Tag Archives: coronavirus

Coronalinks 5/18/20: When All You Have Is A Hammer, Everything Starts Looking Like A Dance

It is the sixty-first day of shelter-in-place. Anti-lockdown protesters have stormed your state capitol, chanting Nazi, Communist, ISIS, and pro-Jeffrey Epstein slogans to help you figure out they’re the bad guys. Inside, the Governor has just finished announcing his 37 step plan to reopen the state over the next ten years. You kind of feel like he should be a little more proactive, but the protesters outside have just unfurled a Khmer Rouge flag, so you hold your tongue.

Meanwhile, a band of renegade economists, tech billionaires, and MIT professors has just announced a bold disruptive Manhattan-Project-style moonshot: send a team of researchers to the swamps of Florida, where legends speak of a Fountain of Youth whose water can cure any malady. But disaster strikes when Florida’s governor announces that exploration is not an essential activity, and threatens to release the quarantine enforcement lions. The nation looks to the White House to solve the growing conflict, but President Trump is too busy evangelizing his latest coronavirus cure: eating those little packets of silica gel in food that say DO NOT EAT. As the Western States Pact and the Eastern Bloc inch closer to war, all that the rest of us can do is strive to stay as well-informed as possible, trying to make sense out of an increasingly nonsensical situation. So:

“Jail Isn’t Real”, I Assure Myself As I Close My Eyes And Drive To The Hair Salon

Here are some CDC graphs that use cell phone data to measure percent of people leaving home over time (source, h/t Kelsey):

On this measure, official government stay-at-home orders didn’t seem to affect the percent of people staying at home, even the tiniest bit.

Facebook is tracking something similar – see their terrible and hard-to-use website here. Here are the data from California:

Can you tell what day a statewide lockdown order was issued? (click here and check the date of the article for answer).

If comparing times doesn’t impress you, we can also compare places. Sweden has attracted international attention for its refusal to shut down business – restaurants and bars there are open as usual. And Nashville has attracted attention as a center of growing anti-lockdown protests by people who think its shelter-in-place order is too strict. But cell phone data finds that citizens of Stockholm and Nashville “have nearly the exact same adjustment in driving, walking, and transit use”.

What’s going on here? On the one hand, lockdowns are poorly enforced and we’ve all seen pictures of people going to the beach unmolested. On the other hand, surely fewer people are going to work, since the offices are all closed? Surely fewer people are going shopping, since the malls are all closed? What about all those pictures of empty freeways during rush hour?

The best answer I can come up with is that most people are risk-averse and started staying at home before the official lockdown. A few risk-tolerant people didn’t, but they are disobeying the lockdown as much as they can anyway. And the cases where risk-tolerant people can’t disobey the lockdown aren’t numerous enough to show up in aggregated cell phone data.

I find this answer pretty unsatisfying, so maybe I’m just misunderstanding what the cell phone tracking data are trying to show, or how much I should expect from them. This might also be a good time to review the ongoing debate about whether reality drives straight lines on graphs, or straight lines on graphs drive reality.


Our World In Data deserves some kind of prize. I think they already have won some prizes, but they deserve better ones. Possibly a Nobel, or a knighthood, or beatification. They give you any information you want, from any country you want, and display it however you want. If you’re trying to figure things out about the coronavirus and not using Our World In Data’s tools and graphs, you’re missing out.

Here’s coronavirus cases per capita across countries, by days after the number of cases per capita in that country passed one in a million. There’s some debate over what we gain/lose by adjusting/not-adjusting for per capita, but I think this is probably the best measure of how good a job different countries are doing at containing the infection:

And here’s coronavirus deaths, by the same measure:

And to save you the trouble of having to divide the first graph by the second graph, case fatality rates:

And since all of these numbers are confounded by testing rates, here’s tests per thousand people:

You’ll hear a lot of bad takes about where America ranks relative to other countries. Ignore them and look at these four graphs. America has one of the highest infection rates of any developed country, trailing only Spain. But it has one of the lowest mortality rates of any developed country, beaten only by Germany, Denmark, and a few other of the usual high performers. It’s right in the middle in terms of numbers of tests, beating eg Netherlands and Sweden, but trailing Germany and Denmark (though it may have an “advantage” on testing since so many people are infected).

Why is US mortality rate so low? The rate could be artificially low it we were unusually good at testing, but we aren’t. It might mean our health care system is unusually good, but that doesn’t seem like us either. I haven’t heard anyone claim that our standards for reporting a death as COVID-19 are stricter than anyone else’s, and our uncategorized excess mortality doesn’t seem much different from everyone else’s. I notice that all highest-mortality-rate countries are European, and that less-developed countries tend to be lower. Maybe it’s something about density? Maybe Europe got a different strain of the virus than everyone else? I only see a few people talking about this (Kevin Drum continues his 100% success rate of having gotten to interesting topics before I did, but see aso here and here), but nobody seems to have much in the way of a theory.

[edit: bpodgursky points out that European populations are, on average, much older than the US]

How effective has lockdown been at controlling the spread of the virus? Three countries have made the news for unusually weak/nonexistent lockdowns – Sweden, the Netherlands, and Japan. All have chosen to keep most of their economy open in the name of “herd immunity”, although they’ve banned very large gatherings like concerts and sports games. If we accept Denmark, Germany, and South Korea as “matched controls” that have made more aggressive efforts to contain the disease, it’s hard to see much of a difference. Sweden’s doing worse than Denmark, but not as much worse as we might have expected. Netherlands is marginally worse than Germany, and Japan marginally worse than South Korea – but it all seems within chance variation. If we’d chosen to use Belgium as a control for the Netherlands (I didn’t because it’s recording mortality statistics in an odd way), Belgium would have looked worse. Countries like Spain and the UK which responded pretty aggressively have more cases than a lot of countries that are barely doing anything at all.

This seems to match the conclusion from the last section: government policy isn’t mattering as much as we think. We thought South Korea and Taiwan were doing well because their governments were so brilliant and competent, but Japan’s government kept denying the problem existed in order to preserve their shot at holding the Olympics, and they seem to be doing equally well.

Switzerland is another weird case. It’s a loose confederation of linguistically French, Germany, and Italian regions. Remember that France and Italy have been devastated by the virus, and Germany has mostly gotten off unscathed. The same is true of Swiss regions; French- and Italian-speaking cantons have been devastated, while their German-speaking neighbors wonder what all the fuss is about. The Swiss government swears that this has nothing to do with policy. From the linked article:

Talking to Swiss media outlet Le Temps on Thursday, the Federal Office of Public Health (FOPH) said the efforts to combat the virus were largely uniform across Switzerland – meaning other situational factors were at play.

The measures were decided at a federal level when the epidemic was already advanced in different ways in different regions,” a FOPH spokesperson said. “This has nothing to do with a difference in the extent to which different cantons have taken action.

This isn’t to say government policy doesn’t matter – just that it adds or subtracts a relatively small modifier on top of a much wider variation.

What’s going on? Various hypotheses – BCG vaccinations, smoking rates, genetics, different viral strains – have come and gone, mostly unconvincingly. Another hypothesis – time – does a little better. Countries with different levels of connection to Wuhan and different luck in terms of superspreader events had their epidemics start a month or two earlier or later. Later countries had the benefit of warmer weather and a more aware population who were already taking social distancing measures on their own regardless of what the government told them (national education level might play into this too). Then lockdown strength added a little more or less on top of this.

This explains a little. But it doesn’t explain NYC vs. the rest of the US, or Japan vs. the rest of the world. Something’s still missing here.


It’s becoming increasingly clear that a big (maybe the biggest) risk factor for coronavirus transmission is speaking. Singing is even worse. The louder you speak or sing, the worse it gets.

Some confirmed early superspreader events were choirs. A lot of others were churches, where everyone gets together and sings hymns full-blast. This person’s explanation for the surprisingly low rate of subway-mediated transmission in Japan is that nobody talks on a Japanese subway.

All this makes sense. Coronavirus has mostly droplet transmission. There are three ways to get droplets: coughing, sneezing, or talking/singing. You do one of those about a thousand times more often than either of the others.

I appreciate how much pressure there is on governors to open up churches, but they should be very careful about this, unless churchgoers can promise to stay uncharacteristically silent. I realize how bad it will look to say that golfing and rock concerts and orgies are allowed but churches aren’t. Still, governors should swallow their pride and stand firm.

(Or they could just troll people. I understand Muslims pray quietly facing the ground, so how about ordering that mosques are allowed to reopen but churches aren’t? Then sit back and watch the fireworks.)

An extremely crackpot theory – could the missing cultural vulnerability factor be how loud and spittle-filled people’s speech is? When I think of a country where people talk very loudly to each other without much personal space, Italy is on the top of my list. Then comes New York City, which muscles its way in even though it isn’t even a country. When I think of countries where everyone talks really quietly and far away from each other, I think of Japan, South Korea, and all those other Asian countries that have mysteriously escaped infection.

Are there counterexamples? I Googled “loudest cultures”, and got people talking about Australians, Africans, and Cantonese Chinese. All those places have done pretty well – though they’re all pretty warm right now, which is a confounding factor. Without being able to find some kind of official data about conversational volume, I’m not sure there’s much I can do with this hypothesis right now.

[EDIT: Commenters point to Georgians and Germans as people who speak loudly yet have avoided coronavirus epidemics]

Love In The Time Of Coronavirus

One of my housemates lives two blocks away from her boyfriend, and hasn’t hugged him in three months. A friend lives a town over from his parents, and hasn’t been able to visit them since March.

Meanwhile, government offices and the media are all talking about how to get hair salons to reopen as quickly as possible, and making detailed lists about what kind of golf courses are or aren’t okay. The governor of California has a four phase plan that discusses exactly what criteria need to be fulfilled before we can have fitness centers, swimming pools, and rock concerts – but not a word about when you can hug your loved ones.

Probably the government just assumes everyone is already breaking those rules and there’s no point in worrying about them. I think this assumption generally holds. My patients are mostly law-abiding upper-class liberals who think of the lockdown protesters in Michigan as basically death cultists – and almost all of them casually let slip that they’ve gone over to their parents’ for dinner, or visited their partner, or even had small gatherings with close friends. The cell phone tracking data is equally pessimistic about the lockdowns reaching too far into the private sphere.

But some people are genuinely law-abiding. Fewer all the time, now that they’ve spent months shut off from everyone they love, while watching everyone else go to the beach with their buddies and face zero repercussions. But the government should acknowledge that these people exist and try to support them. Give the slightest acknowledgment that in between declaring marijuana dispensaries an essential activity and saying that even though nobody else is allowed to work Elon Musk can reopen his Tesla factory because he’s famous, someone is also making a plan for when you can see a friend again. Even something like “once we reach Phase 2 of the reopening, you may visit one person outside your household per week” would ease a lot of people’s misery. Even if this isn’t the best idea from a epidemiological standpoint, they should do it anyway, because otherwise people will visit people outside their household and lose all respect for the law in general.

Information Wet Markets

At last, coronavirus prediction markets have arrived. Check out CoronaInformationMarkets.com and start investing, unless you live in the United States which is an authoritarian Nazi communist Luddite hellhole and bans you from contributing. Some highlights:

“What percent of the global population will be estimated to have contracted COVID-19 by the end of 2020?” – LESS THAN ONE PERCENT is at 11%, BETWEEN ONE AND THREE PERCENT at 18%, BETWEEN THREE AND FIVE at 39%, and GREATER THAN FIVE at 32%.

“Will hydroxychloroquine be approved as a treatment for COVID-19 by the FDA by October 1 2020?” – YES has 32%, NO has 68%.

And “Will a vaccine be approved before the end of 2020” – almost exactly split, 48% YES, 52% NO.

If you’re too chicken to bet real money, or you live in a fascist antiintellectual statist kakistocracy like the US, you can go to Metaculus, which continues their great work soliciting and aggregating predictions made with fake Internet points.

“When will a SARS-CoV-2 vaccine candidate that has demonstrated an efficacy rate ≥75% in a n≥500 RCT be administered to 10M people?” – median guess, October 10, 2021

“When will the Dow Jones set a new all-time record high after the coronavirus crash of February 2020?” – median guess, December 24, 2021.

“When will Disneyland reopen?” – median guess, August 29, 2020

“Will it be reported that Donald Trump tested positive for COVID-19 in 2020?” – median guess, 25%

There’s also a section on the total number of worldwide coronavirus cases in each quarter of 2020. Q1 (January 1 to April 1) saw about a million worldwide. Today we’re at 4.6 million. Metaculus thinks that by July 1, we’ll be at 7.5 million. By October 1, 9.8 million. By year’s end, 10.9 million. Except that I got those by adding results from different quarters together, and an alternative question that just asks that directly gets 16.2 million. Come on, people! Do some arbitrage! It’s almost like you’re not sufficiently devoted to winning fake Internet points!

Wash Your Hands!

Weird Sun Twitter does handwashing timing mnemonics:

When All You Have Is A Hammer, Everything Starts Looking Like A Dance

Everyone is hoping for a definitive solution to coronavirus. A vaccine, or a good antiviral, or a test + trace regimen so well-coordinated that it stops the virus in its tracks.

Suppose that after X years, we realize there is no definitive solution. We are faced with the choice of continuing restrictions forever, or lifting the restrictions, letting lots of people die, and getting herd immunity the hard way. What then?

If we lift the restrictions, the same number of people will die as if we had never instituted any restrictions at all, and also we will have wasted X years. We will have gone X years with millions of people poor and unemployed, millions of others locked in their houses and unable to have fun – and it won’t have saved a single life.

If there’s a 50% chance of a definitive solution in one year, is it worth staying locked down until then? What about a 25% chance in five years? 10% chance in ten years? If there is never a definitive solution, are we willing to stay locked down forever?

Also: if a lockdown lasts a long time, what’s the average R0 during that phase? One possibility is that it’s less than 1, in which case the virus will “die out” locally (although it probably won’t go extinct smallpox-style – too much opportunity for other countries to reinfect us). Another possibility is that it’s more than 1, in which case lockdown isn’t working and we get continued exponential growth ending in lots of deaths and eventual herd immunity.

Is there a possibility where R0 is exactly 1? Seems unlikely – one is a pretty specific number. On the other hand, it’s been weirdly close to one in the US, and worldwide, for the past month or two. You could imagine an unfortunate control system, where every time the case count goes down, people stop worrying and go out and have fun, and every time the case count goes up, people freak out and stay indoors, and overall the new case count always hovers at the same rate. I’ve never heard of this happening, but this is a novel situation.

If that were true, right now we’re on track to gain herd immunity in 30 years. This would be another worst-of-all-worlds scenario where we have all the negatives of a long lockdown, but everyone gets infected anyway.

Sing, O Muse, Of Arbit-Rage

There’s a morbid joke about the news, which goes something like:

10,000 Africans in a famine =
1,000 Chinese in an earthquake =
100 Europeans in a plane crash =
10 Americans in a terrorist attack =
1 pretty white girl getting kidnapped

(if you want to go a different direction, you can add “= 0.1 black people murdered by cops”)

Coronavirus has killed about 100,000 Americans so far. How bad is that compared to other things?

Well, on the one hand, it’s about 15% as many Americans as die from heart attacks each year. If 15% more people died from heart attacks in the US next year, that would suck, but most people wouldn’t care that much. If some scientist has a plan to make heart attacks 15% less deadly, then sure, fund the scientist, but you probably wouldn’t want to shut down the entire US economy to fund them. It would just be a marginally good thing.

On the other hand, it’s also about the same number of Americans who died in the Vietnam War plus the Korean War plus 9/11 plus every school shooting ever. How much effort would you exert to prevent the Vietnam War plus the Korean War plus 9/11 plus every school shooting ever? Probably quite a lot!

Maybe part of this is that heart attack victims are generally (though not always!) older than 9/11 victims, so the cost in DALYs is lower. But the bigger problem is that there’s no arbitrage in the market for lives. Some normal good, like Toyota Camrys, sells for about the same price everywhere. There might be minor variations based on how far you go from a Toyota factory or something, but overall you wouldn’t expect the same Camry to sell for ten times as much in one city as another. Someone would arbitrage – buy the Toyotas in the cheap city and sell them in the expensive one! But the same reasoning fails when it applies to lives. Life has no single value denominated in dollars, attention, or outrage. So when we search for metaphors to tell us how bad 100,000 deaths from coronavirus are, our conclusion depends entirely on what metaphor we use. “It’s like 15% of heart attacks” sounds not-so-bad, and “it starts with the Vietnam War and gets worse from there” sounds awful, even though they’re the same number. There’s no way to fix this without somehow making all our intuitions collide against each other and equalize, which sounds really hard.

Suppose you reopened the economy tomorrow. You tried as hard as you could to put profits above people, squeezed every extra dollar out of the world regardless of human cost. And then you put a 1% tax on all that economic activity, and donated it to effective charity. Would that save more people than a strict lockdown? If a lockdown costs $5 trillion, then the 1% tax would make $50 billion. That’s about how much the Gates Foundation has spent, and they’ve saved about ten million lives. Ten million is higher than anyone expects US coronavirus deaths to be, so as far as I can tell this is a good deal.

On the other hand, the US spent about $5 trillion on the Iraq and Afghan wars. Even optimistically assuming this helped prevent some terrorism, it’s a no-brainer to say we should have accepted the cost in terrorist attacks and spent it on stricter COVID lockdowns instead.

Is spending resources on the coronavirus lockdown a good idea? A good idea compared to what? Compared to using resources efficiently, goodness no, not at all. Compared to putting the resources in a giant pile and setting them on fire, yes, definitely. Compared to usual practice? Usual practice basically involves alternating betwen the two previous options inconsistently; the answer depends on how long we spend in each category. At this point, we are too incompetent for questions about our preferences to even make sense.


The Marginal Revolution folks and the Mercatus Center are doing really amazing work to try to use economics to coordinate the pandemic response. Here’s a report by Caleb Watney and Alec Stapp about how the government should use purchase guarantees to boost production of essential medical equipment.

Patrick McKenzie of Kalzumeus, a Westerner in Japan, talks about his work trying to get them to realize the severity of the virus and take some response. Written on April 21, when Japan’s situation was at its worst and it seemed like he had presciently ferreted out an undercover epidemic. Since then, Japan has gone back to mysteriously defying gravity, so I’m not sure how to think about this now.

Mark Andreessen made waves with an article arguing that the coronavirus shows America needs to learn how to build things again. I also appreciated Ezra Klein’s response, which was that America already knows how to build things but is blocked by government dysfunction (I doubt Andreesen disagrees with this framing). Klein highlights the term “vetocracy” for all the features of modern society which give us a bias toward inaction. Clearly true and important, although a fuller treatment (which I hope to give!) would have to talk about the advantages vs. disadvantages of bias for action vs. inaction (the very end of this post can perhaps be interpreted as a paean to vetocracy). See also Mark Lutter: Build Institutions, Not Apps.

Related to the bat discussion from last post – a new paper finds that viral zoonotic risk is homogenous among taxonomic orders of mammalian and avian hosts – in other words, despite how it looks, bats don’t spread disproportionately more viruses to humans than any other animal – there are just a lot of bats. But see also this contradictory past study.

Earlier I asked whether some savvy early coronavirus investors had dealt a blow to the efficient market hypothesis. Now that the dust has cleared, I agree with this post saying the EMH still looks pretty good – although sometimes it moves in mysterious ways. Also in me-being-wrong news, evidence continues to come in about whether smoking is a risk factor for coronavirus, protective against it, or all the studies are biased and we have no idea. Something in this space will probably end up on my Mistakes page one day, but I’m going to wait until I can be absolutely sure I know what.

Everyone expected prisons and homeless shelters to be devastated by coronavirus, since they had lots of people together in close quarters and little ability to escape. Although these institutions have not had great times, they seem to have weathered the storm better than a lot of people would have predicted, mostly due to a high rate of asymptomatic infections. Why?

There was a lot of talk a few weeks about about Eastern European success at avoiding the coronavirus. Then Russia and Belarus’ case numbers exploded; both are now doing as bad as any Western European country. Poland, Romania, Czechia, and others continue to be oddly quiet. I suspect random variation – Russia and Belarus looked good until they weren’t – but I guess we’ll find out soon.

This article is kind of critical of Dominic Cummings, but the criticism is that he inappropriately pressured scientific bodies to order a UK lockdown ASAP, instead of letting the scientists take however long it took to evaluate the evidence in a proper scientific way. I like due process and checks-and-balances as much as the next liberal, but I also think you should be allowed to break the rules in an emergency and then let the people affected by your choice decide whether they want to show you mercy based on time proving you right, or punish you to the full extent of the law based on time proving you wrong. In this case he was right and deserves to be celebrated.

Some studies of remdesevir, not very encouraging. Hydroxychloroquine is basically dead in the water at this point, sorry Donald.

Some people are worried that coronavirus might be overblown and doctors are just classifying random other stuff as coronavirus deaths. The best antidote to this claim is this look at excess mortality over the average for this time of year worldwide.

You’ve probably already read this, but the story about how Trump’s premature praise for hydroxychloroquine caused some supporters to overdose on hydroxychloroquine-containing fishtank cleaner got a lot more complicated – they were actually both anti-Trump Democrats, and the woman is now under investigation for murdering her husband and inventing the hydroxychloroquine story to cover it up. This seems like one of those things which is probably a metaphor for life.

There’s been some worry about coronavirus reinfection – maybe people who have already gotten it aren’t immune and can get it again? A recent Korean study tried to put those fears to rest, showing that they were mostly testing errors. Professor Shane Crotty says he has studied the immunology of coronavirus and come to the same conclusion a- after infection, the immune system is able to create antibodies to it which prevent further infection for a while. Two data points don’t prove anything, but this is how things work with most viruses, so the burden of proof is on anyone who thinks COVID-19 is different.

Are coronavirus victims so old and sick they would just die from something else soon anyway? Two studies were recently reported as saying they would have lived at least ten more years, but read Scoop dissecting them in the comments section here.

A few weeks ago people were talking about the “iceberg hypothesis” – maybe detected coronavirus cases are “just the tip of the iceberg”, and there have been so many asymptomatic people that we’re nearing herd immunity already. Recent studies haven’t been kind to this proposal. Both France and Spain have about 5% seroprevalence, which means official counts are only off by a factor of ten, about what we already expected. It also means true mortality rate is still about 1%, also what we already expected (and high enough to result in tens of thousands of deaths before anyone gets herd immunity). No icebergs here. A Santa Clara study seemed to show 2% seroprevalence, which actually was much higher than expected and would be consistent with the iceberg theory, but Andrew Gelman is very much not impressed. Greg Cochran gives the hypothesis a postmortem here, and also is not impressed with claims that we might be able to naturally and easily achieve herd immunity before about ~70% of people are infected.

Elon Musk has reopened Tesla’s Bay Area factory. Although the rest of California is gradually reopening, the Bay Area is playing it extra careful and has asked everyone to stay home until at least June 1. Except, apparently, Elon Musk, who declared the factory was reopening regardless of what anyone said, and that “if anyone is arrested I ask only that it be me”. For some reason, the county did not arrest him, and now it seems to have retroactively legitimized Musk’s action. I like Elon Musk and I support the right to civil disobedience, but the government should absolutely have arrested him. They wouldn’t necessarily have to give him twenty years to life or anything, just arrest him enough to make it clear that there are laws and you get punished if you break them. [EDIT: see here for discussion of why he wasn’t arrested]

Nate Silver crunches the evidence and finds that (contra what I wrote last time) there is no evidence that voting by mail gives one party an advantage. So how come Democrats are so excited about it and Republicans so anxious to prevent it? Do they know something Silver doesn’t know? Or are they really and truly just concerned about their principles, with no ulterior motive?

The latest from EA on best ways to donate to the fight against coronavirus. Summary of the summary: Fast Grants and Development Media International. Fast Grants has a minimum donation of $10,000 (they are smart people and I assume there is a reason for this); some people were previously trying to pool their donations to reach this amount but I don’t know where the latest active pools are.

Coronavirus has killed 90,000 Americans so far. Donald Trump tried to put this in context by saying the seasonal flu sometimes kills 60,000 people a year. There are a lot of problems with this comparison, but one I didn’t realize is that coronavirus death toll only counts confirmed cases, whereas flu death toll counts estimated cases, ie a guess as to how many cases we would find if we had perfect detection. The number of confirmed flu deaths – a fair comparison to the 90,000 confirmed coronavirus deaths – is about 10,000 yearly (and remember the coronavirus hasn’t been around a whole year, or even a whole flu season). The article also is not convinced that the 60,000 flu death statistic is a fair attempt at estimating reality as opposed to a made up number that signals how much the CDC wants us to worry about the flu, and the author suggests the CDC officially lower the flu death toll in order to signal that we want people to worry less about it compared to coronavirus (how many simulacrum levels are you on? You are like a little baby, watch this…)