COVID-19 Update: 2020-May-02



Thanks to all for kind donations to support the blog. If you want to help, please visit the donation link below.

This blog is made possible by readers like you; join others by donating at My Wee Dragon.


41 responses to “COVID-19 Update: 2020-May-02

  1. A little related assessment:

    This may be of interest to our community.

    I will be updating. I need to figure out how to drape uncertainties on those plots.

    Sorry for the self-promition, but I’m excited to bring FDA into the discussion.

    [Response: This particular self-promotion is “on topic.”]

  2. Interesting how fast Nebraska is shooting up, especially since they have an election shortly.

  3. It seems way to early to ease the lockdown.

  4. What your numbers bring home to us folk the other side of the pond is that while the measure of per-capita-death-toll may place the US below many European countries, (as per this table of reported Covid-19 deaths/million pop.)
    Belgium . 684
    Spain …. 534
    Italy …… 470
    UK … …. 419
    France .. 378
    Holland . 295
    Sweden .. 265
    Eire . … .. 264
    USA … …. 207

    But while the US sits below these European countries, only ten of the states sit ‘above the US average’. Thus we find 73% of the US death toll is carried by states with 22% of the US population and in the extreme, NY state has suffered 36% of the deaths with only 6% of the population.

    New York … . 1,267
    New Jersey . … 888
    Connecticut . .. 700
    Massachusetts . 576
    Louisiana .. ….. 433
    Michigan … ….. 405
    DC … . … … .. ..356
    Rhode Island … 302
    Pennsylvania … 221
    Maryland … ….. 212

    • Although there is the persistent suspicion that the US is missing a lot of Covid-related deaths. That’s still a bit of a ‘known unknown’ AFAICT. I guess a lot of that sort of stuff will be worked out retrospectively WRT excess mortality studies.

      • @Doc, and all,

        Also, because of counting problems, and irregular periods for release, epidemiologists are using rates to guide policy, smoothing counts first before calculating them. Prof David Spiegelhalter explains.

      • Doc Snow,
        The missing Covid-19 mortality as well as non-reporting of infection cases is something that is a bit of an issue here in UK. As well as silly numbers for Case Mortality Ratio pointing to the vast majority of infection cases going uncounted (particularly in England where the CMR is above 20% and double that of the rest of the UK), there is a pile of anecdotal & other evidence to suggest that a very high death toll outside hospitals is passing under the radar, not least of this is coverage of the Excess Mortality numbers which look startling enough to be fake, except this factcheck page (which shows a New York graph for comparison) suggests otherwise.

  5. Greg Wellman

    Using the site, there appears to be a strong weekly cycle in reported deaths. Presumably this is a reporting artifact, but does anyone know why?

    • @Greg Wellman,

      I don’t know specifically, but Prof David Spiegelhalter (link to podcast just above) says Mondays were hot, so the presumption is that counts of deaths were not reported at the weekend, but only on Mondays. Whether or not that explains the entirety of what you see at worldometers, I do not know. Can you document this strong weekly cycle? You’d need to use something robust, like a multitaper method. There are codes out there to support this, e.g., the multitaper package of R, or spectrum.mtm.dpss in Python.

      • Greg Wellman

        Yes, I’ve since been told that these “fast” statistics don’t have to line up with the exact dates of death that will eventually make it into the “slow” statistics. So it’s easy for a hospital or a county or other reporting organization to defer some of all of the weekend numbers to Monday or Tuesday, resulting in a big dip in the weekend numbers.

  6. Al Rodger

    Thanks for telling a bit about ‘the other side of the pond’.

    I agree with you that the death toll per capita distorts the real situation.

    This imho is due to the fact that the per capita factor does not properly reflect the reality: an epidemic or pandemic is a matter of transmission, which does not depend only on medical and social factors in a country, but is influenced by the population’s density in that country as well.

    It is thus not very meaningful to compare countries like the US with 33 inhabs/km² with countries like Belgium with a more than ten times higher density. Comparing it with e.g. Australia would be meaningless as well.

    And therefore it does not make much sense to compare the countries by dividing cases or death tolls by the countries’ population size.

    Thus to show how European countries heavily affected by SARS-COV-2 do behave in comparison with the US, I used the daily case increments, uniformly scaled to percentiles, and underlined their centred, 7 day running means:

    Apart from Germany, the US and Russia, the countries Belgium, France, Spain, Italy, Sweden and the UK were chosen because on May 1, they shared (in different sequence) the top five of both the death toll per capita and the death toll per total cases ratio.

    It is interesting to see how different SE / UK / US behave compared with BE / DE / ES / FR / IT.

    RU seems to be in front of worst things…

    J.-P. D.

    • @Bindidon,

      What I find keenly interesting about transmission dynamics is how the virus got introduced in the first place. We now have a confirmed case in France in late December 2019, and the Swedish public health and medical people think they have circumstantial evidence SARS-CoV-2 was circulating in Sweden in November 2019. These cast doubt on our narrative and understanding of the virus’ early days in contact with people, even upon its origins. There is, for instance, an explanation which says that it was circulating in the human population and something flipped to release its virulence.

      Citation: A Deslandes , V Berti , Y Tandjaoui-Lambotte MD , Chakib Alloui MD, E Carbonnelle MD, PhD , JR Zahar MD, PhD , S Brichler MD, PhD , Yves Cohen MD, PhD, “SARS-COV-2 was already spreading in France in late December 2019”, International Journal of Antimicrobial Agents (2020), doi:

      Citation: Gambaro, et al, “Introductions and early spread of SARS-CoV-2 in France”,

    • Bindidon,
      I can shed some light onto the UK’s ‘daily case increments’ and why they are not falling despite entering a seventh week of lockdown.
      Identifying a new case requires a test and simply the UK, despite Boris’s yesmen telling us how we are a/the “world leader” on testing and how well we have done given the circumstances:-

      “We invented the test in January. In February we got the number of tests up to 2,000 [a day]. In March we multiplied that by five times to 10,000. Then we set the 100,000 target. … So we have basically caught up with Germany that started with this massive capability. We are miles ahead of South Korea now. Absolutely.”

      , dispite all this rhetoric, the UK has been running behind the curve since the start. Through the first three weeks of March (your dy 32-53) they had barely enough test-capacity to check the symptomatic patients entering hospital were positive. So the cases through those weeks are simply those admitted to hospital.
      Today they are testing six-times as many people as they did in early-March (well short of the 100,000 tests they say they carry out) and have spare capacity beyond the falling requirements of hospital admission screening. While this has reduced the %-positive tests, there is enough infection floating around for the number of cases identified to maintain the peak rates in recent weeks.


      The one big important unknown is the actual daily infection rate. If we are going to test our way out of this mess, we do need to know.
      But such numbers are absent from any official statement. They just are ignoring the problem.
      (As an example, the Chief Scientific Adviser Patrick Vallance yesterday reportedly said at the Health and Social Care Committee (which is examining Management of the Coronavirus Outbreak) that he would “not expect to see antibody levels much above mid-teens [percentage]” in any [UK] region, this based on 5-week-old testing which suggested “something like a 10% antibody positivity in London” and perhaps half that elsewhere. (The discussion was all about levels of immunity and presumably [I await the upload of the transcript] still stuck in the much-discredited viewpoint of achieving herd immunity.) Given the recorded level of Covid-19 infection agross England is now averaging 0.2% and the daily infection rates are running above 4,000/day, if there is now actually 15% infection, then presumably the infection rate would be (4,000 x 15/0.2 =) 300,000/day. If it was based on data from 5-weeks ago in London the multiplier would be a bit bigger – 10/0.1.)

      • “We are miles ahead of South Korea now. Absolutely.”

        Yes, if by that is meant “miles ahead in cumulative caseload, active caseload, and death toll!”

        Jeeze, Louis! Dumbest thing ever! (Not uttered by Donald Trump, that is–and even so, it’s dumber than some things Trump has said.)

        Testing *when it matters the most* is crucial, and that, South Korea did in spades. That’s why, this morning, they are 38th on the global list for cumulative caseload. Per Worldometers, South Korea’s 7-day moving average for new cases, as of yesterday, was all of 8. The UK may well avoid having that number reach 8 thousand; only once (April 10) did it hit a single day increase that large. But that’s really nothing to brag about.

        It is, of course, true that the UK is “miles ahead in tests,” too–even if it’s misleading. The only reason the UK (and many other jurisdictions) are now ahead of South Korea in testing is that South Korea hasn’t needed to do mass testing for weeks now, because they effectively contained the epidemic.

      • Al, sorry if that rant seemed as if it were directed at you. It wasn’t. I know you know all that, too. But ai! The stupid, it burns!

      • Doc Snow,
        Indeed the stupid burns!!

        I do find Boris & the Yesmen unredeemably dreadful in that they happily cover up incompetence with a goodly covering of well-polished turd. At least with Donald T Rump you can explain his incompetence by his obvious personal failings. Boris is likewise a dangerous bullying fool but the vast majority of the UK has yet to notice.

        As you say, the reason UK needs so many more tests is because they allowed Covid-19 to spiral out of control, and after all this time they still haven’t learnt the lesson.

        Test-wise, through the pandemic South Korea have on-average tested 59-times for each positive case, their daily test/case never dropping below 12-times. They ramped-up testing from the hundreds to over 10,000 in just a week.
        The UK testing has averaged only 5-times for each positive reported case (and ther’s probably 5-times more going unreported) dropping down below 2-times for each positive at the peak of the crisis> After their grand achievement, they now only manage the same as the S Korean minimum of 12-times per positive, and that with most positives going unreported. All this because it took them not a week to ramp-up from hundreds to above 10,000 but 5 weeks, a rate of increase that did not keep pace with the spread of the disease.

        And, as quoted above, the morons make all this to be some great achievement. Sadly, at times they appear to believe their own propaganda and add to their pile of cock-ups accordingly. Still they are quite expert at polishing turd.

    • Yes. It’s also an issue within nations, and certainly here in the US. Lifeways are very different in New York or Boston than here in rural South Carolina, and not least in our dependence upon automobile travel, which while carbon-intensive relative to public transit, does inherently involve social distancing. (Until, that is, you need to buy gasoline…)

      Our household was already ‘socially distanced’ to a considerable degree, which has the advantage for now that we feel much less disruption than many people do, and consequently less emotional distress. (Much of the latter actually arises from witnessing the monumental ineptitude of the federal response, apparently culminating in our so-called president simply giving up on everything except top-line economic measures. But I digress.)

      It’s hard to know how the various factors play out quantitatively. To what extent does the dispersed nature of life here actually inhibit spread, and to what extent is that effect negated by the existence of ‘pinch points’ involving provisioning of one sort or another? It’s certainly true that early spread, before social distancing, was quite rapid in our rural county: we were for the first few weeks the leading county for Covid-19 cases in absolute terms.

      We’re still #3 on a per-capita basis, and I note that by far the worst-hit county–Clarendon, with ~7 cases per thousand–has about half the population we do. On the other hand, many of the least-affected counties are also rural–the truly urban ones occupy the mid-to-upper-midrange, more or less. Which brings us back to the question of how representative those official numbers are; the smallest, poorest counties presumably also have the weakest health infrastructure and the least capacity to detect cases. So maybe some of them are hit worse than would appear.

      • @Doc Snow,

        Yes, diffusion of (this) disease in rural areas is poorly understood, due to lack of information. Part of the problem may be poor public health staffing, training, and infrastructure. There may be misdiagnoses, which is natural, because, in the absence of virus testing, antibody testing, or devastating symptoms, it’s easy to confuse COVID-19 with pneumonia or some kind of influenza. Before testing, it’s strongly suspected there were cases, even deaths, which were miscounted.

        This combined with symptom-free transmission or nearly symptom-free transmission makes it difficult to know in rural areas where they stand. It will probably be possible to discern this well after the fact by looking at excess deaths versus ages — standard actuarial stuff — but that’s not helpful in managing.

        It’s easy to think COVID-19 is not too bad if cases are not definitively counted. But from what’s known about the earliest cases of transmission, the virus really gets around. The latest surprise is that it was definitely in France in late December. Swedish officials suspect, but don’t have concrete evidence, that it was in their population in November. The early history of COVID-19 is still being written. Whether or not the origin was in China, or elsewhere and imported to China, or even if there was something crazy like a dual emergence, these are all being newly discussed.

      • @Doc Snow,

        The other index of how bad COVID-19 is in an area is how slammed the funeral homes are. While Massachusetts is much denser than what you describe, the funeral directors are saying “We’ve never seen anything like it.”

  7. Steve Lusky

    “dependence upon automobile travel, which while carbon-intensive relative to public transit, does inherently involve social distancing. (Until, that is, you need to buy gasoline…)”
    Ah, another advantage to charging at home instead of fueling with gas or H.

    The USA charts with triangle in circles is very informative. Would be nice to see a version that updates daily, and has an option to show the trend for the past two weeks. But thanks for these charts as they are.

  8. Hi Tamino,

    Good blog. I have read this code review of the Imperial College model of coronavirus spread:

    I am familiar with climate science and have some understanding of how these models work, but I am not familiar with epidemiology. I was just wondering about your opinion on this, given this.

    I am aware of a growing number of voices of dissent about the lockdown and I am skeptical of this piece given its agenda laden conclusion. However, I would like to know your thoughts on the validity of the model and the criticisms of it, so that I can make an honest assessment for myself over this. Is it a concern or is it not?

    • Michael Sweet

      I noticed that the model the White House cited estimated 70,000 deaths by August a few days ago when recorded deaths were over 60,000 and deaths were over 1,000 per day. They obviously chose that model because it low balled the death rate.

      When they started to open up the economy that model jumped to around 130,000 deaths by August.

      I am sympathetic to the modelers. How do you model social distancing? If we distanced as well as the Europeans we would have many less deaths than if we all go out with our guns and no masks. Trump and Pence never wear masks as a bad example. The recent surge of cases in the White House, where no-one has worn masks in the past, might get them to at least insist that everyone else wear masks.

      It is similar to climate models. How much carbon will be emitted? How do you determine the damage done when you do not know how much carbon will be released?

      • I am less sympathetic to the modelers in question.
        (well I am sympathetic to the idea its freaking hard to do and get right, especially as the major thing that would need predicting is what changes in policy will happen (which will be based on what the model tells them…) AKA if some model says infections will go down, (if the gov listens to that “model”) policy will change to loosen restrictions and then they wont…
        So basically every model, if it is treated as authoritative by policy-makers, is retrospectively made wrong.

        This link compares12 sets of models.

        one I gotta say… if I worked where they made it, my name wouldn’t be on it. It fails to pass the sniff test and shows clear wrong historical “belief” (encoded as an assumption) that any day a now gov policy will smash R well below 1.

        Well, when I tried to drill-down and work out the evidentiary basis for that I found crickets.

        Even the LANL model when I read the model description here
        the ‘more realistic’ LANL model really just assumes some randomness in how fast R decreases but that it does decrease at the rate they suggest it just largely coded in.

        How little real information other than that things are exponential up or down depending on R. How little the models model can be seen by looking at the comparison site linked first (above in this post) with semi-log axis turned on.

    • John,

      The comment you link to gives pretty-much my take on the OP it is describing. Hey, if one model run suggests 400,000 dead bodies and another go makes it 480,000, is there reason to get worked up about the model? Or should the potential half-million dead bodies be the concern?

      People who object to lockdown are not arguing that, say, schools should be re-opened immediately or that cafés and restaurants should be allowed to open when they’ve apply social-distancing measures. Yet those are the sorts of things the big models are considering. They are not calculating how big the pile of dead bodies would be if lockdown ends on Monday-week.

      As A UK resident (where anti-lockdown protests have been no more than silly news), there is a lot of Covid-19 stuff that gets me wanting to throw things at the telly. But I don’t see models being at fault.
      In my view, the UK problem is the poorly defined agenda set by politicians and their scientific advisers that have allowed the UK no other path than lockdown yet without having a proper appreciation of how you get out of lockdown. The UK is likely not alone in this predicament.
      But we Brits are not concerned by this. Boris will save us. He will get a vaccine*/anti-body-test*/mobile-app*/cure-all* [*delete as appropriate] to save the day because Boris’s Brexited Britain is a world leader everything including in Covid-19.

      • @Al Rodger,

        Concerns need to be properly separated.

        The fundamental reason a really crude device like mass reverse quarantines needed to be applied, with obvious economic consequences, is that the governments who needed to resort to it were completely unprepared for the pandemic. These have been studied, written about, and urged to be defended against for the last 30 years, especially the last 20 years, since SARS-CoV-1.

        As near as I can tell, and I’ve noted this elsewhere, the only reason that was not done is because the governments in question had other priorities (military buildups, for instance), were too cheap, or their public was too cheap. I’d say the economic hardship being wrought upon the countries with such governments and their publics is exactly what you get when ill-advised long term choices are made. I have some sympathy, but not that much. The blame lands on the people who could afford to pay more substantial taxes, but preferred not to do so.

        You cannot prepare for these things as they develop. They move too fast and little is known about them.

        The same is going to happen with climate disruption.

        I include my state, Massachusetts, in the USA, within that condemnation. Clearly some have been better prepared than others.

      • Hi Al,

        Those are my thoughts too, but I wanted to defer to others who may have more knowledge than me.

        I didn’t mean to link to that comment, but the blog. To clear up any potential confusion, my comments have been directed at the blog, not the guy in the linked comment.

    • Tamino will have his own opinion. I wait to hear what he says, if anything.

      I looked at the top of the cited lockdownsceptics post and the highlighted comment. And I was a software engineer (then a test engineer, then a statistician, now a statistician/data scientist/quant), and am familiar with a lot of code, including codes that are used in academics, and geophysics (hydrology, mostly), and the R code base.

      The criticisms of the post and comment are about form. Sure, in delivered products, having properly structured and documented code is a good idea, primarily from the company owning the code base, or in good open source code. But it isn’t sufficient to be sure that the code works, particularly if it is quantitative code, that demanding good numerics and algorithms, good testing, and proper operation.

      Facts are, in industry, with the advent of Internet-quality code, and particularly the Agile and scrum methodologies, the quality of code I have seen is atrocious. A lot of open source code is that way. Obviously, it works, is perceived to be useful, is useful, but the lack of good documentation costs the owners/users/adopters a lot. With quantitative code, particularly things like the R code base, the code depends a lot on people understanding the mathematics of the calculations sufficiently to understand what the code is doing. For example, many R packages don’t test or document all the corner cases, and I have had to open up the code itself to understand what they were doing, and take the code and fix the corner case myself.

      So, simply having a code base at Imperial which is not in the best shape is pretty much typical for industry these days, so I don’t see that as a vigorous criticism. Industry also tends not to release their data and code for public view, which is decidedly unscientific.

      About the only thing that I find troubling is that this base, if the reviewer at lockdownsceptics got it right (I don’t know; I haven’t looked at Imperial’s code; I won’t), is that much of the code is written in C++, rather than in something like R or Python 3, and I would recommend the former. (Reason? Like it or not, the fundamentals of Python’s numerics in numpy aren’t that great. I have specific examples. Email me.) This is for transparency and machine independence.

      Apart from that, I cannot see how a proper criticism can be mounted without knowing how to operate the code, and that requires either more documentation than the critics have accessed, or much more familiarity with the code base than the critics wish to invest.

      Also, relating to some of the other comments about other models and projections, the core model used by the Gang of the Orange Mango to come up with their 100,000-200,000 estimates was never identified, let made available for peer review. By juxtaposition, the presentation of that in the Gang’s briefing was implied to be the University of Washington, which it decidedly wasn’t, and if you want an example of how good code is managed, check out University of Washington’s Github or Johns Hopkins.

      And if you want a good insight as to what went into modeling at University of Washington, and many epidemiological models, listen to the recent episode on Five Thirty Eight where University of Washington’s chiel modeller is interviewed: Dr Chris Murray.

      So, I think there’s nothing there. I also observe, after 45 years in the business of doing principally quantitative software, that many software engineers don’t know much about it. Many, for example, think that the way to solve a system of linear equations is to invert the coefficient matrix and do the matrix-vector multiplication. And I found an optimization bug in the GNU C compiler 7 years ago which produced an incorrect numerical result. I even isolated it. The group maintaining the compiler was uninterested, at least at the time.

  9. I am too. I realise that there are a lot of unknowns in how we react, and good reactions can severely curb the death toll.

    However this piece indicates that there is something wrong in the analysis, with inferences that outputted values seem to be wholly unreliable even if no variable is changed.

    Now, I am very skeptical of that blog. It appears to have an axe to grind, as evident by the last paragraph, common among many of science’s attackers. Most of its issue appears to be with the quality of the code itself. I am familiar with code from academics and I am aware of how it can be added to, deleted and altered by research students throughout the years, which makes it seem less streamlined than a code written asap by a Google programmer. So, this bothers me less than some on there.

    However, my knowledge of modelling is quite fledgling and wondered what the thoughts were on this piece. Whether Ferguson’s code is as bad as made out, despite the red flags from this blog’s credibility in my eyes, and why the blog may be wrong/right.

    This is prescient for me, because I have been in long debates with a climate change denier elsewhere who continues to argue that we can’t trust climate models, because they’re unreliable. They would argue that the variability in climate models make them untrustworthy, and argue consistently that climate models are inaccurate, pointing to every source possible which says that models need to be improved as some kind of evidence for this.

    Americans have my deepest sympathy in this situation, your position isn’t good, thanks to the clown in charge. However, I am British and while things aren’t brilliant, they’re better than the States at present. I think the blog is British, as are many of the commenters. Neil Ferguson claimed that without action the UK would have 250000 deaths, but that will not happen. Most back the lockdown, but the skeptics are growing restless and are seizing on blogs like this to show how we’ve done it wrong, how the economic recession will endanger more lives, and that we should open up. I worry that these noisy people will cause havoc in public discourse, as they have done over Brexit, and will do more to damage the image of expertise in favour of their man down the pub bluster.

    • @John,

      If the code were, hypothetically, a modern Bayesian analysis, because these often use stochastic integration to identify posteriors, yes, every time it was run you would get a slightly different result. If it was set up incorrectly, you might get results which differed substantially, but that would be because the setup was incorrect, not because the model was incorrect.

      Look up MCMC and work with the semi-declarative systems BUGS and JAGS. Or the R package MCMCpack. There are things like this for Python, too, e.g., PyMCMC.

      Don’t criticize, is the message to them, unless you understand it.

    • @John,

      Most back the lockdown, but the skeptics are growing restless and are seizing on blogs like this to show how we’ve done it wrong, how the economic recession will endanger more lives, and that we should open up.

      Also, SARS-CoV-2 isn’t finished thrashing the UK yet. There are still deaths. And, after a limited reopen, there could be many more.

      The lesson from Korea is sobering.

      • The UK’s trajectory reminds me of the US’s, actually. They are both in a pattern of quasi-linear increase, and both show very slow declines in daily cases. Both appear to me to be a long way from peak active caseloads, even if the re-openings don’t re-aggravate the epidemiological injury. Good thing for Boris that Russia is around to make the UK look less bad! Russia figures to become #2 in the world by surpassing Spain’s cumulative caseload in a week or so. I expect the UK will eventually surpass Spain as well, though it will take a while.

        On this side of the pond, Canada may be getting closer to a peak–they actually ran a slight decrease in active cases the other day, and while they have yet to repeat the feat, that metric clearly shows some serious flattening. We’ll see.

        But there are a lot of countries moving all too rapidly up the lists. The one that worries me the most is India; they are poised to surpass the Canadian cumulative caseload, and of course they have an enormous downside in terms of vulnerability. Pakistan isn’t looking too good, either, and I never like to see nuclear rivals in bad national moods. (A comment with wider application than just the Indian subcontinent.)

      • @Doc Snow,

        As Dr Erin Bromage recently observed,

        There are very few states that have demonstrated a sustained decline in numbers of new infections. Indeed, the majority are still increasing and reopening. As a simple example of the USA trend, when you take out the data from New York and just look at the rest of the USA, daily case numbers are increasing. Bottom line: the only reason the total USA new case numbers look flat right now is because the New York City epidemic was so large and now it is being contained.

      • Yup. Here in South Carolina, if you look at the state new cases bar graph and squinch your eyes, you can convince yourself that there *could* be a small declining trend over the past few weeks.

        But I seriously doubt it would pass statistical muster. And here we are, opening up restaurants for inside dining–albeit with social distancing that will probably make it pretty uneconomical as a business proposition. The coverage is all of the ‘reopening.’ It remains to be seen how much buy-in it gets from the populace; we sure don’t plan on doing anything different than we have been over the last few weeks. But we’re lucky in that that is an option for us.

      • Doc Snow,
        The UK case/day numbers do not provide a very useful measure of the level of infection as the testing methodology has changed with time.
        From the point when theu began reporting the numbers for the different ‘Pillars’ (which was a week after the peak in recorded cases), what they call “Pillar 1” testing (hospital patients and now also hospital staff) shows a reduction in cases of 60% while “Pillar 2” (essential worker tests) has increased by 240%. The result is a roughly constant level of recorded infection. The unrecorded infection is almost certaily still far greater than the recorded infection.
        The UK death rate (which is still under-reporting) is probably the best measure of UK infection as the methodology has not changed, although the numbers will be running some two-week behind the infections.

      • Doc Snow,
        Further to the UK COVID peak, they have just published the latest “excess deaths” report for w/e 1st May although this doesn’t cover Scotland & N Ireland. This data is now showing a definite reduction in mortality (and continues to mirror the ‘reported UK mortality’ with the excess deaths running some 70% higher than COVID deaths) with the peak mortality 4 weeks behind the lockdown (23rd March) which was presumably a week ahead of peak infection.

      • I agree that US statistics have been so utterly dominated by New York state, and New York state’s by New York City, that it you almost have to subtract New York out to make any sense. That said, in the US, there are several stories. LA and MI have both done a good job flattening the trend and turning over the daily occurrences–and in MI, it is despite the heavily-armed knuckleheads protesting the closures. CA unfortunately seems to be losing the thread. MD and VA seem unable to get it together despite sane policies. Rural America has yet to be hard hit, but they seem to be incapable of learning from other’s mistakes.

        Globally, it looks like the US and Russia–and to a lesser extent the Arab states will be the basketcases. We are nowhere near out of the woods.

      • snark, don’t forget India. They are now (IIRC) #12 in caseload, and have a population not materially less than China’s–not to mention governance that’s not uniformly great, and gobs of poverty. Their situation concerns me deeply.

        Al, thanks for the elaboration on the UK numbers, and why they are even less helpful than might appear. With the continuing spread of the pandemic and its deepening into the developing world, I think the statistics are only going to keep getting less and less precise and informative. I mentioned India in this connection, but there are numerous other countries in point, from Peru (currently hot on India’s heels, as far as we can tell) to Pakistan (whose nuclear weapons are presumably pointed at India). Not a good situation at all.

        On the other hand, at least Canada’s nascent claim to have peaked seems gradually to be solidifying, with a third daily decrease in active cases, and a 44% ratio of active to cumulative. Those numbers at least should, I think, be pretty solid.

      • Yeah, India is worrisome. They have been restricting internal travel–and that is a good thing, because, while dilapidated, India’s transit system moves a lot of people a lot of miles.
        An interesting observation I’ve made is that pretty much everywhere you look, the virus follows economic activity–even more so that population. It may be the first pandemic of a truly global economy. In India, that might actually be a good thing. India’s economy is quite poorly integrated. Villages have few connections to the cities and to the global economy (other than cell phones). If people can’t make it back to their villages, the disease may stay confined to large cities.

        It might be interesting to look at cases per $million in GOP on a regional basis rather than cases per million inhabitants.