Extreme events generally have multiple causes

There were a number of good responses to my previous post, which asked why the death rate in Italy was 100 times higher than in Scandinavia. One answer was that Italy got its cases earlier, and thus patients have had more time to get seriously ill. This hypothesis is supported by the fact that the ratio has already fallen to more like 50 to 1.

But that’s probably not the entire story. France and German cases track each other extremely closely:

And yet France has 91 deaths while Germany has just 8.  In this case, I assume that France has a larger share of infected people who are over 70 years old.

When comparing Scandinavia and Italy, I suspect at least three important factors:

1. The Nordics who are infected tend to be healthy people who traveled to Italy, whereas the Italian deaths are mostly among elderly people who are less likely to travel.

2.  Italian people infected with the virus have had the virus for a longer period of time, as noted above.

3.  A larger share of infected people in Scandinavia have been tested.

Another possible factor is worse medical care in Italy due to hospital overcrowding, although I suspect that’s only a marginal factor thus far.

The lesson here is that extreme events usually have multiple causes.  This is also true in economics.  If extreme events had single causes, then presumably they’d occur much more often.  In my book on the Great Depression I argued that there were multiple policy errors, on both the supply side and the demand side.

People younger than me won’t remember Beamon’s Leap, but I do:

While Beamon received mostly accolades, there also were detractors. The critics harped on the conditions — a following wind of 2.0 meters per second (the maximum allowable velocity for a record), a lightning fast runway and, most important, the thin air of Mexico City. Beamon’s defenders point out that the other competitors, which included the world record co-holders, had the same factors going for them and they didn’t jump close to Beamon.

What a killjoy!!

PS.  Trump should immediately remove the travel ban from China.  We are now in far more danger from visits by Canadian tourists than visits by Chinese tourists.  This should be done even if the original ban was entirely justified (which seems plausible.)

PPS. Over at Econlog, I discuss Boris Johnson’s recent policy changes, and also explain my claim that the US will never experience hundreds of thousands of deaths from coronavirus.

PPPS.  It’s early, but there is some evidence that Norway and Denmark are beginning to get a handle on the crisis.



37 Responses to “Extreme events generally have multiple causes”

  1. Gravatar of Andrew Clough Andrew Clough
    14. March 2020 at 13:08

    Why do you think that hospital crowding is a marginal factor? Given the stories of triage I’m hearing out of Italy I could easily see it being a factor of 2 or more.

  2. Gravatar of ssumner ssumner
    14. March 2020 at 13:55

    Andrew, You may be right. I think it’s less important than the other factors, but perhaps I underestimated it.

  3. Gravatar of Håvard Ihle Håvard Ihle
    14. March 2020 at 14:23

    Regarding the encouraging numbers from Scandinavia the last few days, the numbers may be a bit misleading as (at least here in Norway) we are now only testing people with serious (I believe this means basically respiratory) symptoms and people in important professions.

    We recently closed all the schools and kindergartens, introduced more strict quarantine rules and are basically closing down all travel into the country. I think they changed the testing strategy after implementing these changes.

    I do believe these new measures will have a large effect on the numbers, but that effect will probably not be that visible until after a week has passed.

  4. Gravatar of Steve Steve
    14. March 2020 at 14:33

    Sumner has a habit of posting fake news on the virus.

    Mortality rates should range from 0.5% to 6.0% depending on access to medical care. Top factors include: 1) ventilator supply 2) drug supply 3) right-to-try

    We are already seeing that range play out, from ~0.5% in Beijing, to 3.0% in Wuhan, to possibly 6.0% in Italy and Iran.

    Infectious diseases are medieval in nature and advanced hospital systems offer little benefit over scalable simple solutions. Oxygen, ventilators, dialysis, plasma transfusions (from survivors), chloroquine, kaletra, remdesivir. Probably a few others I’m missing.

    If the virus isn’t contained, it is likely every country will lose 3% of its population. ~50% infected, ~6% dead, many more with organ damage.

    The are enormous statistical artifacts in the range of results being seen elsewhere, including shortages of tests, who qualifies for tests, the long 3-5 week duration of the disease, exponential growth of infected sans containment, and access to medicine. Virtually all the variation in reported mortality can be explained by these factors.

    As for why Italy got hit so hard so early? It might be bad luck of a superspreader in January. It also might have to do with religious practices and multigenerational housing. The same applies to Iran.

  5. Gravatar of Mads Lindstrøm Mads Lindstrøm
    14. March 2020 at 14:34

    Also in Denmark, only the seriously ill are tested.

    Anyway, given an incubation period of 2 to 14 days, one should only really see results after 2-4 days, and the signal will be very weak. When the testing also changed, it just seems impossible to know if the measures has worked or not. But give it a week or 10 days more and we will know.

  6. Gravatar of Mads Lindstrøm Mads Lindstrøm
    14. March 2020 at 14:35

    “Also in Denmark, only the seriously ill are tested.”. That is, they changed the testing two days ago to only test the seriously ill. Sorry, about not making that clear in the first comment.

  7. Gravatar of Michael Rulle Michael Rulle
    14. March 2020 at 15:55

    Beamon’s leap was shocking at the time. It made no sense. The mph was irrelevant and the altitude might have had a few inch benefit (see Hamlin, Hopkins, Hollings—2015 “Effiects of altitude on Performance…….). On the same day in Tokyo in 1991, Carl Lewis and Mike Powell both exceeded 29 feet and are still 1st and third all time. Beamon’s jump is very likely the greatest track and field performance of all time.

    PS—Hospital Crowding.

    Here in NJ hospitals are empty, far lower than normal. We will know in a month just how bad it will get. We are still in the catch-up phase of counting so doubling is happening pretty quickly. It’s not really doubling, the counting is resulting in doubling. Which is why we need randomized testing weekly. No catch-up—-just actual change.

  8. Gravatar of Michael Rulle Michael Rulle
    14. March 2020 at 16:22

    Read your essay on The Hill. Everything you said is what you have been always saying and I believe you.

    I am now going to annoy you.

    The Hill gets a reasonable number of clicks. You need to strap on your propaganda hat—or if you prefer a gentler word “persuasion” hat. I like that you criticized Summers—-perhaps make note he has been away from academia too long and has spent the last 20+ years with Hedge Funds and in politics. Name people who should support your view (called “appealing to a higher authority”) like Bernanke for example and others.

    People don’t believe in markets—-using them as a reason convinces no one—-unless you provide support from others. I think you believe the following—-you are fighting an actual economic war—-you need to recruit troops. Why did all your readers respond so positively to your battle cry a few essays ago?

    This is serious stuff. The Hill will let you Propagandize (which does not mean lie, but persuade aggressively) so next time go for it.

    This is really serious stuff, right/


  9. Gravatar of Michael Rulle Michael Rulle
    14. March 2020 at 16:28

    PS—Of course I “believe’ in markets—-but you must realize most people don’t—and they don’t even know why they don’t—-they may say something like “behavioral finance” bla bla—-in fact, while supporting what markets tell us, work in some “behavioral finance”——there must be something they say that fits:-).

  10. Gravatar of Benjamin Cole Benjamin Cole
    14. March 2020 at 16:49

    I think when it’s all over, Great Britain will prove out as the model to follow.

    For reasons that are mystifying, the House has rejected the Trump Administration plan for a payroll-tax (FICA) holiday through the remainder of 2020.

    The Federal Reserve is armed with popguns.

    This does not look good for the economy in 2020, nor the stock market.


  11. Gravatar of Aleksander Aleksander
    14. March 2020 at 17:09

    “France and German cases track each other extremely closely”

    What do you think this means? I think it means they have tested a similar number of people. That could be because they are both testing most of their infected. But it could also mean that at least one of them is only testing a fraction of their infected population. The discrepancy in death numbers that you cite indicate the latter.

    In fact, unless the real death rate is ridiculously high, or we have completely misunderstood the incubating length or something, current rates of deaths/day suggest that both countries have only tested a fraction of their infected population.

  12. Gravatar of Christian List Christian List
    14. March 2020 at 17:28


    great analysis on Econlog. I don’t know if Johnson has changed his “strategy” yet, but he will have to do so in very few days indeed.

    I read a CNN report on the UK a few hours ago and Johnson’s “strategy” is shockingly chaotic at the moment. Sometimes he says so, then he says the opposite. This has Trumpian dimensions. And it will end up like Trump: with a big bang.

    Really nice, very accurate post, and new nice ideas, too. But don’t be too harsh on Scott. His corona posts have improved a lot in just a few days. He is a marvel of learning.

    I don’t know how the French test, but we Germans test to a rather limited extent. The test kits are finite, so at the moment we only test people where there is concrete suspicion. So usually people who first of all have symptoms and secondly have had known contact with Covid-19, either directly through a person or through a risk area from abroad.

    The criterion “risk area from abroad” is becoming more and more questionable with each passing day, because large parts of Germany are now a risk area themselves, but are not evaluated in this way in the test procedure.

  13. Gravatar of Cove77 Cove77
    14. March 2020 at 18:58

    Johnson has back tracked on stadium crowds, NI diverging from Republic re schools though that should change in a few days

  14. Gravatar of Christian List Christian List
    14. March 2020 at 19:43

    The more I think about it, the more important your last point “share of infected people tested” seems to be.

    Scandinavia must have the highest share of infected people tested.

    With the difference between the factors concerning the death rates one can calculate relatively exactly how many infected people countries like Italy, France, Spain must have missed.

    This also explains why the virus is spreading so fast.

  15. Gravatar of Aleksander Aleksander
    14. March 2020 at 20:51

    Christian List:
    “we Germans test to a rather limited extent.”
    That’s how pretty much all countries do it. South Korea claims to be doing much more extensive and systematic testing, and their “confirmed cases” number is also the only one I know among countries with many deaths that is somewhat compatible with the death numbers.

    But even with the South Korean case number, I don’t understand why the raw number is so widely publicised. The real number is obviously higher. Why don’t the South Korean experts just publicize an estimate of the actual number of cases, based on randomised testing of representative samples?

  16. Gravatar of Aleksander Aleksander
    14. March 2020 at 21:09

    Christian List:
    “Scandinavia must have the highest share of infected people tested.”
    Yes, I think this is plausible. Norway, for example, have had cases for several weeks, but only had their first death a few days ago. I haven’t run the numbers, but I still think they have one of the lowest deaths/confirmed cases ratios in the world (about 0.25%).

    “With the difference between the factors concerning the death rates one can calculate relatively exactly how many infected people countries like Italy, France, Spain must have missed.”
    If the Scandinavian death rates are accurate. But you should wait a few more weeks until more people have died there, as most infections are still likely too recent. I still think the best sources of mortality rate are Diamond Princess (1%) and South Korea (0.9%, although they claim the upper bound is 0.6% based on studies there).

    But yes, it’s not that hard to estimate number of infected based on mortality rate and deaths/day. There are, for instance, clearly more than 100,000 infected Italians, and tens of thousands of infected Spaniards. It’s just a little hard to confirm whether your calculations are right, even after the fact.

  17. Gravatar of ssumner ssumner
    14. March 2020 at 21:33

    Havard and Mads, thanks for that info. Of course it’s also possible that behavioral changes from a week back are beginning to show up in the data, even before effects of new policies. But I take your point.

    Michael, Yes, I was only 12 at the time, but had enough instinctual understanding of statistics to know the jump was special.

    Ben, You said:

    “Great Britain will prove out as the model to follow.”

    LOL. You still don’t know they’ve already given up on the strategy?

  18. Gravatar of Benjamin Cole Benjamin Cole
    14. March 2020 at 22:14

    Scott Sumner-

    Perhaps the Brits are changing on stadium crowds. Yes, science is bending to hysteria…something new? Are you on the side of the hysterics or the scientists?

    Let’s see if the Brits keep schools open, and in general follow a strategy of moderating towards herd immunity. I guess the Brits anticipate an 80% infection for the country as a whole, eventually.

    What do you think the infection rate was for crew members of the Diamond Princess?

    Covid-19 is described as highly infectious. The Diamond Princess crew ate and slept in common quarters on a prison-ship for few weeks. My guess is 100% infection rate (that means they have antibodies to Covid-19 in their bloodstream). A handful of illnesses worth mentioning, and a 0% death rate among the crew.

    In the US, we can destroy our economy and collapse the financial system again, but eventually we will see the same migration towards herd immunity. It is likely inevitable. After a lockdown, then what? Every lockdown ensures you have carriers and fresh infections, leading to the next lockdown. I guess each lockdown will be justified with photos of “first responders” in hazmat suits and flashing lights.

    Well…perhaps the US will create a permanent national health-security complex, ala the global military security complex the US already has. The health-security complex will forever be stamping out coronavirus….as herd immunity is never obtained.

    I suspect the second shoe has yet to drop on Wall Street, and if so, and if corporate bonds begin to crumple, then we see other financial institutions tumble too.

    I hope you are short on equities.

  19. Gravatar of msgkings msgkings
    14. March 2020 at 22:57

    One thing that really worries me about all this…let’s say the likely outcome is we muddle through this, deaths only like a bad flu season…so the lesson learned is ‘shut it down’ works.

    Are we going to stop the country/world every time there’s a viral outbreak? Can we afford to do that every couple years?

  20. Gravatar of ChrisA ChrisA
    14. March 2020 at 23:08

    Re the disparity in South Korea and Italy deaths, the answer appears to be due to the age of the people who are infected, see this analysis here which appears convincing to me: https://medium.com/@andreasbackhausab/coronavirus-why-its-so-deadly-in-italy-c4200a15a7bf

  21. Gravatar of Ray Lopez Ray Lopez
    15. March 2020 at 00:28

    @readers – note Sumner’s cheap tricks:

    1) attacking posters that don’t agree with him with fake news, specifically, Ben Cole correctly anticipated the UK strategy and now Sumner says overnight (as in Saturday) the UK has abandoned that strategy because they modified it slightly by possibly banning stadium crowds. Sumner hates to lose so he makes up stuff to ‘win’ online. Once he even got his daughter and wife involved by mentioning them in an obscure internet dispute I had with him to win the argument; how low can one stoop?

    2) playing the race card: “Nordics” more healthy that Italians, Germany > France, etc. The fake, pretend German “Dr” List enjoyed that, but even he was embarrassed enough to backtrack a bit on why Germany has done better than France. It must be embarrassing to be a German? Or maybe not (which is itself embarrassing, white socks with sandals all).

    3) Sumner opines on things outside his narrow area of expertise, and that would be NGDPLT with certain restrictions like assuming strong money non-neutrality (boy, talk about a narrow field of expertise). When I was involved with litigation that was a fool’s strategy. About as foolish as me holding myself out as an expert in NGDPLT (which I don’t do, I just point out inconsistencies).

    @msgkings – good point, and indeed that’s one of the biggest dangers of “Big Government” and crisis. If it works due to chance alone (Covid19 burns out, as did SARS, more or less on its own) then Big Govt takes credit for it; if not, they double down on even more restrictions. One reason that government cracks down on liberties even during peacetime (e.g., permits needed for protests, antitrust takeover review approval for certain national interests, banning largely harmless companies like Huawei under hypothetical national security concerns, etc) under the theory that they need control in case there’s war, aka “wartime powers” as it’s known in law.

  22. Gravatar of Benjamin Cole Benjamin Cole
    15. March 2020 at 01:55

    Achtung! Christian List (and anybody else).

    A fellow Teutonic traveler talks common sense:


    You got to love that accent.

    The lockdown in Hamburg means you can’t go to a concert but you can ride a crowded train or bus.

    When Donald Trump becomes the voice of reason in this world… liquidate everything and run for the hills!

  23. Gravatar of Riccardo Riccardo
    15. March 2020 at 04:23

    Scott, you mostly make good sense (to me) over a wide range of topics. But when you say “the US will never experience hundreds of thousands of deaths from coronavirus” and then go on to explain it’s because the Republicans would never allow such an outcome in an election year… well, doesn’t that assume they have the chops to do what’s needed to “flatten the curve” before there’s a healthcare meltdown? For several years you’ve been pointing to how Trump & Republicans have gutted key competencies in the government, saying that the US is becoming / has become a Banana Republic. How many Banana Republics can correctly respond to something like this, something we’ve never seen in our lifetime? It’s not like Trump can now suddenly push a button (regardless of how much money he finds to throw around) to marshal the resources needed to make a difference because it takes real grownups to do the hard work of implementing effective plans. For example, did you see the photos yesterday of incoming travelers being queued for hours at major airports around the country? Enforced overcrowding of thousands in small spaces at the worst possible time, not for any health screening, but because the bureaucracy did not modify its standard immigration and customs processing in anticipation of the predictably large influx of returning American travelers. Or what about last night’s White House news conference where the Surgeon General calmly assures everyone we’ll be fine because the COVID-19 fatality rate is low, and especially low for young people – implying that we only need to wash our hands as a public service to keep Granny safe – completely missing the point that it’s not the fatality rate per se that makes this pandemic dangerous but the fact that its exponentially growing caseload will soon overwhelm all healthcare services. If the contagion is not quickly slowed way down, the 5-10% of those infected who need medical care won’t be finding any, and the “low” death rate will likely go higher (nevermind that 1% of, say, 1/3 of the US is a already a huge number of dead). But the Whitehouse clown show last night basically told young people it’s Ok for them to keep on clubbing among themselves so long as they avoid visiting grandma. It simply could not have been worse messaging at a time when every day matters to reduce the coming impact. So I hope you’re right and there won’t be hundreds of thousands of deaths. But if we dodge the bullet it will have nothing to do with Trump & Republicans deciding they can’t afford it in an election year. They are simply in way over their head.

  24. Gravatar of Benjamin Cole Benjamin Cole
    15. March 2020 at 04:35

    I wonder if some other governments are doing the Great Britain thing, but not saying so.

    For example, you can’t go to a concert in Hamburg, but you can get on a crowded train or bus.

    Now, really…

  25. Gravatar of Michael Rulle Michael Rulle
    15. March 2020 at 05:37

    @scott sumner

    I have mentioned this a few times. Do you think it is a good idea to do formal randomized testing to estimate how many have the virus? My thought is—-we are currently “counting” new cases——thus conflating—-or not seeing——the difference between true growth rate of cases versus the growth of counted cases.

    Even sample sizes as low as 1000 (prefer 3-5k) would give us a picture of “today’s” rate of infection. Doing this weekly or more frequently would be a more accurate picture of the true growth rate of the number infected.

    Do you agree or not? Please give reasons! Thanks

  26. Gravatar of Michael Rulle Michael Rulle
    15. March 2020 at 06:34

    Re: Why Lawrence Tribe is the worst kind of hack

    Trump announced Fed partnership with Google etc. Simultaneously, Verily, a sub of Alphabet announced it was doing something similar for the Bay Area. The phenomenally corrupt media (not all media—the phenomenally corrupt ones) leaped to say “Trump Lied”. Tribe jumped right on board.

    Google then went national to say what was obvious—-they are doing what Trump announced.

    Think how stupid Tribe really is——-for him to think the entire WH would be that dumb or grotesque to lie about that—-when they are appearing daily to communicate what they are doing.

  27. Gravatar of Michel Rulle Michel Rulle
    15. March 2020 at 06:53

    PS. Tribe does not appear to know the difference between Alphabet and Google. What a corrupt jackass.

  28. Gravatar of Michael Rulle Michael Rulle
    15. March 2020 at 06:55

    PS—forgot you were moderating—I support it. Re Tribe—-he appears to not be aware of the difference between Alphabet and Google. Jeez.

  29. Gravatar of Patrick R. Sullivan Patrick R. Sullivan
    15. March 2020 at 07:39

    “For several years you’ve been pointing to how Trump & Republicans have gutted key competencies in the government, saying that the US is becoming / has become a Banana Republic. How many Banana Republics can correctly respond to something like this, something we’ve never seen in our lifetime?”

    I laugh.

    But what can any government do, really? As someone else mentioned, people can’t really isolate themselves from one another. If they don’t go to NBA games, they’ll go somewhere else…where they will mingle with strangers. In a nation in which 80% of people live in cities, how do you ‘socially isolate’?

    Anyway, Joseph Sternberg had an interesting piece in the WSJ recently;


    Italy lags other large European countries in provision of acute-care hospital beds, furnishing 2.62 of them per 1,000 residents as of 2016, according to the Organization for Economic Cooperation and Development. In Germany it’s 6.06 and in France and the Netherlands it’s 3.15 and 3 respectively. That year, Italy devoted around $913 per capita to inpatient acute and rehabilitative care, compared with $1,338 in France, $1,506 in Germany, and $1,732 in the U.S.

    U.K. policy makers understand what such analyses portend—because underinvestment in Britain’s creaking health-care system is even worse. The U.K. spent the princely sum of $901.70 per capita on acute care in 2016, according to the OECD. British data don’t distinguish acute-care beds, but a comparison of available beds overall isn’t any more favorable to the U.K. (or to Italy). In 2017, when Germany provided 8 beds per 1,000 residents and France offered 5.98, Italy managed 3.18 and the U.K. only 2.54.

  30. Gravatar of ssumner ssumner
    15. March 2020 at 08:44

    Ben, “Science” has been saying for quite some time that global warming and coronavirus were serious threats. Idiots like Trump have been denying the scientific consensus. Which side have you been on? Let me guess.

    ChrisA, Yes, that’s part of it. But not all.

    Riccardo, I agree that Trump is as incompetent as you say, but we still have a first world infrastructure, which has not been totally destroyed. Trump will be a bystander in the effort to control this epidemic, he’s simply not a factor. Much will occur at the state and local level.

  31. Gravatar of ssumner ssumner
    15. March 2020 at 08:49

    Michael, We should prioritize doing as much testing as possible. Instead we have symbolic gestures like travel bans.

  32. Gravatar of Michael Rulle Michael Rulle
    15. March 2020 at 09:43


    I am feeling shock. Was watching Cuomo. He was trying his best and no hate toward him. But I literally had no idea at all what he was saying——-heard various things like “no ventilators”, no beds, 6 foot rule——

    On Friday Ohio said they thought 100000 people have (an estimate) the disease in Ohio—-???!!!. Isn’t that good? Zero deaths. Or did they all just get it last week. Now they are backing off from that—-great

    Watching some O’Hare airport person freaking out over how crowded it is——really? No kidding.

    It is literally all corona all the time—-and people are literally saying different things. 57 people have died.

    The only thing we can do is wait ——but during that time as 50 deaths grow to 3000 we are going to go crazy——

    I still,cannot,envision an exit strategy. My daughter lives in Brooklyn. A building with 500 people. They have been told they “may” have been exposed—-the building, not her. What does that even mean? So they are self quarantining.

    I think politics has forced politicians to feel like they all need to respond and be in front of cameras. This is a disaster.not the disease. The panic per disease ratio will shatter all records ever.

  33. Gravatar of Ray Lopez Ray Lopez
    15. March 2020 at 09:57

    @Michael Rulle – calm down, you’re hyperventilating, and none of your posts make sense. Relax, Sumner is not moderating these posts, if he did I’d be long gone by now.

  34. Gravatar of Aleksander Aleksander
    15. March 2020 at 11:02

    Michael Rulle:
    “Even sample sizes as low as 1000 (prefer 3-5k) would give us a picture of “today’s” rate of infection.”
    This would only work if more than 1/1000 of the population was infected, which is only plausibly true in Italy and San Marino. Maybe Spain.

    But if you don’t need daily snapshots, I support this method. 50,000 tests on a representative sample in the course of a week or so should do the trick. I’m sure lots of countries are doing this small-scale in specific regions.

  35. Gravatar of Mads Lindstrøm Mads Lindstrøm
    15. March 2020 at 12:59

    Scott Sumner:

    “Michael, We should prioritize doing as much testing as possible. Instead we have symbolic gestures like travel bans.”

    I do not see how having a travel ban would explain the lack of testing at all. I would think, it would be two different parts of the federal government. Perhaps I am overinterpretating the “Instead” in your sentence above.

    What is more, did Chinese travel ban not postpone the outbreak by a couple of weeks? Maybe I am overestimating the effectiveness of the travel ban, but it do seem a good strategy to limit travel from highly infected places. If not the travel ban, how do you explain the still low (compared to Europe) us numbers?

  36. Gravatar of ssumner ssumner
    15. March 2020 at 13:20

    Mads, You said:

    “Perhaps I am overinterpretating the “Instead” in your sentence above.”

    Yup. The term ‘instead’ does not mean “because”. We could easily do both, as you suggest. But we aren’t.

  37. Gravatar of Jason Jason
    15. March 2020 at 13:36

    Regarding Denmark, the first 300 or more cases were all Danes returning from ski trips Austria or Italy, which well explains why the cases so far are skewed towards younger people. The measures taken may or may not be enough to keep it that way. We’ll know more in a week or so.

Leave a Reply