Tyler Cowen recently made this claim:
That is the abstract of a new NBER paper by Andrew Atkeson, Karen Kopecky, and Tao Zha. You will note that when it comes to Covid-19 cases, the superior performance Europe had enjoyed over the United States seems to be evaporating.
I view case data as unreliable, especially when making international comparisons. I prefer mortality data, which shows the US to be doing far worse than Europe, a gap that continues to widen dramatically:
In fairness, deaths lag infections by several weeks, so it’s possible that the US/Europe gap will soon begin to narrow. My prediction, however, is that a month from now the gap will be much wider than today. I further predict that when vaccines become available (probably in 2021) the gap will be still wider. And I predict that studies will show that the gap reflects more than just different definitions of Covid-19 deaths, that excess death data will present a similar picture.
This graph is quite striking in that is shows that Covid-19 is currently a mostly Western Hemisphere problem. The absolute number of deaths in countries like India is fairly large, but in per capita terms most of the deaths are occurring in the Western Hemisphere.
Western Hemisphere countries don’t have a lot in common. The US is rich and Honduras is poor. One thing these countries do have in common is high murder rates—by global standards. Of course Canada is the exception. But notice that Covid-19 deaths rates in Canada resemble Europe more than other Western Hemisphere countries. Perhaps Canada should be viewed as a European country stuck in the Western Hemisphere.
PS. Elsewhere I’ve argued that America increasing resembles what used to be called a ‘banana republic’. Perhaps that term in no longer PC, but the phenomenon hasn’t gone away.
READER COMMENTS
Jon Murphy
Aug 30 2020 at 1:40pm
Do you know how “probable” deaths are counted in Europe? Are they counted in Europe?
JFA
Aug 30 2020 at 2:40pm
One way to see if measuring probable covid deaths makes a difference to the overall number would be to look at excess deaths and compare that to reported deaths. For the US, the number of observed deaths above the 95% threshold on expected deaths (i.e. excess deaths) from March 2020 through Aug. 15 (the date of the latest CDC data) is 183,392. That’s with a few states on longer time lags and with delayed reporting of deaths in general. By August 15 there were 168,000 reported Covid deaths, and as of today (8/30 there are 183,000 measured Covid deaths). How the US categorizes potential deaths doesn’t seem to matter that much to the US numbers.
Here are a few sources for other countries (these sources use their own models for excess deaths; I just used the CDC’s definition): https://www.economist.com/graphic-detail/2020/07/15/tracking-covid-19-excess-deaths-across-countries, https://www.nytimes.com/interactive/2020/04/21/world/coronavirus-missing-deaths.html
Sweden, France, Switzerland, and Germany have excess deaths very close to the reported number of Covid deaths. The number of reported Covid deaths in Spain and Italy (where Covid hit hardest) is at least 15 percent lower than the number of excess deaths. So reporting potential deaths is probably not that big of a factor, except maybe for Spain, but any error is probably not going to equalize Europe’s outcomes to the poor performance of the United States.
Jon Murphy
Aug 30 2020 at 3:13pm
I see. Thanks!
Scott Sumner
Aug 30 2020 at 9:52pm
Thanks JFA.
Alan Goldhammer
Aug 30 2020 at 1:54pm
I read the paper this morning and was unimpressed. I’ve read so many non-epidemiological model papers since mid-March and they all have flaws of one kind or another. The assumptions in the current paper are wrong but I guess using ‘death’ as a marker makes it easier to do model constructs. I usually mention these papers in passing in my newsletter as I don’t have the patience to pick apart all the problems.
RPLong
Aug 31 2020 at 9:33am
A few years back, there was a push from within Canada to create a Canadian-European-Union economic zone or free-trade zone. Unsurprisingly to anyone familiar with Canada, it failed because Canadians didn’t want it. The commentary was interesting, though. Canadians didn’t want it because they didn’t want Canada to become “too American.” I remember one op-ed saying that Canada was more European than Europe, at least when it came to trade.
Alas, another missed opportunity for greater trade liberalism. Perhaps if Canadians start to see free trade as a European ethic, rather than an American one, they’ll warm up to it.
Warren Platts
Sep 1 2020 at 4:05pm
If free trade was a European ethic, they would not practice mercantilism.
Todd Kreider
Aug 31 2020 at 5:30pm
“Perhaps Canada should be viewed as a European country stuck in the Western Hemisphere.”
Aug 30th….deaths…….if U.S. population
Germany……….9,300………..37,000
Canada……..9,100……..78,000
Netherlands….6,200………..118,000
France………….30,600……..153,000
Sweden…………5,820……….186,000
U.S……………..187,000…….187,000
Italy…………….35,500……..195,000
U.K……………..41,500……..203,000
Spain…………..29,000…….203,000
Belgium….. …..9,900……..297,000
Scott Sumner
Aug 31 2020 at 8:21pm
This is called changing the subject. Do you disagree with the data in the graph I presented?
Todd Kreider
Sep 1 2020 at 4:31am
You wrote: “I prefer mortality data, which shows the US to be doing far worse than Europe, a gap that continues to widen dramatically:”
I prefer to stay in reality.
JFA
Sep 1 2020 at 9:28am
So in your reality, Europe is only made up of 8 countries?
JFA
Sep 1 2020 at 9:52am
Also, if you just take the countries you listed and used them as a comparison group, they still do better than the US. For the group of countries you listed (minus the US), deaths per million is 442 vs 555 for the US. And even though you were using mostly the worst performing European countries, the deaths per million is still 20% lower than the US. If you leave out Canada, the comparison groups deaths per million is 462 (still 17% lower than the US).
The jury is still out on who will have a superior performance because this thing is going to go on for a while, and the US is looking good on the case fatality rate (though that’s potentially biased by less than stellar testing in the US (it has a lower number of tests per confirmed case relative to the comparison countries you listed)). But currently, the scoreboard does not have the US doing well.
Scott Sumner
Sep 1 2020 at 1:33pm
Todd, I showed that Europe’s mortality rate is only one half the US level. I’m not sure what point you are trying to make.
Even worse, the gap widens further every single day.
Warren Platts
Sep 1 2020 at 4:39pm
The numbers in your graph are correct. However, averages hide a multitude of sins. Several European countries–Belgium, Spain, UK, Italy, Sweden, (and San Marino & Andorra)–have higher death rates than USA.
Within USA, six states account for fully half of all deaths (New York, New Jersey, California, Texas, Florida, Massachusetts). Four states (all traditionally liberal Democrat states) have death rates more than twice the national average (New Jersey, New York, Massachusetts, and Connecticut).
Meanwhile, the bottom six states (Wyoming, Alaska, Vermont, Hawaii, Montana, Maine) collectively account for 441 deaths as of today (0.2% of all SARS2 deaths). Yes, these states have small populations (collectively 2% of the total population). However, their deaths/million population is 76.5, a mere 13% of the national average, and far lower than either Canada or the EU.
Thus there is a huge amount of variation within the data. It is very skewed. Note that the median U.S. deaths/million among states is ~335–less than 60% of the national average.
I understand the game being played here: by painting the nation with a broad brush, the failings of blue state mayors and governors to deal with a virus that likely originated from a Chinese laboratory can be laid on Trump’s doorstep. But a little bit of digging shows that is really unfair..
derek
Sep 1 2020 at 9:18am
So relatively high murder rates in the Americas are somehow reflecting a factor that explains higher coronavirus transmission? Weird.
Todd Kreider
Sep 1 2020 at 10:51am
You have used the word “performance” a few times when what you want to use is “outcome”.
Neuroscientist/statistician Karl Friston’s group has shown that government policies have had little impact on Covid-19 outcomes.
“His models suggest that the stark difference between outcomes in the UK and Germany, for example, is not primarily an effect of different government actions (such as better testing and earlier lockdowns) but is better explained by intrinsic differences between the populations that make the “susceptible population” in Germany — the group that is vulnerable to Covid-19 — much smaller than in the UK.
“As he told me in our interview, even within the UK, the numbers point to the same thing: that the “effective susceptible population” was never 100%, and was at most 50% and probably more like only 20% of the population.
https://unherd.com/2020/06/karl-friston-up-to-80-not-even-susceptible-to-covid-19/
Todd Kreider
Sep 1 2020 at 10:52am
That was supposed to be to JFA.
JFA
Sep 1 2020 at 1:40pm
I think there is merit to using performance, but “outcome” might be better in some cases (much like “suggests” is a better term than “shown” when discussing any results about Covid related epidemiology). That being said, I don’t really buy Friston’s idea of “dark matter” epidemiology. Shut downs really do seem to have effects on the spread of Covid (it seems a big coincidence that the decline in cases starts 2-3 weeks after lockdowns are imposed in just about every place it is tried), and personal behaviors (which can be influenced by policy) have an effect on the spread of Covid. I think getting testing off the ground earlier and faster in the US would have aided in containing the spread of the disease without causing huge disruptions. These are the considerations that I think warrant some reference to performance.
Todd Kreider
Sep 1 2020 at 3:03pm
“Shut downs really do seem to have effects on the spread of Covid (it seems a big coincidence that the decline in cases starts 2-3 weeks after lockdowns are imposed in just about every place it is tried)…”
Norway says its lockdown didn’t help, was a mistake, and that they won’t do that again. A graph of New York shows their lockdown did nothing to slow the virus, although I’m not sure if the virus in NYC was slowed down to some extent.
Japan didn’t have a lockdown but on April 7 the government asked 40% of the population – the largest cities – to stay home as much as they could, although the culture there really makes working away from an office very difficult. On April 16th, the government extended the stay at home if you can request to the entire country yet cases had already stabilized at increases of 9% a day the April 7 emergency was declared and on the way down before the April 16 emergency was declared. Especially when a few day lag time of reporting cases is considered, it looks like the stay-at-home request didn’t have much effect (if any) on slowing the spread of the virus. Once the stay-at-home request ended in late May, the increase in cases per day stayed at 0.3% for six weeks before getting up to 3% a day in late July. Without any special actions, that fell to 1% a day last week and keeps decreasing.
JFA
Sep 1 2020 at 8:49pm
“A graph of New York shows their lockdown did nothing to slow the virus”. New York and NYC issued various policies between March 12 and 17 (including the stay at home order) and daily cases peak on April 5 or so (about 3 weeks after the policies). The same occurred in China when lockdowns were imposed and 2 weeks later cases reach a peak (the peak was reached quicker because the lockdown was probably tighter: https://www.phrp.com.au/issues/march-2020-volume-30-issue-1/on-a-knifes-edge-of-a-covid-19-pandemic-is-containment-still-possible/). The lock down of northern Italy occurred on March 8 with a nationwide lock down issued on March 9; cases in Italy peak on March 28 (three weeks after the lock down orders). In Spain a nation lock down is imposed on March 14 and 2.5 weeks later (March 31) daily cases peak. Germany imposed a lockdown between March 20 and March 22, and on April 5 (about 3 weeks later) cases peak. Norway imposed a lockdown on March 12 and cases peaked on March 30 (about 3.5 weeks later). I could go on. Now as things are opening and people are getting more comfortable going out, you see cases in many of these places increasing. If the early decreases were just due to the virus running its course you wouldn’t expect to see any increases.
Were lockdowns the best policy? Almost certainly not. Were they the only policy? Definitely not. We could have been like Japan and told everyone to stay at home, talk really really quietly, and wear masks (side note: I had a friend who went there at the end of February and he said that the areas that would usually have been packed at night were deserted). But to say that lockdowns (if they are actually complied with, otherwise they wouldn’t be lockdowns) have no effect is just being willfully obtuse.
Todd Kreider
Sep 2 2020 at 12:48am
A lockdown can have only a potentially short -term effect and every epidimiologist knows this. Fauci said in late January that “…historically it doesn’t do much good, which is true.”
Here is a graph that compares New York state with Sweden in terms of deaths (bright solid lines) and cases (faint dotted lines): https://fee.org/articles/why-sweden-succeeded-in-flattening-the-curve-and-new-york-failed/
The Norwegian government says its R0 was near 1.1 when it locked down so said it was clearly a mistake not to follow Sweden as they initially planned.
I have friends in Japan and even in early April many people were taking trains and subways in Tokyo. One friend sent me a photo of the inside of her train with the caption: “NOT social distancing!” However, she said that Tokyo was very quiet then after 9 pm, even on weekends.
In Europe, case increase currently has a wide variance. Cases are barely increasing in the UK and increasing at only 0.2% a day in Sweden. Germany’s and Italy’s case increases have also only slightly moved up to 0.5% a day. Cases in Spain have been higher at 2.0% a day and France is at over 3% a day.
Europe as a whole has cases increasing at 0.9% a day, whereas U.S. cases have been increasing at 0.4% a day.
Todd Kreider
Sep 1 2020 at 3:41pm
I forgot to add with the Japan part that despite all of that, Japan’s per capita deaths have been 60 times lower than the U.S., U.K., Spain, Italy and Sweden. I’d think Fiston’s “dark matter” hypothesis is all that could explain such a huge difference.
P Burgos
Sep 1 2020 at 7:57pm
Isn’t Canada the only country in the Western Hemisphere to have not really practiced slavery/forced labor as a part of their economy? Did they ever have a racial caste system like the US and Latin American countries? Additionally, how many other countries in the Western Hemisphere have parliamentary systems of government, as opposed to presidential systems?
Michael Rappaport
Sep 2 2020 at 6:09pm
This debate seems peculiar. Perhaps I missed it, but the debate overlooks a key factor. The most important risk factor for covid deaths is obesity. The U.S. has much more obesity than European countries. How can you compare without considering this key factor?
Comments are closed.