“The” fatality rate for Covid-19

I use scare quotes because obviously “the” fatality rate for Covid-19 varies widely between countries. In Italy, 23% of the population is over 65, and only 1.1% of Covid-19 deaths have been people younger than 50. In some African countries the share of population over 65 is only 2% or 3%, so obviously they’ll have much lower infection fatality rates than Italy. Nonetheless, it’s worth looking at some data:

Weekly testing by the UK Office for National Statistics (ONS) indicates that 6.8 per cent of the population in England has been infected (and thus developed antibodies to the virus). The latest national survey in Spain published last week indicated that 5.2 per cent of the population had been infected.

Based on current fatalities, those ratios imply a fatality rate of roughly 1.1% in each country. But that underestimates the fatality rate for two reasons. A few of those currently infected will eventually die, and some Covid-19 deaths went unrecorded. So the actual fatality rate is roughly 1.2% to 1.4% in the UK and Spain.

One of the highest confirmed rates was in New York City, where it hit 22.7 per cent, according to a study by NY State Department of Health and NY State University at Albany.

This also implies a fatality rate of roughly 1.1%

On the other hand, all three of these rates seem to be above average:

The “infection fatality rate” — the proportion of those infected who die — depends on local circumstances but is typically in the 0.5 to 1 per cent range. A study by Imperial College London of the epidemic in China found an IFR of 0.66 per cent. Meta-analysis by Australian epidemiologists who pulled together 25 studies around the world calculated that the average IFR was 0.64 per cent.

What explains the difference? First, Western Europe and NYC have older than average populations, so you’d expect a higher case fatality rate. The percentage over 65 in Spain (19%) and the UK (18%) is much higher than China (11%). In places like Singapore and Qatar the illness has hit young and healthy migrant worker communities, and the fatality rate is far below 0.1%.

I’ve generally assumed that the fatality rate in the US (where 16% are over 65) is about 1%, and I’m sticking with that claim. I view claims of 0.3% fatality rates as way too low for the US. Some were based on flawed studies of infection rates in California.

If the US public had not taken any measures to avoid Covid-19 then 100 to 200 million would have become infected and 1 or 2 million would have died. But that was never going to happen, as people do take measures to avoid infection. You can’t do epidemiology without economics. There is lots of social distancing even in countries such as Sweden, which never had a mandatory lockdown.

From the beginning, the debate over lockdowns has been a distraction from the issues that really matter—voluntary social distancing, behavior changes (no hand shaking, washing hands), near 100% mask wearing in crowds and public indoor spaces, test/trace/isolate, etc.

PS. In a choir practice in Washington State, 52 of 61 attendees came down with Covid-19, so I don’t believe claims that most people are not susceptible.

A COVID-19 superspreader unknowingly infected 52 people with the new coronavirus at a choir practice in Mount Vernon, Washington, in early March, leading to the deaths of two people, a new Centers for Disease Control and Prevention (CDC) report finds.

This was an older group, and a younger group would have suffered from less illness for a given number infected. But there’s no doubt the illness is highly infectious.

PPS. Yesterday, I went out to eat at a nice restaurant in Tucson. The diners were mostly old, and almost all arrived wearing masks. The waitress wore a mask. They sat at tables in a large outdoor plaza, generally about 12 feet from the nearest occupied table. There was a brisk breeze. Not a high risk setting.

Off topic: James Hamilton says the actual unemployment rate may be 19.7%, although the improvement in May was real. As flaws in the data get corrected, the fall in the measured unemployment rate will likely be slower than the fall in the actual unemployment rate.



30 Responses to ““The” fatality rate for Covid-19”

  1. Gravatar of Josh Josh
    9. June 2020 at 09:33

    To the extent we want to aggregate fatality rates (though I think we shouldn’t do that – we should leave them disaggregated), the number I want is the “age-weighted” fatality rate. Ie assuming people of every age are infected proportional to their representation in the local population, what would the fatality rate be. I think a lot of differences in fatality rates (at least in the developed world) are not due to the proportion of old people in the population but their proportion of the infected. Nursing homes being the biggest variable here.

    I wouldn’t dismiss differential susceptibility because of this single church choir data point. I don’t think the model is something like “80% are entirely immune” (or whatever number). It’s that there might be a distribution of the amount of viral load it takes to infect each person. And that distribution might span orders of magnitude. Such a case would mean that the highly susceptible are over represented among early infections and that R drops naturally as they develop immunity. But it also could mean that in a really high viral load setting like a church choir, 80% can still get infected.

  2. Gravatar of bb bb
    9. June 2020 at 09:55

    I agree that there would have been a lot of voluntary social distancing without the mandated lockdowns. The NBA and NCAA are perfect examples. However, I do think the lockdowns likely eliminated some super spreader events. We’ll never know what would have happened had NY shutdown a week earlier like the DMV or California did. We do know that Sweden has roughly 5 and 10 times the number of deaths per capita than Denmark and Norway respectively. That should not be minimized.
    I was personally grateful when my city shutdown the schools, because there no viable way to voluntarily shutdown the schools.
    I don’t see how you can declare mandatory lockdowns a distraction. I’ve read many stories about superspreader events- you reference one in this article. How many superspreader events were avoided due to mandatory lockdowns? It’s too early to know that answer.

  3. Gravatar of Thomas Hutcheson Thomas Hutcheson
    9. June 2020 at 11:11

    “You can’t do epidemiology without economics.”

    Well you can and many people did, but it was bad epidemiology. The question is, will future models incorporate endogenous behavioral change?

  4. Gravatar of Thomas Hutcheson Thomas Hutcheson
    9. June 2020 at 11:19

    @ bb

    I would not have said “distraction” ether but you do not need “lockdowns” to regulate high-risk activities.

  5. Gravatar of Christian List Christian List
    9. June 2020 at 11:47


    We also have these results in Germany: Singing together seems to be particularly bad, for example in choirs or in church services. Aerosols are apparently more important as initially thought. It’s basically an airborn disease now (not just droplets), and closed rooms seem to be particularly suitable for transmission.

    I think the protests are not a major concern (from a healthcare perspective) as long as they are taking place outside. Looting should be fine as well, as long as masks are being worn. They shouldn’t start singing the American anthem though, but this danger doesn’t seem to exist.

  6. Gravatar of Todd Kreider Todd Kreider
    9. June 2020 at 12:48

    A German study concluded the CFR in Germany is below 0.36% and the lead author concluded that the study was set up to be conservative so that he thought the actual CFR there will be between 0.2% and 0.3%.

    The CDC said on May 20th that the best estimate for the U.S. CFR is 0.4% and the IFR is 0.26%.

  7. Gravatar of bb bb
    9. June 2020 at 13:23

    We may have our terms mixed up. I interpret “lock-down” to mean activities being regulated by government mandate. Most obvious form of lockdown to me is restrictions on large gatherings. Do you have a different definition of lockdown, or do you envision a different way of regulating high risk-risk activities, such as bars and nightclubs or churches?

  8. Gravatar of Justin Justin
    9. June 2020 at 13:51

    -“The CDC said on May 20th that the best estimate for the U.S. CFR is 0.4% and the IFR is 0.26%.”-

    Seems low. 0.29% of NYC’s total population was killed by COVID-19 as of today. It’s hard to imagine IFR < 0.5% at this point for anything other than a very young population. Scott's estimates seem reasonable.

  9. Gravatar of Todd Ramsey Todd Ramsey
    9. June 2020 at 18:54

    It’s unlikely 52 people were infected by one person at the infamous Skagit Valley choir practice. Three people became symptomatic the day after the practice, faster than the CDC says possible.

    In addition, most of the 52 attended choir practice one week earlier. It’s likely that one person infected several people at the first practice, who then spread it further at the second practice.

    The CDC report considers the possibility of spread at the first practice, but dismisses it without discussion. I think that’s because they wanted to scare people into thinking one person could infect 52 people in one sitting.

  10. Gravatar of Todd Kreider Todd Kreider
    9. June 2020 at 19:18

    “Seems low. 0.29% of NYC’s total population was killed by COVID-19 as of today. It’s hard to imagine IFR < 0.5% at this point for anything other than a very young population. Scott's estimates seem reasonable."

    Ok, you go with Scott, the economist, and I'll go with the CDC. Seems fair.

  11. Gravatar of Grant Grant
    9. June 2020 at 20:03

    “PS. In a choir practice in Washington State, 52 of 61 attendees came down with Covid-19, so I don’t believe claims that most people are not susceptible.”

    After looking at household transmission data from contact tracing studies, I too believed a lot of people were not susceptible. However we’ve seen nearly 100% of some prison populations get infected.

    Nowadays it’s clear it’s contagiousness that varies greatly between individuals. Some people just don’t transmit the virus very well if at all. A husband can get sick and not give it to his wife.

    Some people test PCR positive, but attempts to culture their virus fail.

    This realization has important implications for the policy responses needed to curb covid growth. Unfortunately it has been under-reported by the media.

  12. Gravatar of ssumner ssumner
    9. June 2020 at 20:12

    Josh, Good points. But I thought some are claiming that 80% are immune.

    bb, I meant that lockdowns distracted us from policies that would have been much more effective. In my view, Sweden could have avoided lockdowns and had fatality rates as low as Norway.

    Some of the East Asian countries did not even close restaurants.

    Some private schools did close down voluntarily, and I certainly favored shutting down public schools. Heck, I’d close them down even if there were no epidemic.

    Christian, I need a mask to protect myself from your lethal dose of sarcasm.

    Todd, You said:

    “and the IFR is 0.26%.”

    So basically 100% of New Yorkers were infected? Okaaaay . . .

    Justin, Exactly.

    Todd Ramsey, That’s possible, but of course it doesn’t undermine anything I said about susceptibility. Also keep in mind that this event occurred very early in the epidemic, so it probably was one super-spreader.


    You said:

    “Ok, you go with Scott, the economist, and I’ll go with the CDC. Seems fair.”

    The people who said masks don’t work, and then changed their minds and agreed with me? The people who said in February that there was no need to test people? The people who now say that more than 100% of New Yorkers were infected? Yes, you go with the CDC, I’ll go with logic and data.

  13. Gravatar of Todd Kreider Todd Kreider
    10. June 2020 at 03:01

    “In my view, Sweden could have avoided lockdowns and had fatality rates as low as Norway.”

    Scott, the Norwegian statisticians and government completely disagree with you and say they should have gone Sweden’s route and voluntarily social distanced.

  14. Gravatar of Todd Kreider Todd Kreider
    10. June 2020 at 03:05

    “The people who said masks don’t work, and then changed their minds and agreed with me?”

    The CDC was correct in the first place that there is no scientific evidence that cloth masks work. A 2015 study that compared health workers using surgical masks and cloth masks found that cloth masks had a viral penetration rate of 97% whereas it was 44% for surgical masks. The authors recommended that the health care workers not use cloth masks.

  15. Gravatar of Alan Goldhammer Alan Goldhammer
    10. June 2020 at 05:29


    I do a daily newsletter and spend about 3-4 hours a day reading preprints and published articles (archive is here: https://agoldhammer.com/covid_19/ ). Despite all the progress, there is so much we don’t know about the virus and infection. There are a number of papers just in the past three days looking at infections in young people and healthcare workers. Lots of evidence of asymptomatic infections but with serological conversion. There continue to be increasing case reports in states that have reopened and this is going to continue as long as prudent precautions are not taken.

    It’s worthless to continue debating who said what about masks and when. those of us who have a public health or infectious disease background (I fall in the latter category), knew from the beginning that masks were a good thing.

    The whole approach in the US has been so screwed up and my worry now is that they are making another big mistake picking the vaccine companies to move forward with government support. Fortunately, it looks like the drugs that control the immune system response are working (remdesivir is not one of these and I don’t see it having much impact in COVID-19 treatment) and with different clinical approaches, serious COVID-19 infection will be better managed going forward.

    I thought at the beginning that the CRF in the US would be between 0.3-0.6% and I’m sticking by that. Of course it’s dependent on accurate diagnoses of infection rate which is still uncertain.

  16. Gravatar of Benjamin Cole Benjamin Cole
    10. June 2020 at 05:32


    Morgan Stanley, looking at today’s CPI release, say the next PCE core should come in at 1% YOY.

    So the Fed buys a few trillion in assets? The globe has $400 trillion in assets, and huge additional capital flows annually. The Fed’s QE program amounts to what?

    The Fed looks aggressive. So does the Bank of Japan. So does a shadow boxer.

  17. Gravatar of Michael Rulle Michael Rulle
    10. June 2020 at 05:45

    One statistic in the US I find interesting. The 7 day rolling average death rate has declined 54% from May 9th-June 9th. The 7 day rolling average of new cases has declined 25% during that same time period. I do not think one can attribute that to our so called lock down, voluntary social distancing or any behavioral issue. It could be attributable to testing more people and we are capturing more young people in our “new cases” sample.

    This does not preclude deaths rising if we stop “social distancing”——but we will have to start recapturing more older people in our testing sampling as well or The social distancing idea will be weakened. .

    However, we don’t capture this info—-or at least I cannot find it. The fact we don’t have this kind of information is ridiculous.

  18. Gravatar of Alan Goldhammer Alan Goldhammer
    10. June 2020 at 07:48

    @Michael Rulle – statistics in the US have been skewed because of what happened in New York. It is also clear that the lock down in New York worked and the extra beds that were set up were not needed. Had they not locked down, those beds and likely more would have been required. It is pretty well documented that cases in other states are either at a plateau or in some cases increasing.

    The virus is not going away and you can either practice prudent public health measures or just let things run their course. Public surveys seem to support the former approach. I’ve noted in the past that there is 1/3 of the public who are over 60 and who have a lot of disposable income and have made the decision not to fly, go to movies, restaurants, large malls without crowd control, etc. This money will not be coming back into the economy anytime soon.

    Your point on testing is correct and it represents a continuing and massive failure.

  19. Gravatar of MikeDC MikeDC
    10. June 2020 at 08:04

    As usual, I look at this from the opposite perspective.

    If the prudent and obvious social distancing and public health measures were going to happen no matter what (and I agree that they were), then “the debate over the lockdowns” isn’t a distraction at all because they were a tremendously costly imposition on people that probably had little benefit.

    A second point worth considering is that the data required to make any sort of sensible conclusions simply isn’t available. I’m a trained researcher with +20 years of experience looking up numbers, and I couldn’t get a comparable, age-based CFR across countries. I spent a lot of time trying to determine, for example, if the CFR for particular age groups varied significantly across countries, but the data just isn’t there (my hypothesis is that a significant amount of the CFR might be based on differing “treatment” approaches: e.g. Belgium is already kind of trigger happy with euthanasia… an 80 year old there may be treated with a morphine drip, while an 80 year old American may be more likely to get “heroic” life-saving measures).

  20. Gravatar of Logan G Logan G
    10. June 2020 at 08:30

    I’m always glad to hear someone is visiting Tucson (you even managed to hit relatively good weather for us for this time of year!).

    In Arizona, broadly from the outset our governor made a shelter in place order without teeth; namely police couldn’t ask where you were going or why, and there was a vast amount of reasons people were allowed to travel outside. Additionally, I live near Tucson’s major ‘active park’ area, a walkable/bikeable river path loop that goes around and through the city (ignore that the river is dry unless it has just recently rained). This area was absolutely packed for most of March and April, and restaurants/ bars have been getting more full. Lastly, the University of Arizona’s large common area and recreation area, a nearly mile long stretch of grass known as ‘The Mall’, has from the get go been packed. However, there has been a recent flurry of concern here (and I saw at Bloomberg there is even an opinion piece) on a bump in our Covid cases. This is ostensibly due to the end of the Governor’s shelter in place order. However, the timeline just doesn’t line up. I could believe it if I had witnessed most people strictly following the order, but people were actively ignoring not only the shelter in place but even things like wearing masks for the entire time since March.

    I am curious what your thoughts are on this? Certainly more people are out and about now (although not a ton more, it is summer after all, and we have an utterly mind blowing smoke cloud covering the city from the fires in the Catalinas unfortunately), but there is also certainly more testing. Additionally, Arizona’s cases, and in particular the cases with the worst outcomes, have horrifically been predominant in the Native American population, I am unaware if some change occurred on the Native American lands that also may have caused a change.

    Tangentially related, Pima County (of which Tucson is a part) was praised as having one of the best response groups to Covid, in developing best practices, prepping for worst case scenarios, and providing guidance and obtaining testing resources. Perhaps Tucson isn’t particularly contributing to the jump in numbers, instead perhaps it is Phoenix (Maricopa County).

  21. Gravatar of Mike Mike
    10. June 2020 at 10:39

    A point seemingly missed by you Scott is recent data showing asymptomatic folks, who were infected, are not showing anti-bodies in 60% of the cases. The infection rate is probably a lot higher than you believe.

  22. Gravatar of LC LC
    10. June 2020 at 11:07


    Off topic but I cam across Chris Schwarz’s presentation on the economy (together with Bill McBride) here: http://boundedfinance.com/

    3 things struck me:

    1. For all the talk of Fed’s liquidity gush this time, the Fed’s balance sheet change is actually < 1/2 of Fed's Balance Sheet change in 2007-2009 (see slide 73). This seems a powerful argument that expectation matters and Fed did a much better job this time than 2007-2009, so much so that they didn't have to use so much "ammunition" in the conventional view.
    2. Fed should adapt NGDP LT going forward to get themselves out of a rock and hard place situation (see slide 77).
    3. Dollar index is flat this year, this shows Fed has supplied the dollar demand from the world this time thus the "safe haven" aspect of dollar is not true.

  23. Gravatar of Justin Justin
    10. June 2020 at 11:07

    –“Ok, you go with Scott, the economist, and I’ll go with the CDC. Seems fair.”–

    Did you see my comment about NYC having already lost 0.29% of its population to COVID? An IFR of 0.26% is too low even assuming 100% of NYC was infected, which is almost certainly not the case. The Bronx death rate is even higher, with 0.32% of the 1.4 million population killed by COVID.

    This isn’t about the authority of the CDC or Scott. If 0.32% of the Bronx is already dead from COVID, the IFR for aging western societies is well above 0.26%.

  24. Gravatar of Todd Kreider Todd Kreider
    10. June 2020 at 11:33

    The IFR of NYC can be higher than the IFR for the U.S. due to demographics differences, which is what the CDC has estimated.

  25. Gravatar of ssumner ssumner
    10. June 2020 at 15:35

    Todd, You said:

    “Scott, the Norwegian statisticians and government completely disagree with you”

    AFAIK they’ve never even been asked about my proposal. I don’t think you understand what I’m proposing. I’m not just saying “don’t do lockdowns”, I’m saying do other things that are more effective. Sure, if all you do is refrain from lockdowns then you’ve got major problems.

    Logan, The police? Here in California we were always perfectly free to go outside. I assumed that was true everywhere in America. Were the police harassing people in some states?

    I’d expect that ending the lockdown would make cases increase, but not by a lot.

    Mike, Link? What does that say about the proportion of NYC residents who were infected? It seems implausible.

    LC, But the Fed hasn’t done all that much better than in 2008, just a bit better. They still have a very weak forecast for 2021 and 2022.

    Todd, Demographic differences? Seriously? That could only explain a tiny bit of the discrepancy. NYC is not all that different demographically. The 0.26% figure makes no sense for a country like the US.

  26. Gravatar of Todd Kreider Todd Kreider
    10. June 2020 at 16:45

    “Todd, Demographic differences? Seriously? That could only explain a tiny bit of the discrepancy. NYC is not all that different demographically. The 0.26% figure makes no sense for a country like the US.”

    Not according to a modeler whose name I can’t remember but has concluded that almost all of the difference in deaths is due to initial conditions of a country and that of course includes demographics. His conclusion was that government policy differences explain almost none of the differences in outcome. (Someone at Marginal Revolution stated the modeler’s name but couldn’t find it later.)

    As I’ve written, 40% of Americans are obese and 10% are extremely obese with a BMI of over 40. Contrast that to Japan and South Korea where 3% to 4% are obese and a tiny percent are extremely obese.

    Also, more than twice as many Americans per capita have heart failure
    than Japanese.

  27. Gravatar of Christian List Christian List
    10. June 2020 at 18:44


    with all due respect, you are such a stubborn idiot. Even the CDC people changed their minds, they’re recommending face masks now, especially cloth. You are more stubborn than the stupidest bureaucrats in the US, and that really means something. Is this a fatal side effect, i.e. one and only effect, of the miracle food supplements you are taking, or where does this idiocy come from?

  28. Gravatar of Todd Kreider Todd Kreider
    11. June 2020 at 04:20

    Mr. List,

    I know you aren’t a curious person, so you must not have noticed that the CDC did not provide one shred of scientific evidence that cloth masks protect a person or those they are around.

    A 2015 study of health workers compared using surgical masks with cloth masks:

    “Results: The rates of all infection outcomes were highest in the cloth mask arm, with the rate of ILI statistically significantly higher in the cloth mask arm (relative risk (RR)=13.00, 95% CI 1.69 to 100.07) compared with the medical mask arm. Cloth masks also had significantly higher rates of ILI compared with the control arm.

    “An analysis by mask use showed ILI (RR=6.64, 95% CI 1.45 to 28.65) and laboratory-confirmed virus (RR=1.72, 95% CI 1.01 to 2.94) were significantly higher in the cloth masks group compared with the medical masks group.

    “Penetration of cloth masks by particles was almost 97% and medical masks 44%.”


  29. Gravatar of Christian List Christian List
    11. June 2020 at 11:41


    studies are slowly being done, but it is of course extremely difficult to get really good studies so quickly for a virus that has only been around for a few months. You’re making a rookie mistake again and again, which is that absence of evidence is not evidence of absence.

    When we don’t have many good studies, which is true for most problems that we face in this world btw, then we just have to use our common sense.

    And where is your specific problem with masks anyway? Masks are extremely low cost, you can make some of them even yourself, and in the worst case scenario they don’t help as well as we thought, so what is your problem?

    The study you cite is a different virus, and more importantly, it does not investigate the mechanism by which masks are mainly supposed to work right now. If you still do not understand such basics, then you really are a hopeless case indeed.

    It’s so funny that you swallow your miracle food supplements en masse, even though in those case there really is zero proof, and then for measures that are very likely to work, you play the fool again. This might not be a coincidence. There may well be a connection, for example in the sense that exactly the same perception and processing errors could play a role in your brain.

  30. Gravatar of Todd Kreider Todd Kreider
    11. June 2020 at 12:42

    Quite wordy of you today, Mr. List. However, you don’t cite one scientific study to show effectiveness of mask wearing for either the mask wearer or those nearby while I cited a 2015 article showing that in a medical setting, cloth masks were worthless. Yet somehow you are the wise one and I am the rookie.

    You aren’t science minded, so it is pointless to discuss the two supplements that are believed to be either helpful or potentially helpful. Ask your neighbor food delivery guy how vitamin D3 in larger doses than the 400 UI is helpful.

    There are 10 Nobel Laureates of Physiology or Medicine backing the research of NR (nicotimide riboside), a vitamin B3 derivative, and last year the FDA allowed Elysium to apply for drug status after reviewing the favorable results of acute kidney injury after heart surgery.

    Another small study showed NR+pterostilbine slightly reversed the condition of ALS patients after 4 months and one year, the length of the study and for $1,000 a year outperformed the latest ALS drug that was approved in 2017 and only slowed decline by 30% over the same time period.

    An earlier small study showed 1,000 mg of NR reduced systolic blood pressure in those with pre-hypertension by 10 points, which translates to a 25% reduction in risk for a heart attack.

    A small study revealed strong evidence that 1,000 mg of NR over eight weeks reduced fatty liver by 20%. As a result, tqo larger studies have started.

    In a few weeks, the U of Washington trial of 30 heart failure patients taking NR at 2,000 mg or a placebo for a few weeks will be published and there is a good reason to think those results will be positive.

    I thought you are a doctor. Why do I have to explain this to you?

Leave a Reply