth - lockdowns yes/no is a very interesting (and extremely difficult) question to answer. But really, for the question of "how should we fight pandemics" it doesn't matter that much - there are other interventions that will have a massively larger effect (and in fact will make doing the research to determine if lockdowns work really hard, as you discovered) that this is like trying to figure out if you should amputate the left or right leg of the guy who got run over by the wheat thresher. The real answer, for society, is taking precautions to avoid the "got run over" scenario, or "let a pandemic get so bad you are debating lockdowns" scenario. The various "Titanic deck chair rearrangement" plans don't move the needle very much.
In fact, given the pandemic's ability to inspire Americans to polarize even the most mundane things, I think that the true answer to your question: "lockdowns? yes/no?" is actually "I think you should stop asking this question." I say this is as a libertarian, first amendment maximalist who has indeed been following this issue and happily debating it with others for a year and a half: this is a rabbit hole that doesn't have... whatever rewards rabbit holes are supposed to have at the bottom of them.
I don't mean to diminish the importance or quality of your work (or anyone else your reference), but for the goal of having the public know what good pandemic response is, this just doesn't help. (I intend no moral or judgement criticism here AT ALL) It is so hard to isolate, so amenable to data cheating, mistakes and incredibly confounding errors (maybe lockdowns work in areas where people believe they work and therefore cooperate, but don't in places where they hate that kind of thing?) that time spent worrying about it just convinces the average person that the question DOES matter, and that other questions (that really do matter more) don't. That is, those people who tried to derail you are exactly right: every minute that you (and all the researchers you cited) spent trying to figure out the truth of this question was time that could have been spent mobilizing public demand for the truly effective pandemic solutions (which surprisingly seem to be far less controversial, and easier to prove they work in practice), for a net benefit.
By any objective, scientific - and indeed, conventional-wisdom-approved measure, there are lots of things (again, many uncontroversial and agreed upon) we could have done better on the pandemic. Politicians and bureaucrats from both parties (or even unaligned) flubbed it to previously unimaginable levels (I used to think the CDC was great!). So the fact that we have no mass movement to Do Better Next Time (and stunningly little promise that any changes have been made, or that anyone at all was held accountable for failures) is shocking, and trying to fix that seems like a much more fruitful use of (anyone's) time.
If someone could figure that out, I would love it. I would love it so much that I would send them money right now to work on it.
It's interesting to see the question of "how much more do you trust someone arguing against their typical ideological bias" in a very different context than I'm used to seeing it in. I work in a corporate business environment, and the analogy my boss likes to use is "PETA vs. the butcher", in that those are two people which you would expect to have very different opinions on the sanctity of animal life.
So yeah, assigning additional credibility towards people arguing against their typical positions is meaningful - but exactly how much additional credibility to give something, and how much more seriously to take the argument, is a very interesting question. On first glance it's the sort of thing that seems unavoidably subjective; every individual arguing against their ideological bent is going to have a varying degree of attachment to their "tribe" and therefore the emotional or social costs of believing things that go against their prior attachments - and thus the amount of additional 'virtue' we would assign them for 'having done science correctly' - would similarly seem to vary on a case by case basis.
Do you award extra 'virtue points' if the head of the NRA argues for gun control in a specific scenario, vs someone who's just a member, vs someone who's just Republican?
What about the opposite scenario, where the head of the NRA argues for total lack of gun control? My statistics background says "this is the null hypothesis, there's no additional information to be gleaned here" but is there ever a point where someone is so ideologically committed to certain positions that even otherwise good-seeming arguments that they are making should be discounted or taken with a grain of salt? Maybe if they're trying to sell you something?
An interesting question that I'm going to be thinking about!
Interesting post. Three comments:
1. What you describe doing in representing A to B and then B to A corresponds pretty closely to the job of paidhi in C.J.Cherryh's (very good) sf Foreigner series. The paidhi first acts as A's agent in representing his position to B, then switches roles and acts as B's agent in representing his position to A.
2. Your experience trying to figure out the net effect of either lockdowns or Marijuana legalization is similar to mine looking at the effect of climate change. There are a whole lot of possible effects, positive and negative, and what one mostly sees is a calculation based on a subset biased towards the result the person calculating wants or expects. Nordhaus gives a lot of weight to very speculative calculations on low probability high cost effects of climate change but, so far as I can tell, never considers low probability high cost effects of preventing climate change — even though the obvious one, the end of the current interglacial, is a climate effect that has happened multiple times over the last few million years.
There is one way of including emotional effects in the calculation — leaving individuals free to decide for themselves whether to quarantine. They have the right incentive with regard to emotional costs, since they are the ones bearing them. It's an imperfect solution because they have too weak an incentive to take precautions — if they get Covid, much of the cost goes to them but much to other people to whom they might pass it. And all of the cost if they already have it and don't take precautions goes to others.
My family in fact quarantined more tightly than the rules require until we were vaccinated.
On your problem with fancy statistics ... . I tried to follow the controversy over the Lott and Mustard paper on the effect of concealed carry, eventually stopped when the statistical arguments got above my level. Like you I have taught low level statistics and, while my mathematical background is much stronger than yours, my statistical background probably is not.
"Apparently pro-lockdown academics who get too close to the public spotlight have been getting harassed by lockdown opponents, and this is a known problem that pro-lockdown academics are well aware of. I was depressed to hear that, though in retrospect it makes sense."
I'm very new to this blog. I love the smart writing. But man do I wonder sometimes what alternate universe the author lives in to not know this stuff has been going on for a year and a half now. Federal, state, county and city health directors have received death threats, with armed protestors sometimes showing up at their house. The health director instantly becomes the most reviled person in their region. Many have quit over it.
One interesting question you can ask someone is what burden of proof would they require to admit they are wrong.
Do you really believe that the mathematician with his zeugmas, crenelated or not, started off with a totally unbiased mind, a tabula-rasa, and let the math lead him to a conclusion? Or do you believe that he started off with some emotion, instinct-driven position and found the data and the math to justify it?
You can’t perform any meaningful post-facto analysis of the pluses and minuses of lockdowns because you can’t perform a test with one America locked down and one not locked down. Comparing states is a worthless exercise – the social, economic and cultural differences between, say, California and Alabama, are simply too great. The researchers who have tried to do that just prove my contention: they begin with a belief and try to find the data to support it, ignoring the fatal limitations of the exercise.
The same is true with data for marijuana usage: we know with some degree of confidence the typical health and behavioral impacts of marijuana usage, but we cannot possibly forecast the net benefits\harms caused to American society at large, and all those who pretend we can and that they have the right methodology, have started out intending to prove their instinctive belief (even when they’re adamant they haven’t) and produced nothing of value.
So, what do we do when decisions have to be made, like about lockdowns? Well, we start off with what we know for sure:
Covid is an infectious disease spread by contact. Covid is potentially fatal or highly debilitating. We have very limited facilities for treating patients with serious cases of Covid. People will respond to images of hospitals denying admission and people dying at home in great distress by self-locking down. Such ad-hoc quarantining will be chaotic and lack legal protections, and will cause great economic damage.
We don’t need Zeugmas to be sure of any of this. All we need to do is to confidently state the common-sense facts, draw the obvious conclusions and act accordingly. For the most part, that, thankfully, is what we did.
One description of why the unknown unknowns can be high impact and unpredictable:
Suppose the US had absolutely no lockdowns. This would probably make the population dislike the incumbent Republican government more, which would probably lead to the Democrats win with a huge majority in the election. Whether this is good or bad is up to your interpretation, but it could plausibly have an impact much larger than the marginal impact of the lockdowns.
On the other side, suppose the US had stronger lockdowns with more bipartisan support from the federal government. This might lead to the incumbent Republican party staying in power in the election, especially if the stricter lockdowns had a large impact on keeping case counts down. Similarly, this could also plausibly have an impact much larger than the marginal impact of the lockdowns.
Of course, the assumptions behind both of those scenarios could be incorrect, but they're at least *plausible*, making the whole situation impossible to evaluate to a binary yes/no.
"If a professor comes up with some study that shows guns decrease crime, and that professor is a gun owner, goes hunting every day, and donates to conservative organizations, is that suspicious? What if another professor shows that guns increase crime, and this professor has never owned a gun in her life, hates hunting, and donates to liberal organizations? It feels like "enthusiastic gun owner" is more of a "marked" group than "non-gun-owner", but that's just a coincidence - if we had been in a world where academia leaned conservative and most professors owned guns, it would be the opposite. But should we really discount the fact that the pro-gun study professor has an NRA bumper sticker on his car? I’m still not sure how to think about this."
This strikes me as the fundamental crisis both of rationality and our current moment. My sense is that most people have already surrendered to the view that political or class alignment overshadows all evidence of unbiased analysis or opinion. And its not like there aren't plenty of real examples of this! But if we can't find any way past this problem we are probably screwed.
When there is enough uncertainty, model outcomes reflect the biases of the author a surprising amount of the time. I'd do a study on this but I'm biased that this is clearly true.
Here is some advice, if you don't like questionable statistical manipulation then never look closely at some areas of climate science. This is particularly true for the ones that make it through the media's new and alarming results filter.
A lot of statistical manipulation can be a sign of a very difficult problem, it can also be a sign that the less complex models weren't giving answers the author liked. For some problems where there were literally new statistical methods invented for analysis it would be nice to see a progression of different model outcomes and how the results were affected. Sometimes raw data is just crap, and you can't polish it no matter how hard you try.
I'm very surprised and sorry that <i>"it just felt too weird and transgressive to focus on something authorities weren't even talking about."</i> That's one of the biggest reasons I started following your writing, Scott - you've been willing to focus on weird but important aspects of situations, even when authorities aren't talking about them!
The toughest thing about this topic for me is the fact that for many conservative places (like Indiana where I live), "lockdown" was largely hypothetical! Yes, stuff closed for a bit, but for most of COVID you could go most places and largely wouldn't even be asked to wear a mask.
How do you study the effects of a mask "mandate" (read: suggestion) that more than half the population simply ignored? Surveys seem like a very thin way of learning things like "a quarter of the people at the grocery store don't wear masks" and "lots of people never stopped getting together at all."
I could tell that paper was exhausting to write, and thank you for going through with it, even if the results weren't quite what you wanted in terms of clarity.
By chance my RSS feed today turned up https://ourworldindata.org/child-mortality-big-problem-in-brief
Quick back of an envelope gives me that if the whole world was as prosperous as W. Europe (where I am) that would save ~3m child deaths/ annum
3m/ annum is roughly the C19 death rate over the last year. I have no idea what would have happened if we'd done nothing and carried on. If we'd doubled the death rate but the economy had carried on growing normally we'd be at break even.
The correlation between wealth (GDP/capita) and longevity, both in life expectancy and child mortality, is overwhelming. I saw some talk about this in 'phase 2' of the conversation last year but it seems to have been forgotten
"Other times I just went for maximally stupid models and saw what happened"
I think you're probably being too harsh on this as a good judgment method, at least in this case. Lockdowns are, in general, very 'maximal' policies; they are not "adding a penny tax to drinks at the bar" or "restrict advertisements for tourism" which are the much more typical policies you might see a government introduce. They are a great cudgel of a policy.
As you state, this is helpful in bounding things, but given the political interests, and the immense difficulty of teasing out impacts, it should bound people to a great degree. If the broad policy of preventing interactions fails to produce broad results, we should be quite suspicious of very careful, complex models. Given the replication crisis and similar scandals, I think a skeptical instinct in these cases would prove a better judge overall.
If one was studying lockdowns, why choose red vs blue states? Why not instead choose US vs China? Because that's an emotional ouchie for many.
I would argue that the targeted comparison reveals a lot about the motives of an author. Even red and blue states aren't that much different. They each have large percentages of the opposition who behave according to their tribal tendencies regardless of government edicts. I'm no fan of China but the real question to me is whether hard core China-style lockdowns are what is needed once widespread transmission is underway, not these flip a red/blue coin and micro-analyze the results.
I live in Florida and we have been bombarded by the national media with partisan accusations constantly. No state mask mandates! Curiously my county was under mandatory mask orders for over a year, as was almost every metropolitan area in Florida. Florida left the decision making to the local areas. Florida-Man-Moron. Aaaaaghh, people on beaches! It was all so tiring and wasteful to read.
Nobody won against covid. Tabulating the losses to 50.1/49.9 is not very interesting. Open a county level national map, set it to cumulative cases/deaths per population and scroll around. The timing was different, but the cumulative results were pretty uniform. There was a big dry forest and it burned down.
"One thing some people fairly called me out on was asymmetric political bias adjustment."
Yeah, I think you are correcting in the wrong direction.
Who gets better at arguing and defending their points: Someone who is surrounded by people that agree with them, or someone who is in a hostile ideological environment?
I think if you want to be an openly conservative or libertarian academic you have to be *better* than your peers within your field. Otherwise you'll get washed out. Just like anytime you see a racial minority succeeding in a field with discrimination, they are almost always above average in that field because they are operating with a handicap.
Scott: Thanks for the effort you devoted to the article. I do not think we are near a final accounting on this issue. And, I doubt that we will ever achieve real clarity because some information that would be very good to have was never gathered, such as: We did not do random tests of healthy individuals to determine what the real rate of infection was; There were nowhere near enough autopsies performed to be able to get an accurate picture of dying with COVID vs dying from COVID.
Further we do not have a good grip on many collateral issues:
"Study: Hospitalizations for eating disorders spike among adolescents during COVID" by University of Michigan
https://medicalxpress.com/news/2021-07-hospitalizations-disorders-spike-adolescents-covid.html
"The number of adolescents admitted to the hospital for severe illness from eating disorders has increased significantly during the COVID-19 pandemic, new research suggests. At one center, the number of hospital admissions among adolescents with eating disorders more than doubled during the first 12 months of the pandemic, according to the study that appears in a pre-publication of Pediatrics. The 125 hospitalizations among patients ages 10-23 at Michigan Medicine in those 12 months reflect a significant increase over previous years, as admissions related to eating disorders during the same timeframe between 2017 and 2019 averaged 56 per year."
"Emergency Department Visits for Suspected Suicide Attempts Among Persons Aged 12–25 Years Before and During the COVID-19 Pandemic — United States, January 2019–May 2021" Weekly / June 18, 2021 / 70(24);888–894 On June 11, 2021, this report was posted online as an MMWR Early Release.
https://www.cdc.gov/mmwr/volumes/70/wr/mm7024e1.htm?s_cid=mm7024e1_w
"In May 2020, during the COVID-19 pandemic, ED visits for suspected suicide attempts began to increase among adolescents aged 12–17 years, especially girls. During February 21–March 20, 2021, suspected suicide attempt ED visits were 50.6% higher among girls aged 12–17 years than during the same period in 2019; among boys aged 12–17 years, suspected suicide attempt ED visits increased 3.7%."
"COVID and schools: the evidence for reopening safely" | 07 July 2021 | News Feature | https://www.nature.com/articles/d41586-021-01826-x
Re. Bias adjustment: I'd say it's valid, in that anyone who breaks from the consensus needs to be more harshly evaluated, even if that consensus might be political in nature; Like how market skeptic economists get raked over the coals
> For example, one person brought up that I (as a younger person) might be underestimating the emotional distress older people feel about COVID, because they (unlike me) are at serious risk of death.
Isn't it the opposite? For me (and my young friends), and lots of people around 25 (in France), the conclusion is that we sacrificed a big chunk of our mental health and youth to protect the physical health of old people.
If you have a good memory, you can play blindfolded chess against 10 GMs and score 50%. Just pair them up and let them play each other with you as a conduit. Derren Brown does this on TV with a twist (he takes 11 and defeats 1 to end up with a winning record).
Sounds like what you want to do/have done!
I don't think "emotional damage" is the right term. It wrongfully implies that the damage of being unable to do normal activities like going to a bar is somehow different from health damage. It's not. The reason we dislike being ill isn't because it gives us some abstract "bad health" state. We dislike it because it stops us doing normal things. That's what health actually means to most people. Ability to do normal activities is a critical component of QALY - a disability that has no harm beyond making you unable to go to work, unable to do recreation, and unable to see your extended family would be considered a devastating health impact, not just be called "emotional damage".
Seeing rationalists everywhere, not sure if Baader Meinhoffing: https://metaforecastings.com/about-us/
> it's so dumb that it couldn't lie even if it wanted to
Don't you mean "couldn't even tell the truth if it wanted to"? Why would you expect a dumb model to tell the truth?
Thanks for writing this one and the earlier one, Scott. There is a lot of clear thinking still to be done about the global Covid issue and moving it forward is important. Also nice technique for getting researchers to talk to each other.
This got too g-d-n long but I think I had a point.
Part of the challenge of analyzing lockdowns in the United States, and some commenters have mentioned it somewhat - I am not sure anywhere in the US "really" did a "real" lockdown. There were narrowings, yes, there were some stoppages - school! work! but the combination of society, government and policy delivered something other than "lockdown." Some call it closures, I'm sure there are other names. Some smaller locales (e.g. Indian reservation) had curfews and "only residents allowed in" policies but this was not widespread.
So, in the US, lockdown it was not, and so the dramatic mental references that attach to that word are not as pertinent. Economic damage, business damage, livelihood damage, educational damage, relationship damage, all to varying extents. We don't need the word "lockdown" to communicate that there were bad consequences. Some of the bad consequences derived from the partial nature of the restrictions; we didn't shut enough down for long enough to make enough difference, and it's possible that fewer shut-days would have been required to achieve the same outcome or better, if the shut-days and open-days were distributed differently.
I remember late Feb/early March 2020 because I had just left a job and moved to a different state. I had enough money to get by while finishing the move and for the first time in a long time I was not freaking out about 100 things. So when I'd get up in the morning and hear birds and there were no contrails and little traffic and then we'd go to the park and walk around trying to be socially distanced, it was like a gift. Which felt dirty in a way, because people were dying, suffering, struggling, but for my specific situation it felt like, ok, this is what it takes to lift the miserable rat race lifestyle off of us and let's make sure we never accept it back. For many people that lift did not occur at all but for some it did. Air quality improved dramatically where I was. My suspicion is that people who could work from home, had some conscience for their fellow man and were able to act on that by conducting isolation, and tended to left/center-left politics, may have felt that lift more than other groups. And so now for some people the word "lockdown" means "that lift." When the people with the means to do so undertook more of a "welfare of humanity" project than perhaps ever before, and it was one uniquely suited to the "personal is political" stance; making personal choices for the benefit of humanity, writ very large. It hit a lot of buttons.
That began to sour a few weeks in though when people were still going out in public but not as often and some of the social graces eroded; people who had gotten very good at pretending not to fear each other now showed furtive glances in stores, plus, what if someone coughed on me? And it had never been good for people without financial security, work-from-home, decent daycare. Then the "welfare of humanity" project shifted to a laser focus on the experience of these essential workers, who needed to be helped. Then the welfare of humanity project, which was already coming apart, became aligned with social justice in the United States and became even more "welfare of humanity." But by now "lockdown" meant "welfare of humanity" for enough people that discussing it was difficult. "Lockdown" did not mean "welfare of humanity" for everyone by any stretch; even with what we had in the US which was not lockdown, even that did not mean welfare of humanity project.
So we have to face that "welfare of humanity" project did cause suffering, did cause harm, and there are a few levels to that. The tradeoff level is tempting because it gets over with faster. But we skip the "what is harm" and "what is welfare of humanity" and "what was lockdown" definitions at our peril.
An element of the harm which I don't see mentioned much is an increase in paranoia in multiple quarters. Paranoia of government, of international politics, of other people. Maybe sublimated fear of Covid itself, maybe just looking at the policy challenges and disagreements and struggle. The "wow, they're incompetent" reaction hit people differently and I think depended somewhat on how much an individual was invested in "welfare of humanity," Some people already thought the government was incompetent and were maybe saddened but not surprised. Others found themselves in contexts where paranoia/criticism was embraced, for the first time they saw behind the Oz curtain. That had some low-information aspects though but I think it partly spawned the Jan 6 capitol situation, and this a-ha moment was pretty widespread. It became "a-ha" versus "welfare of humanity" and there we linger at the social level. Covid-era has caused wakeup calls in different ways for different people; some who had some trust in society prior to Covid looked at "welfare of humanity" and "shutdowns" and realized they were at the mercy of people they didn't trust, and reacted to that.
I think "welfare of humanity" jumped the shark in early summer 2020 but I have not yet mapped exactly how. The "a-ha" has not always reflected reality and has moved further away in fact, but to me it is less hollow, it still has fuel, and so the rabid mistrust will continue. But "welfare of humanity" euphoria causes hangover, but since it also coincides with a left goal, or many left goals, it cannot be put down either. In July 2021 now it is paranoia versus hangover, with some liberal policy accomplishments.
> mostly (completely false) claims that lockdowns cause more suicides, lockdowns devastate businesses, etc.
As far as I can tell the first is (probably) not true (though we don't really know, who knows if the toll takes a few years to set in, Tyler Cowen thinks emotional connections that were cut during the pandemic won't recover), whereas the second is probably not entirely false, at least as a first approximation?
>And it's so stupid - emotional damages! People being annoyed that they can't go to the bar
A few sentences later you write how weird it is that the libertarian people are writing angry op-eds, where did you think this anger came from? I mean "emotional damages" seems like such an obvious thing; this is obviously what people are also emotional about when they hate lockdowns; why else would they be emotional? No offense, but how is this in any way surprising?
Great post, well-written, fascinating, as always — but I'm sorry, I've been completely sidetracked by something you mentioned.
There's a chance cannabis decreases IQ? By how much is it posited to do this, and is it permanent? Because I smoked a lot of marijuana back in the day, and my IQ is the only good thing about me*, and I would like to know how much dumber I possibly made myself.
*Actually, I'm kind, handsome, charming, and humble, too, but I won't mention that stuff.
I think this is a great post simply for admitting the complexity of the situation. The only thing I've found odd about this whole thing is that nobody – yourself included as I recall – contrasted the response to this pandemic to the response to the Spanish flu a hundred years ago. Hundreds of thousands of people died in America alone, and in a very COVID-like way, yet as far as I know there were never any involuntary restrictions anywhere, people just went "well that sucks and is terrifying" and went about their normal day until the flu became nothing more than... the flu. (I've seen an Everett True cartoon where Everett canes the living shit out of a guy for not voluntarily wearing a mask, but that's about the peak for stringency of enforcement to my limited understanding.) However, I'm far from an expert on this, and I've been hoping in vain all this time for someone who does know to do a carfeul compare-and-contrast on this.
"it's so dumb that it couldn't lie even if it wanted to"
I have an aunt like that.
I really love meta posts like this. You have a really interesting way of thinking about things. I don't always agree with your specific opinions but I am always fascinated by how you think about them.
History always ends up oversimplifying stuff, because actual history is far too complex to document in detail. I expect we will eventually converge on just using total deaths per million within jurisdictions over a fixed period (not yet over) as the sole metric of success.
Thank you for both the original and meta analyses!
Re the bias question: the expert class is almost by definition devoid of poor people, young people, and uneducated people. So it doesn't seem conspiratorial to suggest that the harm done by lockdowns to those groups of people is probably underestimated in expert analysis.
Wouldn't it be strange if that were _not_ the case? This seems like a more important issue to me than specific ideological biases, like "that guy writes in Mother Jones" or "that lady owns a lot of guns."
I don't think there's been an adequate accounting of the many ways that lockdowns greatly improved a lot of people's lives, partly because it's taboo to talk about a lot of people having a nice time while others are dying, and partly because even if it was asked about, people are reluctant to admit it because of said taboo.
Yet SOOO many women I know will secretly say that 2020 was one of the best years of their life, that the lockdown was a huge relief, and that it vastly improved their physical and mental health. A lot of guys do too, but definitely way more women. Granted, I am talking about middle class people who didn't lose their jobs. They just suddenly got an extra 15 hours of their life back each week, which was previously spent in unpaid but required labor (commute, etc.). And now because of the lockdowns, they have far greater bargaining power and flexibility at work. I understand the impact was very stratified based on demographics, but I think there is a major underestimation of all the positives, because it's just not very nice to talk about how a bunch of relatively secure middle class suburban people suddenly had more enjoyable lives, while older people were scared and essential workers were having a terrible time.
I don't know if there will ever be good research on this so long as it's taboo to admit, and someone who lost a family member or job is going to scream at you and tell you what an enormous jerk you are.
The junior associates at my firm are all getting a $15k raise and the ability to work flexibility from home or reduced hours -- things that were unthinkable pre-pandemic -- just to induce them not to leave. For them, it looks to me like the whole thing net positive, and not by a small amount.
I was initially appalled by the lockdowns. I could not believe that this country is so senior-focused that we were going to sacrifice the livelihoods of young and poor people on the alter of keeping a bunch of near-death retired people safe, when they were the ones who could most easily stay home. However, I'll admit that 2020 was a fantastic year for me personally, and continues to have positive impacts on my life. Most people I know say the same.
Also, are people really as polarized on Covid as some of these comments are making it seem? Not in my experience. There is a hysterical small percent on either tail end who are get very upset, but the vast majority of people are not and have never been like that. Most people complied with the orders without much grumbling, but also didn't particularly care whether others followed them or not. Out of about 300 familial, professional, and friendship associations I have that span the entire political spectrum, I knew maybe 7 or 8 who took adamant and angry stances on one side of the other. Most everyone else was sort of mildly but not overly concerned and rather live and let live about it. Every right wing Republican I know got the vaccine, and a large portion of the left-wing people would admit privately that they thought the whole thing was overblown and they weren't particularly concerned about catching Covid (and most of them broke the lockdown rules anyway). I think the supposed polarization is a bit of a internet creation. You can ask someone a yes or no question on a survey, but are they measuring the strength of the response and how strongly one feels about their answer?
I was also surprised that the two Astral Codex surveys on Covid asked questions solely about negative impacts on the respondent's life and not a single question about positive impacts.
If marijuana reduced collective IQ by 1 or 2 points but made everyone a lot more chill, I think it'd probably be worth it.
I really enjoyed this meta process reflection on doing that long piece. I hope you do this kind of thing again, the process reflection, not necessarily the long piece, though that was good too.
It's an important insight that when you get closer to something, you find the naive concerns about it are irrelevant to what's actually going on. I find this is true across a wide range of topics and disciplines. Consequently I highly recommend getting close to things as soon as possible.
For example, if you think you might want to be a lawyer, go shadow three different ones for a day right away, rather than spending extra time studying for your SATs so you can get into a pre-law program, so you can get into law school, so you can get an internship, so you can get hired 20 years from now and then finally discover that you hate something about the job that nobody ever mentions. (John Taylor Gatto was big on this, and recommended "a thousand different apprenticeships" starting in elementary school as part of his education reform platform)
Or consider the debates about ethics in self driving cars; should they swerve to hit the aged nun or continue off the cliff? Great for journalist thinkpieces, but completely irrelevant to the ways that self driving cars are actually programmed.
Heinlein recommends getting close to lots of things: "A human being should be able to change a diaper, plan an invasion, butcher a hog, conn a ship, design a building, write a sonnet, balance accounts, build a wall, set a bone, comfort the dying, take orders, give orders, cooperate, act alone, solve equations, analyse a new problem, pitch manure, program a computer, cook a tasty meal, fight efficiently, die gallantly. Specialization is for insects."
But I find it's also important within narrow specialties; I've seen people argue about a flaw in a piece of hardware in endless meetings without ever having laid eyes on one. When I started carrying one around to all the meetings, the problem was understood much faster. Or when I've gone off daydreaming for days about a potentially difficult part of a problem, only later actually making an attempt at a solution and finding the hard parts are elsewhere.
I'd say the effect is roughly logarithmic; huge gains when you go from zero encounters to one, then diminishing returns as you spend more time with a thing and become an expert in it. But it's amazing how long people will go before getting those first few encounters.
This was a really interesting and helpful post and I'd enjoy reading more like it.
Also, it made my day that Scott has apparently had classical rhetorical terms drilled into him at some point, just as I have.
Scott, you have been talking to the wrong Australians. I am surrounded by them and they don't share the views you report: quite the contrary.
Re: the CS/AI/stats thing - I've tended to notice that something similar applies in psephology (study of elections). Nate Silver was primarily a statistician, not a political scientist.
Over here in Australia, our best-known psephologist (employed by our public broadcaster, the ABC to analyse redistricting and provide election coverage) was primarily a computer scientist who went into election research on a whim:
https://www.abc.net.au/news/redirects/backstory/television/2019-05-14/antony-green-on-30-years-of-analysing-elections/11110578
Another well known election analyst in Aus was primarily an invertebrate ecologist (Kevin Bonham).
One who used to write a lot about psephology and polls (now moved onto COVID-19 stuff) graduated in econometrics (Bryan Palmer, pseudonym Mark Graph).
One who used to have an active pseph blog but is now mostly on Twitter (Pollytics/Possum Comitatus, aka Scott Steel) was primarily an economist.
I know at least one pseudonymous pseph who was primarily a philosopher (unable to reveal which, it's in confidence).
Of the remaining well known ones, I think three are primarily political scientists (Peter Brent, Ben Raue and Malcolm Mackerras). So that's a 4-3 in favour of non-polisci to polisci (or 5-3, if you count the pseudonymous one) amongst Australian psephologists.
I think there's probably something - maybe the uncertainty, maybe the politics, maybe the status - about fields like election analysis and live pandemic epidemiology which attracts people from other fields to come and contribute.
“This question was too multi-dimensional.”
^ this (this whole section) is a concise summary of what it’s like to work in data science, and it makes me sad.
My feeling is that, if the validity of a study hinges upon the right way to "crenulate the zeugma", then it is very likely that the study is too tentative and/or the effect is too weak, and should not be used as the basis for public policy.
>Whenever I talked to Aussies or New Zealanders, they just really wanted to stress that they had ascended beyond such primitive mortal concerns, defeated the virus their own way, and were somewhat annoyed that the rest of us were squabbling about the relative merits of our inferior plans rather than focusing on how great they were. Sorry, guys.
We Aussies boned up the initial phase as badly as anyone else (refusing to close all borders long after it had spread beyond the PRC), we were just lucky it came here late. An amount of the rest I am unable to quantify is downstream of that.
Re the emotional damage, I find it striking that you're overlooking the biggest source of emotional anti-utilities on the pro-lockdown side. The pain of loosing a loved one years before they would otherwise have died.
I don’t get your comment about Aus/NZ. The key tool we used once cases got too many to contact trace was lockdowns. We used them repeatedly and they worked every single time. A few times we tried to avoid them and things got out of control. See Melbourne last winter and Sydney now. It seems very solid evidence that lock downs work extremely well.
Isn't it fascinating how conservatives, usually the ones more concerned about hygiene and infections, threw all of that overboard without blinking in this particular case, when the normal tendency would have been unusually useful to stay with?
I'm surprised nobody has touched what I thought was the most interesting part of your post - the bit where you find two opposite impossible-to-measure outcomes completely overdetermine your complex model and any work you do refining anything other than these parameters is completely swallowed up. I built a similar model to you for lockdown and found the same result you did - your assumption around the emotional impact of lockdown dwarfed everything else you might assume about the disease, including deaths and the emotional impact of deaths (actually I found that if it seriously and permanently harmed productivity / capita by knocking people permanently into an unemployment trap then this was plausibly of the same magnitude(ish) as the emotional effects of lockdown if you assumed weak emotional effects, but emotional effects were still a big deal)
At the risk of raising yet another highly controversial topic, I tried to do what you did for lockdown with Brexit. I included maybe 20 parameters in my model, and then discovered that the 'will it increase or decrease economic growth over ten years relative to no Brexit' was literally the only parameter that made any difference whatsoever; for any reasonably set of preferences a difference in the answer to this question of 0.01% either way was greater than literally any possible combination of outcomes to the other 19 parameters.
Therefore I don't think this is an isolated example, and it seems likely that most big political decisions have this character (on a sample of n=2, but a couple of other rationality tropes are similar - x-risk, cryonics, EA and so on). I really think this must be an understudied problem in decision making, because otherwise it would be all anyone talked about when making decisions.
For the record, I thought it was clear your conclusion was that lockdowns weren't worth the emotional cost, but I guess putting it in bold, TLDRing, and adding a witty partisan headline would have helped.
"This question was too multi-dimensional. As in, you could calculate everything right according to some model, and then someone else could say "but actually none of that matters, the real issue is X", and you would have a hard time proving it wasn't."
Scott, I hope you will never let something like this discourage you from treating a subject as multi-dimensional. It's easy to underestimate how different the utility functions of different people are. Some people will not care much if they (or others) die from Covid, but find the risks of LongCovid terrifying. It is quite tempting to say "they don't understand what they are talking about" in such cases, but I think that we should really not give into that temptation.
Giving a multi-dimensional treatment helps all readers, not just those who happen to have the same utility function as yourself. I remember your post "moral cost of chicken vs. beef". It convinced me that we have a strong moral obligation to replace beef with chicken. This was exactly the opposite conclusion from yours! (I had this opinion before reading your post and already acted on it, but your post made me much more certain about it.) This is just because we have different value functions.
I am not discouraging you from giving your opinion and conclusions. But I would be careful leaving things out because they seem not so relevant. (But obviously, there are limits to this.)
On your liberal vs. libertarian question, dont you think that in Iran the far majority of psychologists would make arguments for the positive psychological effects of Niqab? If then the one secular dude, who writes about Russell and Ataturk all the time, all of a sudden is the only guy to find evidence to the contrary, is that suspect?
A senior stateman reveals the inside story on how Adorno, Huxley and Rees conspired to destroy America and impose a New World Order: https://monthlyreview.org/castro/the-world-government-part-1/
"(...) To keep all populations outside the minority elite under an almost permanent state of oppression and make them love their servitude, (...) the main tools used to achieve that were some vaccines which altered the functions of the brain and the medicines that the State forced the population to consume. In Well’s opinion, this was not a conspiracy, but rather a ‘one-world brain’ which would function as a police of the mind."
"that's just a coincidence - if we had been in a world where academia leaned conservative"
There is a widespread liberal argument that this isn't a coincidence, but that academia leans liberal because liberal positions are more often factually correct and conservative positions are more often factually wrong.
As a fully-fledged member of the Blue Tribe, I can't possibly come to any neutral judgment on this, but it does seem to be worth consideration.
Re: discounting a conservative/libertarian's study against lockdowns and not discounting a presumed-liberals study in favour of lockdowns - it's not as simple as just "study aligns with beliefs". There are a number of factors all pointing in the same direction.
It seems obvious on the face of it that *of course lockdowns work to some degree*, and the question is just how much. It would actually be pretty bizarre if lockdowns didn't attenuate the transmission of an airborne agent. It would be like finding a study that flossing didn't prevent tooth decay - it seems pretty obvious that it must, so we will naturally be skeptical of the result.
There is a general consensus that lockdowns work, so they have to provide enough evidence to show they're right and everyone else is wrong.
It's also true that the "conservative/libertarian" position on many science-related issues is patently false/absurd. Climate change isn't real, for instance.
So it's not one pointer like it seems, it's a few which conspire. It's "study aligns with prior beliefs" (possible bias to find desired result), it's "group is known for lying about science" (low prior credibility for members), "result contradicts obvious likelihood" (higher burden of proof), and "study at odds with general scientific consensus" (higher burden of proof).
The liberal, pro-lockdown study in the hypothetical case of a researcher who's equally biased only has to contend with "study aligns with prior beliefs", and in fact any doubt is *attenuated* by "group is known for trusting science", "result in line with obvious likelihood", and "study concurs with general consensus".
Of course, the study can be good or bad quite apart from all of that, but I think that's the explanation for why "conservative/libertarian finds evidence lockdowns are bad" seems on the face of it worse than the liberal pro-lockdown study without reading either. It applies more or less equally to the pro-gun/anti-gun position, or a number of other issues split along red tribe/blue tribe lines.
If it was extremely difficult for a well-connected, intelligent, and thoughtful person such as yourself to analyze the pros and cons with the benefit of a full year of hindsight, imagine how hard it was for the public health authorities, and government, to do so last spring and summer.
That is, I think two additional conclusions you might reasonably draw are:
(1) It's really REALLY hard to get public health measures bang-on correct, when you are looking forward and dealing with an infectious virus, the human immune system, and dynamically responsive human populations in an ideologically, economically, and intellectually diverse population of 300 million. Perhaps "didn't utterly screw it up" is about the highest standard of success we can reasonably expect, and we should critique our public health authorities and government appropriately*.
(2) Maybe a lot of governing of 300 million people with a view towards eliminating bad luck and improving everyone's lot in subtle and difficult to measure ways -- like their emotional well-being! -- is incredibly difficult to get right, and our default expectations about that should be deeply skeptical, no matter how persuasive the hypotheses are -- and we should craft our political institutions with an eye to that skepticism.
---------------
* It isn't cost-free to have expectations for government that are much higher than it can reasonably achieve, because when it routinely falls far below expectations, that leads to broad cynicism about even the things it might be able to do well *and* it leads politicians to start lying a great deal more, on the grounds that it might actually turn out to be easier to obfuscate the data to avoid the lie's exposure than it is to get genuine results done that are good enough to win re-election. I think we see both these trends in politics today, and I think both can at least in part be traced back to unreasonably high expectations among voters about what government can actually be reasonably expecte to get done and get done right.
I wonder why emails were more helpful than comments on the post. I have a couple of hypotheses.
1. Commenting may be a form of status signaling, in which you align yourself with one side and bash the other. Hence, it is less prone to neutrality than an email.
2. Emails are generally longer and more detailed than comments, and hence can be more helpful/insightful.
3. Researchers in the field, who on average would have more informed and accurate things to say, may be reluctant to discuss their stands on a public forum, and hence emailed Scott directly.
Although 3 seems to be the most important reason, I wonder if 1 and 2 played a part at all.
I wonder if the point about "300 million people going to the bar" is valid here.
Imagine that there are 1000 billion creatures in the universe. If the pain of the death of one person is is roughly equal to the inconvenience caused by stopping 300 million people from going to the bar, then this calculus can be extrapolated to "death of one person= slight ticklish sensation for a second to 1000 billion people". Can we really say that the two are morally equal?
I think pain is mostly personal, and cannot be "added up" across persons.
I doubt that pro-lockdown authors are facing a lot of heckling and anti-lockdown authors are facing little. Rather, concern about a deadly viral pandemic and concern about how strangers will treat you on the street/internet are probably correlated. Both have ambiguous and unclear yet occasionally fatal or life-ruining danger, for which the main defense is to stay inside.
Also, the question of whether a conservative producing conservative science or a liberal producing liberal science, etc, can be trusted, I suspect you have to ask this on a case-by-case basis, and I don't think that any one ideology is more trustworthy. Find a few scientists who seem to be particularly honest, credible, diligent, listen to what they say and ignore everyone else.
For me, the huge issue with lockdown is not emotional damage, but something related : it's establishing a precedent for a central power to monitor and control personal behaviors and limit essential freedoms to achieve difficult to measure societal benefits... This is a very bad precedent, like anti terrorist measures but even worse. The west looks a lot like China did 5 years ago, when all progressive media were telling how Orwellian it was. The same media now preach safety above everything else without worrying much about citizen tracking and restrictions...we are on a very bad slope here, and we sleep faster and faster regardless of the actual epidemy strength... Which make the epidemy looks more like a convenient excuse...
About math and models, more specifically predictive models:
I think worth mentioning that math is not the hard part in predictive models. You make your hypotheses, you set your parameters, you do the math, you read the results. The math here is often simple. Or maybe it is not, if you want a beautiful solution as closed-form expressions; but we do not need that to read the results, we only need an accurate enough numeric value. The models are usually a set of differential equations or some kind of discrete dynamic system: our everyday computers can solve most of them numerically in the matter of seconds.
The hard part, really, are the hypotheses and parameters. About parameters, John von Neumann put it this way: “With four parameters I can fit an elephant, and with five I can make him wiggle his trunk.”
But the real stopper is hypotheses. Models have a boatload of them, most of them implicit.
For example, when people were about computing the herd immunity threshold: the main parameter was the average number of people anybody may contaminate. Key word: average. Introverts will meet fewer people, with a lower risk of contamination, the opposite for extroverts. Does it average out? No: common sens tells us that extroverts will be contaminated early, then immune, hence slowing the propagation. And models designed to test that common sense confirm it: random graphs where 90% of the nodes have two links and 10% have 12 have a lower herd immunity threshold than graphs where all nodes have three links.
Furthermore, some epidemiologists knew that. There were articles about it published, even before the current crisis IIRC. But when you look at the models, the average is the only distribution parameters. That translates into the implicit hypothesis “everyone's as much an extrovert as everyone else”. Which does not match reality. But since it is implicit, it is rarely discussed.
The important thing to remember about predictive models is that they can tell us what can possibly happen, but they are too many uncertainties for them to tell us what will happen.
If you can read French, or stomach an automatic translation (it should work reasonably well for that kind of text), the following blog article, and the related articles, tells it better than me:
http://www.madore.org/~david/weblog/d.2020-04-14.2650.html
It connects with the Euler remark: when Euler says “𝑒^𝑖π+1=0 therefore God”, he really is saying “𝑒^𝑖π+1=0 AND beauty implies God, therefore God”. We can call bullshit on the implicit hypothesis, which is not math.
About unexpected consequences:
I can argue that the Nazis caused the Manhattan project, which caused civil nuclear power earlier, which might allow us to barely escape the worst of climate change. So, Nazis good? Certainly not, it was not on purpose.
The conclusion I take is that History is chaotic, and worrying about hypothetical indirect consequences too far removed is a waste of time. Cannabis may decrease intelligence? Maybe, but people more relaxed may hit their kids less, making the next generation smarter.
The lockdowns, by changing our relation to work and welfare, may have opened the road to Universal Basic Income and Universal Healthcare, and caused the end of poverty before 2040. Or they may have strengthened the choke-hold of tech corporations on society and led us to the standard cyberpunk dystopia.
I do not have a strong argument for it, but my intuition tells me that we need to assume that the indirect chaotic effects average to 0, because otherwise we cannot do anything. It is not very satisfying; but if psychohistory was easy, it would have been invented already.
About another concern with regard to lockdowns:
You certainly remember that after 9/11, authorities have pushed for security theater, widening the powers of the executive branch and law enforcement, and it happened again after each episode of terrorism. For example, the Crypto Wars were re-kindled by the need to decrypt a terrorist's phone.
The phenomenon is simple: power attracts people who want power, if they can get more of it they will, and events that scare the population are a good occasion to consolidate power.
As you pointed out, lockdowns in Europe were probably more strict than in the United States. To implement them, governments had to have emergency laws voted. When the emergency subsided, some of these laws ended, and some of them have been kept and made permanent.
One of the concern I and many other people had about the lockdowns were exactly this: they being the excuse to weaken legal protections for fundamental rights, with a good measure of ratchet effect: once the population of an area had been locked down under police control to save lives from a pandemic, and society has not collapsed, we can be locked down under police control again for other reasons, say a terrorist manhunt, or protests against an unpopular reform of tax pensions.
It was especially visible in France, where the implementation of the lockdown were quite stupid, to the point that German newspapers called us Authoritarian Absurdistan. Parks and forests were forbidden. To go to groceries, we had to fill and sign a form for ourselves; police chiefs have bragged how many people they have fined at the very doors of convenience stores. And whoever complained about it were swiftly called covid denialists and compared to Trumpist anti-maskers.
The weakening of fundamental freedoms and checks-and-balances, both in law and in mentalities, is a long-term consequence of the lockdowns that I fear might feel deeply.
We are carrying out a real-time experiment on this in Australia as we speak. Both Melbourne and Sydney are having lockdowns because of delta outbreaks. Melbourne lockdown was very quick and strict, Sydney hesitated. As a result, Melbourne looks very close to proving conclusively that it is possible for a 5 million people city to control delta and (hopefully, with a little luck) bring it down to 0. Sydney at this point seems to be losing control.
The difference between the 2 cities is the ideology of the state governments, Labor in Melbourne, right-wing Liberal in Sydney.
The utility arithmetic of emotional consequences doesn't sit right with me...
I don't think you can assign a negligible negative utility to some experience, then just naively multiply that utility by the number of people having that experience to create some huge negative utility.
For instance: If a genie would let me save one person's life on the condition that a hundred people had to stub their toes, I would take that deal in a heartbeat. My decision doesn't change if it's a million people. Or a billion. Or a trillion.
Somehow, my intuition is that if a utility is small enough, it doesn't simply add up over several people.
Similarly, I *don't care* if a billion people are sad about not getting to go to night clubs for a year, if that's the price to pay for saving lives. For the same reason that I would happily sacrifice a billion servings of caviar to stop covid. "Oh no I don't get to eat caviar" times a billion. Get over yourself you spoiled brat. Night clubs are a luxury, and you should be happy to give up that luxury for the opportunity to save lives.
Maybe write a post about the emotional damage subject specifically? You're right that it would be transgressive. Anyone who brings it up is in for some shaming as being selfishly concerned with their own comfort "while people are dying!"
These kinds of calculations always subtly ignore the value of simply not being forced to do something. You don't have to be a raging libertarian to realize that "mere" personal freedom has some real moral value. Not being forced to do something is an inherent good.
That's why I think it's much better to do trade-off-type thought experiments to judge the rationality of lockdowns. Find some phenomena, X, that roughly matches the distribution of mortality risk as covid (plus some padding for the "unknowns", if you want), and then 1) see how much effort people currently expend avoiding such risks and then compare it to relative hardship of lockdowns, and/or 2) ask people if they'd vote to lockdown to avoid the risks of X. (Of course, when polling you have to avoid mentioning covid, lest you taint it with politics.)
Asked this question on LessWrong: https://www.lesswrong.com/posts/Rm5FdDz8rv9kiqdSQ/do-we-have-a-term-for-the-issue-with-quantifying-policy
I feel like your analysis had an asymmetry, where you acknowledge that R reduced mainly because of voluntary behavior changes rather than lockdowns, but then attribute all the costs to lockdowns when most of the costs were also caused by voluntary behavior rather than government policy.
I feel dumber for having read this. Sweden, Afganistan, other places that simply ignored covid, provide irrefutable proof that covid while a real cold/flu like disease, us no more dangerous than flu. There are no excess deaths, or the effect is very small. Death rates are roughly same as 2017, 2018
Any kind of statistics other than overall amount of people dying, are meaningless. US gov got initially freaked out by COVID, and threw tons if money at it. Where there is free money, there is grifting.
Situation: 10 grandmas walk into ER, presenting fever, headache and coughing. These are symptoms of both flu, and covid. But treating COVID patient milks insurance for 100k, while treating flu patient hardly milks 20k. Do you think our hospitals are going to classify these 10 grandmas with flu, or with covid? "But but but the PCR test show covid". Do you undetstand how PCR test works? Its exponential magnification. With 45 cycles you will detect covid on a ham sandwich with 50% probability. With 50 cycles every test on any substance will come out as covid.