Category: Data Source

Here is a Christopher Balding tweet storm, excerpt:

Iceland has done almost 14k tests on an island of 360k so more than 3% of the total population…They have more than 800 confirmed cases, 10k people in quarantine, 800 in isolation, 18 hospitalizations, 6 in ICU, and 2 dead…About how many people SHOULD have corona if the spread etc numbers are accurate. As of March 27, Iceland would be expected to have more than 46k people that have corona. Emphasis this is on an island of 360k and 800 confirmed cases.

What is going on in the Icelandic numbers?  What accounts for this apparent heterogeneity?  Dosage?  Is it that Icelandic clustering is mostly in one easy to control central city and the rest already is “socially distanced,” even in the best of times?

I know there are some MR readers in Iceland, and presumably they read the Icelandic press.  Can anyone shed light on why the death rate is not higher in Iceland?  Is it that the death rate is about to burst a week from now?  Alternatively, you might think the Icelanders have kept their hospitals up and running — important for sure — but that doesn’t explain what seems to be a quite low rate of reported cases.  Or is it that Iceland’s second largest city is so tiny — Akureyri at 18,925 inhabitants — that the virus doesn’t have many easy chances to recirculate once cut off for a while?

Similarly, Sweden hasn’t restricted public life very much and they do not seem to be falling apart?

How much better is Staten Island (less dense) doing than Manhattan (more dense)?

Some reports indicate that in hard-hit Westchester County,. NY, the rate of hospitalization is about one percent (8-10 percent in some other places).  Alternatively, here is serious talk that the death toll in Wuhan is 20x official figures.

How much of the heterogeneity results from the kind of mixing you get?  One account of the low German death rate is the young and the old were never pushed together so much by the policy response.  One account of the high Italian death and hospitalization rate is that the initial quarantine was only regional and thus it spread very dangerous forms of mixing throughout the larger country.

It is possible that Cambodia, Thailand, and Vietnam still will be hit hard, but so far the signs do not indicate as such.  Warm weather may play a positive role, though that remains speculative.  The latest weather paper appears credible and indicates some modestly positive results.  Of course weather won’t explain the relative Icelandic and Swedish success, if indeed those are truly successes.

By the way, on the “everyone already has it” theory, a semi-random sample of 645 from Colorado showed zero positives.

So where is all this heterogeneity coming from?  Is it all just bad data?  That seems hard to believe at this point, and Iceland seems like a plausible source of reasonably good data.

As for concrete conclusions, these heterogeneities should make us more skeptical about any models of the situation.  But it would be wrong to conclude that we should do less, arguably risk-aversion could induce us to wish to do more, including on the lock downs front.

It is also worth pondering which heterogeneities are “baked in,” such as heat and age structure of the population, and which heterogeneities can be altered at the margin, such as forms of social mingling.  It is at least possible that studying these heterogeneities could make policy far more potent.

Overall, I do not see enough people asking these questions.

Here is the Clive Cookson FT piece (with an irresponsible headline).  Here is the Lourenco new Oxford study, only a few pp.  Miles Kimball offers analysis and numerous references.

Here is the Bendavid and Bhattacharya WSJ piece that perhaps has had the biggest popular influence.  They argue that many more people have had Covid-19 than we think, the number of asymptotic cases is very large, and the fatality of the virus is much lower than we think, perhaps not much worse than the flu.  But their required rate of asymptomatic cases is implausibly high.

The best evidence (FT) for asymptomatic cases ranges from 8 to 59 percent, and that is based on a number of samples from China and Italy, albeit imperfect ones.  Icelandic data — they are trying to sample a significant percentage of their population — suggest an asymptomatic rate of about 50 percent.  To be clear, none of those results are conclusive and they all might be wrong.  (And we should work much harder on producing better data.)  But so far there is no particular reason to think those estimates are wrong, other than general uncertainty.  You would have to argue that the asymptomatic cases usually test as negative, and while that is possible again there is no particular reason to expect that.  It should not be your default view.

Marc Lipitsch put it bluntly:

The idea that covid is less severe than seasonal flu is inconsistent with data and with the fact that an epidemic just gathering steam can overwhelm ICU capacity in a rich country like Italy or China.

Furthermore, the “optimistic” view implies a much faster spread for Covid-19 than would fit our data from previous viral episodes, which tend to come in waves and do not usually infect so many people so quickly.

So I give this scenario of a very low fatality rate some chance of being true, but again you ought not to believe it.  The positive evidence for it isn’t that strong, and you have to believe a very specific and indeed unverified claim about the asymptomatic cases testing negative, and also about current spread being unprecedentedly rapid.

Here is Tim Harford’s take (FT) on all this, he and I more or less agree.

By the way, Neil Ferguson didn’t walk back his predictions.  That was fake news.

So we still need to be acting with the presumption that the relatively pessimistic account of the risks is indeed true.  Subject to revision, as always.

I have been corresponding with a working group regarding the covid-19 situation in Japan.  They shared a draft of their white paper with me while attempting to circulate their revisionist conclusions in policy circles.

The speed premium is indeed increasing quickly.  The white paper has not materially changed since when I first saw it. Since then, the Olympics were postponed and experts in Japan have described the outbreak as “rampant.”  The working group feels that society needs to prepare, and that this outweighs the desire to wait for additional official confirmation.

The authors are an international team based in Tokyo. They cannot attach their identities to the white paper at present.  They are not medical researchers. They have reviewed their conclusions with a medical researcher and others.  You can weigh the evidence of their claims.

Here is the document (no, it is not malware), and here is the opening bit:

The governmental and media consensus is that Japan is weathering covid-19 well. This consensus is wrong. Japan’s true count of covid-19 cases is understated. It may be understated by a factor of 5X or more. Japan is likely seeing transmission rates similar to that experienced in peer nations, not the rates implied by the published infection counts. The cluster containment strategy has already failed. Japan is not presently materially intervening at a social level. Accordingly, Japan will face a national-scale public health crisis within a month, absent immediate and aggressive policy interventions.

There is a great deal of further detail, including the numbers, at the link.  Sobering.

Here is a relevant tweet thread started by Moritz Kuhn, many interesting comments.  For instance Moritz writes: “What is more, it may provide a warning sign for those countries where the elderly and the young live close together, how important it is to contain the virus there early on. These countries are within Europe in particular such as Serbia, Poland Bulgaria, Croatia, or Slovenia.”

Also on Italy, Dan Klein writes to me:

1. They kiss, hug more, converse longer.
2. Young people live with their parents, family more.
3. They smoke somewhat more (packs smoked per capita twice that of Sweden). Smoking weakens the lungs. But also we smokers finger and thumb our cigs and then put them into our mouth. Wash hands first!
4. For these and whatever adventitious reasons, Italy was early to the problem, and it spread before people learned to adjust behavior.

We will learn more soon.