Revisiting the mistake

Back on April 4, I made this claim:

I also predict that in a month or two, when we have a good grasp on the likely death totals from this epidemic, or at least this wave of the epidemic, there will be news stories showing how many of those deaths could have been prevented merely by starting the social distancing at the national level at the same time it was done in Washington (which was about 2 weeks earlier). And the numbers will give you a sick feeling in the pit of your stomach. I don’t know the exact numbers, but if there are say 100,000 deaths, it might be the case that on the order of 80% could have been eliminated by starting two weeks earlier. And no, this is not a dig at Trump (or de Blasio), the country was not mentally prepared for doing something like this two weeks earlier. But it won’t change the reality of the fact that we as a society made a huge error. Social distancing may not even be the right answer, as some tough guys claim, but given that we ended up doing it anyway, not doing it two weeks earlier was super costly. Hiroshima costly.

And let’s not even talk about the tens of thousands of preventable deaths each year from kidney failure.

Time for some soul-searching?

Even as I typed “Hiroshima” I wondered if my claim would eventually look foolish. About 10,000 Americans had died at that point.

Today there are 94,000 official Covid-19 deaths and many additional unofficial deaths. The eventual toll will likely exceed 150,000.

But what about the preventability question. Could 80% of these deaths have been prevented at virtually no extra cost? Quite possibly.

I’m not sure it was accurate to claim that Washington began social distancing about 2 weeks earlier than other states; it’s much more complicated than that. But it did start social distancing somewhat earlier. And Washington has since gone from being by far the worst hit state to well down in the pack. Each week it slips further down the list in terms of deaths per million (right column):


There are other pieces of information. The combined population of Denmark and Norway is a bit higher than Sweden, and yet those two countries have barely 20% as many fatalities as Sweden. I don’t cite this figure to take sides on the Swedish policy, rather to indicate that if we were going to do social distancing anyway, then (in retrospect) it would have made more sense to start 2 weeks earlier. Two weeks is a long time when caseloads are exploding upward.

Tyler Cowen linked to another interesting study. Here’s the abstract:

We test whether earlier social distancing affects the progression of a local COVID-19 outbreak. We exploit county-level rainfall on the last weekend before statewide lockdown. After controlling for state fixed-effects, temperature, and historical rainfall, current rainfall is a plausibly exogenous instrument for social distancing. Early distancing causes a reduction in cases and deaths that persists for weeks. The effect is driven by a reduction in the chance of a very large outbreak. The result suggests early distancing may have sizable returns, and that random events early in an outbreak can have persistent effects on its course.

I view all of this as a collective mistake (including myself), which might explain the relative lack of moral indignation, at least given the death total. We prefer to point fingers at specific villains (as with Bush/Iraq, although even in that case Bush’s role is greatly overstated.)

Social distance and trash the economy?

or . . .

Allow a huge death toll?

There are good arguments both ways. But for the sake of God don’t do both!

Alas, we did both.

PS. Creating a kidney market is still extremely important. And on that issue I do point fingers (at deontological ethicists.)



29 Responses to “Revisiting the mistake”

  1. Gravatar of Michael Rulle Michael Rulle
    20. May 2020 at 11:06

    That would be very bad——but so much of what is written suggests otherwise. I have no idea what is true or is not–and maybe this would be correct if we get a vaccine very fast—–But—most theories suggest we are just spreading the deaths out—remember “prevent hospital overload”? It is as if the number of deaths are set in stone—its just when they occur—-but even as that has been written—can that be true? Yes–if you believe –a) no vaccine, and b) herd immunity and c) it does not go away until one or both happen. We will know for sure after it does not matter (hopefully sooner due to vaccine)–we know so little

  2. Gravatar of Ray Lopez Ray Lopez
    20. May 2020 at 11:25

    Another excellent Sumner op-ed. Part of the reason is humans are bad at assessing risk, and in the USA, unlike many other parts of the world, people are not used to dealing with tragedy on a national scale (the NY Times made this point, comparing Croatia, Greece, which know hardship, to the USA; the former are more prone to do lockdowns without complaint).

    Another reason is here:

    The Five Rules of Risk
    80,586 views • May 20, 2020

    1. Voluntary risk is more acceptable than involuntary risk

    2. Acceptance is inversely proportional to prevalence (the more people exposed to a given risk, the less it is acceptable)

    3. Disease is a yardstick (people do risky things based on what their chances of catching disease is, unrelated to the activity, i.e., with disease being the opportunity cost) @7:03 – fighting in Vietnam War was no more hazardous in terms of death than dying from disease on average (across the entire population, not just for young men, and notice death in Vietnam was rarer due to Medivac helicopters). Hence the public accepted young men dying in Vietnam.

    4. Novelty increases perceived risk (hence college kids in 1970s, before Three Mile Island, listed nuclear power as the riskiest activity, ahead of motor vehicles, alcohol and smoking, due to its novelty

    5. Numbers are numbing: people assume a logistic “S-shaped” curve for value of lives saved vs number of lives saved, when it should be either linear or arguably even exponential (when millions die, society breaks down). This is Stalin’s quip: “The death of one man is a tragedy, the death of millions is a statistic”.

  3. Gravatar of Aladin Aladin
    20. May 2020 at 12:36

    I’m not sure how much I trust the 2 week earlier claim … things happened when they happened. The flip side is, we could have started socially distancing 2 weeks later, and the crisis would be exponentially worse.

    It doesn’t seem like deaths are orders of magnitudes higher in areas with stricter lockdowns vs less strict lockdowns. So it is not a matter of public policy, because a lot of this is being driven by people choosing to self isolate and distance or not, not government making that decision for them.

    So if it’s a matter of how people behave, saying everyone collectively is going to start self isolating 2 weeks earlier is nonsense. Through what mechanism would that happen? Economics uses incentives to drive behaviors of large groups of people, what incentives would you impose?

  4. Gravatar of Plato’s Revenge Plato’s Revenge
    20. May 2020 at 12:43

    @Michael Rulle: I don’t think lockdown just spreads fatalities around, and not because of hospital overload. We’re bound to make progress treating the disease, even without a vaccine. The learning curve is likely very steep right now, and we don’t need hundreds of thousands of patients to pick these low-hanging medical fruits

  5. Gravatar of H_WASSHOI (Maekawa Miku-nyan lover) H_WASSHOI (Maekawa Miku-nyan lover)
    20. May 2020 at 13:17

    Compared to the totally useless atomic bombing of Hiroshima, which was dropped when Japan was already deciding how to end the war, COVID 19 was not as much of a lost profit for the economy as a whole, but it was expected to kill old people who were not working, so they decided to let it die, didn’t they? The roots are the same as those of Hiroshima in that the future utility of the individual is not considered. Maybe it’s not so much malice as just incompetence.

  6. Gravatar of dtoh dtoh
    20. May 2020 at 13:49


    Look at the social distancing metrics by state at the IMHE site. In virtually every state, social distancing preceded any government action, and the subsequent government action had almost no impact on the timing or shape of the curve of mobility (social distancing).

    If I were to hypothesize on causality, I would say.

    When the incidence of the disease rises to a certain level, it CAUSES people to socially distance themselves.

    When people begin to voluntarily socially distance, it CAUSES government to implement redundant and unnecessary mandates to force people to do what they are already doing voluntarily.

  7. Gravatar of dtoh dtoh
    20. May 2020 at 13:51


    When you were talking about Hiroshima, were you saying it was costly on a net basis (costs minus benefits were high.) Or just that the costs were high?

  8. Gravatar of bb bb
    20. May 2020 at 14:25

    @dtoh, Schools are a big factor. We along with most parents we know chose not to send our children to school on the last day they were open. I don’t know if people keeping their children home influenced the decision to close the schools, which came out later that day. But keeping our kids out of school without government action would not have been a desirable path for me. I think some government action is required, if for no other reason than for schools. And the externalities are huge.
    I do believe that the localities should have wide discretion. I have no problem with rural parts of my state choosing to open up more quickly, but I would be very upset if my local officials weren’t allowed to stay shut down a little longer. For instance, it be heartbreaking to me if Atlanta became the next hotspot because the governor imposed a state-wide opening on them, which might happen.

  9. Gravatar of @dtoh @dtoh
    20. May 2020 at 14:47


    I’m by no means a strict libertarian so I agree, but in general I think, like you, that authority should be devolved as far out locally as practical and that government should limit regulation to the most dangerous activities….bars, religious services, etc.

  10. Gravatar of bb bb
    20. May 2020 at 14:59

    @dtoh- well said

  11. Gravatar of ssumner ssumner
    20. May 2020 at 15:00

    Aladin, You said:

    “It doesn’t seem like deaths are orders of magnitudes higher in areas with stricter lockdowns vs less strict lockdowns.”

    I didn’t say they were. Voluntary social distancing is more important than lockdowns.

    On the other hand, Sweden argues against your claim.

    Wasshoi, Yes, incompetence.

    dtoh, I wasn’t making any moral judgment about Hiroshima. And I agree about voluntary social distancing, as I’ve indicated many times in other posts. The lockdowns are overrated. I tend to oppose mandatory lockdowns, except in the most extreme cases.

    bb. I support closing public schools. But I view that as government inaction, not government action. 🙂

  12. Gravatar of dtoh dtoh
    20. May 2020 at 15:09

    I wasn’t asking about your moral judgment. I was asking about your utilitarian judgement.

  13. Gravatar of Ray Lopez Ray Lopez
    20. May 2020 at 15:18

    Here’s the graphic that shows what dtoh is talking about, with ‘voluntary’ social distancing in all states even before government mandated quarantines.

    That said, NY state could have saved more lives had they been an early closing state rather than a late closing state.

  14. Gravatar of Christian List Christian List
    20. May 2020 at 16:40

    I think the closure of the nursing homes was a key measure. In the US, 50% of the dead seem to come from nursing homes, in Sweden the figures seem to be similar, and in Canada as much as 80%.

    Germany and Austria have sealed off nursing homes and hospitals. Visits were completely forbidden for weeks, from a very early stage. Since 1-2 weeks visits are allowed again, but only one visitor per patient, and you have to register, plus gloves and face mask etc.

    This is a method which does not harm the economy and which seems to be rather effective.



    any news on your Australia vs. New Zealand bet yet?

  15. Gravatar of Benjamin Cole Benjamin Cole
    20. May 2020 at 17:41

    Meh. You can look at local, regional or national fatality rates from C 19 and draw your own conclusions, based on your priors. It is a cousin industry to macroeconomics.

    Sweden is doing about the same as Great Britain.

    Does Southeast Asia already have some sort of herd immunity or were they particularly successful in blocking the virus from ever entering the region? The latter seems improbable giving given limited state capacity and the necessity of most people to keep working.

    Even in those areas that have suppressed the virus, the exit strategy remains problematic. You then have a naive population and a novel virus as a standing situation.

    There is an old French expression, “It is urgent we do nothing.”

  16. Gravatar of BC BC
    20. May 2020 at 22:35

    “Could 80% of these deaths have been prevented at virtually no extra cost?”

    To evaluate the true cost of preemptive social distancing, we must consider what would have been the cost of preemptively social distancing for two weeks for other potential pandemics (SARS-1, H1N1, swine flu, bird flu, etc.) only to find out that the preemptive social distancing hadn’t been necessary after all (assuming that we wouldn’t have been bombarded with claims that the preemptive social distancing was what prevented an outbreak from emerging in the first place).

    Maybe, only NY/NJ needed to start two weeks earlier since most cases across the country have been linked to NY. Maybe, Washington’s early social distancing is what saved CA?

    Another risk of preemptive social distancing is that one burns through one’s social distancing budget two weeks earlier. Loosening social distancing two weeks early could be worse than starting two weeks later. We’ll have to see how Denmark/Norway compare to Sweden after they have completely emerged from lockdown and achieved similar immunity to Sweden. One can’t judge a marathon after the first few miles.

    Also, excessive social distancing can have a backlash effect of perhaps paradoxically reducing well-targeted social distancing. For example, the unreasonable distancing policies in Michigan triggered a wave of protests and may lead to less voluntary and useful social distancing by provoking negative sentiment and less trust? Similarly, premature social distancing could potentially lead to premature ending of social distancing, i.e., starting two weeks earlier might lead one to stop more than two weeks earlier.

    Finally, maybe we need to exclude nursing home deaths in considering the incremental value of society-wide social distancing beyond measures targeted to nursing homes.

  17. Gravatar of Postkey Postkey
    21. May 2020 at 00:02

    In the UK.
    “An earlier lockdown could have prevented three-quarters of UK coronavirus deaths, modelling suggests
    Researchers say that if the UK had imposed lockdown seven days earlier, its death toll would be on a par with 8,000 in Germany”

    Was it the ‘government’s’ ‘fault’?
    “While critics have slated the Johnson government for not following the science, the real problem seems to be that he did. As a special report by Stephen Grey and Andrew MacAskill at Reuters explains: . . . ”

  18. Gravatar of Scott H. Scott H.
    21. May 2020 at 02:37

    Mr. Sumner you wrote:

    “rather to indicate that if we were going to do social distancing anyway, then (in retrospect) it would have made more sense to start 2 weeks earlier.”

    Yes, it’s too bad nobody had thought of a plan about what to do if a deadly, contagious virus starts to break out. Otherwise back in January (January!) we could have written “Well, if R0 is above X then that means we’ll need to shut down international travel and quite possibly the national economy.”

    Why couldn’t we write the first sentence of a plan?

  19. Gravatar of Michael Rulle Michael Rulle
    21. May 2020 at 05:30

    I am pretty sure most of Scott’s readers have looked at distributions in various statistical studies. I am starting to believe we have all been starting at the left tale—-like around 3+ standard deviations in—-thinking we are looking at the whole distribution.

    For example, 99.9% of Americans have not died from Covid. Also, 99.9 % of Brits have not died from Covid. Same with Japan.

    99.99% of Americans did not die from Covid if under 65. Now imagine using the “lives saved” for what we have done! That is a good one.

    You get the point. When the tail is so tiny—(-I need to look up what Deirdre McCloskey thinks)—-despite my own obsession with death rate differences between Japan and US, —-is trying to determine what is even in the tail so small, we are bound to make serious errors?

    I purposely am being statistically idiotic——but am I really being idiotic? We do know that 80% of people who were working are now not working. (63% versus 51% I believe are the relevant numbers).

    I have lost track of how many trillions we have committed to this thing. Let’s call it 4 trillion. So we were spending 83.3% of what we are spending now.

    So, perhaps rather than me focusing on “counting and politics” and locking down two weeks earlier versus 2 weeks later——maybe I need to keep it simple

    Whoever thought of this lockdown idea (some like to blame Neil Ferguson) and tossed out the word “exponential” to an innumerate humanity perhaps has caused this extremely dangerous situation.

    Can we put the air back in the balloon—-or has the balloon popped?

    All should be permitted that was permitted before. Let’s all of us morons decide for ourselves.

  20. Gravatar of Michael Rulle Michael Rulle
    21. May 2020 at 05:38

    PS—-20% not working, 80% of before are working.

  21. Gravatar of Michael Rulle Michael Rulle
    21. May 2020 at 06:03

    PPS—-Ms. Mccloskey seems to think it obvious lockdown was good—-but she also fears we are flirting collectively with “Coercionism”—her Word for socialism. So—-not sure what that means for policy—-As it appears we have somewhat engaged in a trade off that may not be totally in our control——my sense is she is “Scott like”. We can recover fast once we decide it is safe to do so. But she does not define what that means—-but she likely thinks it will be obvious when we see it.

  22. Gravatar of Michael Rulle Michael Rulle
    21. May 2020 at 06:30

    PPPS—-more odd stats

    Canada says 80% of Their Covid deaths are in “nursing home”

    NY Times says 1/3 of US deaths in Nursing Homes

    PA says more people over age 95 died than people under 60 from Covid.

    The median age of the Lombardy death wave was 81.

    The secondary counting problem of the old dying, is the high number of Covid-morbidities. It would not be bizarre to say Covid is not as bad as flu—-if we looked only at those under 65.

    But how does one protect the old? What special way is that done? I cannot think of any—-other than extreme measures that are unacceptable.

  23. Gravatar of ssumner ssumner
    21. May 2020 at 13:24

    dtoh, They are the same.

    Christian, What did I bet? I forgot.

    BC, Again, I wasn’t talking about public policies. I was talking about practices.

    Scott, That’s way too cryptic for me.

    Michael, For the Nth time, comparing actual deaths to actual financial cost is meaningless. We have BOTH.

    Try comparing deaths with no social distancing to financial costs with social distancing.

  24. Gravatar of anon/portly anon/portly
    21. May 2020 at 14:11

    “And Washington has since gone from being by far the worst hit state to well down in the pack. Each week it slips further down the list in terms of deaths per million….”

    And King County is no longer even the worst hit county in Washington – Yakima and Benton (2/3 of the Tri-Cities) counties, on the East side of the state, are now the worst.

    And the Positive Test rate at UW Virology keeps dropping – a couple of days below 3% this week. The days when it was over 10% seem like weeks and weeks ago. (And they were).

  25. Gravatar of Scott H Scott H
    21. May 2020 at 14:54

    Maybe this will make my point clearer:

    If we had a plan for a pandemic, then all we had to do is to pull out the plan to see what to do. We could have been sitting here saying “Oh snap! This looks like we’re going to shutter international flights and shut down the economy because that’s what the plan says to do when R0 is above a certain number and mortality rates are above a certain number”.

    Why didn’t we have that conversation? Because there wasn’t any plan. They never even got the first and most fundamental if/then statements of a plan off the ground. This is total dereliction of duty for the hodgepodge of agencies that were charged to create plans.

    We can talk about what Trump, you, me, and the state of Washington should have done when caught unawares. But the elephant in the room is that there was no plan.

  26. Gravatar of Scott H. Scott H.
    21. May 2020 at 15:00

    Curse you auto fill! Can I get my Name edited back to Scott H. on that other post please. Thank you.

  27. Gravatar of dtoh dtoh
    21. May 2020 at 15:42

    If they’re the same, then by saying it was costly you have made a utilitarian judgment and by extension a moral judgment.

  28. Gravatar of ssumner ssumner
    21. May 2020 at 16:50

    anon, Interesting.

    Scott H. I agree.

    dtoh, Not without considering counterfactuals. All options were costly (at Hiroshima), and in dealing with the coronavirus.

    Maybe we are talking past each other. I just mentioned Hiroshima to give a sense of the death toll, not to make a moral judgment on Hiroshima. If one supports the bombing, then one could argue the Japanese government is ultimately to blame for the “cost”.

  29. Gravatar of dtoh dtoh
    22. May 2020 at 16:18

    The point I was making is that you need to look at net cost (benefit – cost) and you need to compare the net cost against the net cost of other alternatives.

Leave a Reply