The definition is that region Y has many more deaths per million than region X. Examples include:
(a) The U.S. and Europe vs. Asia (Taiwan, Singapore, Hong Kong, Japan).
(b) NYC vs. the rest of the U.S.
(c) Italy and Spain vs. the European countries that did much better.
(d) Sweden vs. the European countries that have done better.
(e) The European countries that have done worse vs. Sweden
What stories can explain these?
For example, could there be a different strain of the virus attacking different regions? This would be a plausible explanation for (a) and (b), but it is less attractive for explaining (c) and very implausible for explaining (d) and (e).
Did the infection spread more rapidly in the Y regions than in the X regions? Note that if it is true, then the Y regions now enjoy more immunity. In the short run they are worse off, but they are less vulnerable to a second wave.
Did the virus strike more at-risk people in the Y regions than in the X regions? This is one of the explanations that I like to believe. I tell it as a story of cultural differences. For elderly people living independently or in nursing homes that have hygienic skill/luck going for them, death rates will be low; otherwise, they will be high. For people who work in jobs that force them to encounter many people in enclosed spaces (including a need to commute by subway), death rates will be high; otherwise they will be low.
Does health care quality make a difference? Would more people have lived in the Y regions had their health care systems not been overwhelmed? That may help explain (c), but overall I get the sense that health care quality is not making nearly as much difference as you would think.
I suspect that cultural differences are more important than policy differences. For example, mask-wearing is more widespread in Asian countries, even without government mandates.
Sweden is held up as an example of a country that allowed people to go to restaurants, but a closer look reveals that Swedes were advised to socially distance and they were very inclined to comply. There may have been at least as much social distancing in Sweden as there was here in states where governors issued strict rules.
If I am correct about this, then Sweden is not such an interesting experiment. The explanations for (d) and (e) may have more to do with relative size of at-risk immigrant working-class populations than government policy. By the same token, I would not expect Sweden to achieve “herd immunity” much faster than places that implemented government lockdowns.
Asian countries are held up by some people as an example of more aggressive at testing, tracking, and tracing. But when one works through the requirements for pulling this off successfully, I have my doubts. Just the number of false negatives that you get from testing would seem to me to present a major challenge to implementation. My bet is that face masks account for more of the difference between Asia and the West than does test, track, and trace. Maybe a lot of temperature-taking also helps, especially if people with fevers shed a lot more virus than people who don’t.
You may be getting tired of me saying this, but we would have a much better handle on these issues if we were conducting rigorous studies instead of relying on computer models. Flying blind is a choice. Sooner or later, epidemiology has to be taken over by real scientists.
“Sweden is held up as an example of a country that allowed people to go to restaurants, but a closer look reveals that Swedes were advised to socially distance and they were very inclined to comply. There may have been at least as much social distancing in Sweden as there was here in states where governors issued strict rules.”
Please provide your evidence, a simple google search indicates otherwise.
https://www.google.com/search?q=sweden&source=lnms&tbm=nws&sa=X&ved=2ahUKEwiKgYz0q-_oAhULWq0KHTNpD7wQ_AUoAXoECCEQAw&biw=1920&bih=937
watch the video with Sweden’s health minister. Maybe a more definitive source than your Google search.
And tone it down. I’m close to deleting your comments again.
I did watch it and have read a lot of articles from reputable sources on the topic (many included in the google news search link that I provided).
My conclusion: their version of lockdown (call it lockdown lite) is better than ours and may better position them for the next steps.
And also, Scott Sumner here:
https://www.econlib.org/sweden-is-not-the-model-taiwan-is/
Arnold,
Thanks again for your daily insights during the pandemic. I would like to ask a question about your interpretation of Sweden’s approach, from the perspective of regional heterogeneity.
Re: “There may have been at least as much social distancing in Sweden as there was here in states where governors issued strict rules. If I am correct about this, then Sweden is not such an interesting experiment.”
Q: Isn’t the Swedish experiment nonetheless interesting, and indeed crucial, simply because Sweden is trying to traverse the pandemic with minimal institutional disruption (e.g., schools are open) and without enacting new or more government coercion, whilst keeping its nerve (under intense international scrutiny), and hoping and trusting not to become a negative outlier in mortality rates?
I would also like to reinforce your observation about regional heterogeneity, re: face masks: “I suspect that cultural differences are more important than policy differences. For example, mask-wearing is more widespread in Asian countries, even without government mandates.” The Sweden’s approach to masks confirms the specificity of social norms, I think.
Sweden’s Chief Epidemiologist, Anders Tegnell (video interview, 15 April 2020), discusses face masks at some length, and expresses skepticism about requiring masks outside the home, except in medical settings, elder-care facilities, and the like. If I understand correctly, the gist of Tegnell’s argument is that a mandate for face masks outside the home would have the unintended consequence of undermining Sweden’s crucial current public-health norm: Stay at home if you are in any way symptomatic! If there were a mask mandate, then a substantial subset of symptomatic individuals would persuade themselves, mistakenly, that masks adequately protects others. These symptomatic persons would break the norm of self-isolation and would go out with a face mask. This behavior would increase other people’s risks of getting exposed to the virus. This anti-social behavioral effect would outweigh any modest protection from infection that masks provide. This is a Peltzman effect, but with the emphasis on behavioral adaptations that impose more risk on others. In sum, Sweden’s approach about masks reflects a commitment to norm-management and behavioral economics.
I wouldn’t discount random luck. There was a time (say late Feb) where no one was paying attention to Coronavirus but we now know it was spreading rapidly. If Y had 100 infections on February 20th and X only had 1, then even if Y and X follow identical paths from that point (identical infection rate, type of people who are infected), Y will have 100x as many deaths today. That’s the nature of exponential spread.
I think a big factor is actually whether the initial cases in an area hit a mostly asymptomatic group or a mostly high death-rate group. In Washington it was the latter (a nursing home), which then called that state’s attention to the infections almost immediately, prompting a quick response. In NY/NJ, I suspect it was mostly asymptomatic (or lightly symptomatic) people who got it initially, which masked the extent of the infections until perhaps 1% of their population was infected.
With exponential growth, the initial state just matters so much early on, and I think that has still swamped everything else in many places. E.g. the difference between 20% and 30% growth over a month is only going to erase a 10x difference in initial conditions. Besides a few sparse areas (e.g. Vermont), or Washington (which started their response early), no place been able to cut the growth rate by anywhere near that much relative to the rest of the country. And even if a disease hit mostly young healthy people vs mostly old unhealthy people in two areas, that’s not going to erase more than a 10x difference in initial infections.
No, the difference between SK and Italy is not due to this, and perhaps the difference between Italy/Spain and other western countries it not either. But I think most of the US heterogeneity is due to initial conditions. And I think most of Europe too.
Maybe it is all random variation. Unit root processes are very unstable. They can easily blow up or die down just by chance.
Going to strict border controls early, for foreigners and citizens alike, with health screening, mandatory quarantines, and follow-up, made big differences between otherwise similar countries. Australia is similar to Canada, but being an island and going strict on border control early means they are doing a lot better.
There is some evidence vitamin D deficiency is correlated with worse covid outcomes: https://www.wsj.com/articles/vitamin-d-and-coronavirus-disparities-11587078141
I suspect that this explains in part the relative better outcomes in warmer locations like Taiwan and Australia. Bright sunlight is also thought to be inhospitable to the virus. One wonders also if women have less mortality than men because on average they prefer warmer home temperatures than men. Somebody should test for correlations between average residential heating costs and covid morbidity. The cold spring in the USA is surely not helping.
Warmer weather also allows greater freedom from being trapped in enclosed spaces in which transmission is more likely to occur. Anecdotally the “hospital in the parking lot “ arrangement resembles the community health clinics in Minas Gerais, Brazil, that I have used: open air waiting rooms and many without HVAC. So far Belo Horizonte in MG has had less mortality than São Paulo and I wonder if that is because BH stopped international flights earlier and also has a higher percentage of the population living without HVAC? Pure conjecture but it is a heterogeneity nevertheless. Overall, I believe that the SUS is superior to the USA health system and I was deeply appreciative of the nobility of a health care system that allowed me free access even when I was only a short term visitor.
It is all well and good to jump on the bandwagon now and condemn Sweden but as Bjorn Lomborg was tweeting the other day, a Science article that took into account that pandemics tend to come in waves determined that a strict lockdown was not the best policy. It will not really be possible to make judgments until we get longer term excess mortality figures.
And Steve Sailer raises a lot of interesting questions about how differences in previous vaccinations might account for heterogeneity.
Quartz magazine suggests high heat and humidity as a possible explanation of why India has so few cases: https://qz.com/india/1839018/why-does-india-have-so-few-coronavirus-covid-19-cases-and-deaths/
I suspect a relatively large percentage of the population also lead less enclosed lives as well and do not have tightly insulated homes with HVAC.
Sooner or later, epidemiology has to be taken over by real scientists.
They are real scientist but this is new and we have to do best as possible! And forgive some transgressions. I bet if you analyzed all FDRs WW2 efforts, we could find all kinds mistakes. (Spruce Goose would be one example!) Anyway private companies have to deal with 90% of data all the time.
Considering the disease started spreading heavily in January 2020, it seems reasonable to state regions, nations, and 50 states are all running a separate experiment with no control subjects. Or we don’t have a long term Null Hypothesis to disprove here. I live in California and if any state should have been burning with spread, it would be California before any other state. Why did spread hit Washington first, it just luck? (Or if you ran 100 different Covid computer models, California burns first 60% of the time but 25% it would have been Washington.)
1) I still believe the Pacific Rim with experience of SARS and other diseases acted first because the population went through the 7 stages of grief quicker. South Korea amped up the test and wear mask almost immediately probably 60% of the population remembers SARS! Experience is a great teacher.
2) And yes you can analyze a lot of data about the US and realize everybody was starting to understand the dangers and were moving accordingly. I went to Red Robin in early March and it was obvious at 8:00 PM the restaurant was 50% less full than capacity which is normally full at that time.
3) US disease and death stats are exceptionally good long term but without a crisis it was 90% fine that it takes 14 months to collect. For decades the death stats move slowly and costs a lot of money to get better up-to-date stats. Now suddenly we need daily stats fast and nobody was ready for that.
4) In terms of spread, use less day-to-day stats and I like the 7 day average of cases and deaths. So I look up LAT and the 7 day average of cases is below last week number! It is declining here.
5) We are still going to have surprises. There will be a meat plant spread (or warehouse ) like Smithfield SD.
I still there two future realities here:
1) In terms of South Dakota, some of rural America could really be hit hard. It is very contagious and little hotspots could pop up.
2) I extremely worried by Global South as their resources, financial & health, are not as strong. South America could be a big hotspot long term but Africa has been more contained. (Note Africa like Pacific Rim has disease spread experience.)
A major source of regional heterogeneity was risk perception. Global travelers avoided travel to Asia and areas perceived to be a high risk for getting the virus. Once the virus started spreading unnoticed in Italy, infected asymptomatic global travelers continued to avoid travel to perceived high risk areas. There was a risk inversion where areas perceived to have the highest risk had the fewest infections. Asia and heavily Chinese cities like San Francisco had fewer infections while Europe and New York had more infections.
To the extent distancing is effective, its effect is surely heavily driven by speed and what I’ll call “Effective Distance.”
In WA state, when the “spread out” urging came (before the orders), several very large local employers (Microsoft, Amazon) had a LOT of people start working from home quite fast (like in the space of 48 hours.) We don’t have much mass transit here, and relatively few multi-generational famalies. I think on the whole we have fewer instances of large close groups crammed into small dwellings. We certainly have poor people, and quite a lot of homeless people, and that later was seen as a problem early on. There also people living in very small semi-shared dwellings called “apodments” in Seattle – but that seems a small scale trend for King county as a whole. So, on the whole, our Effective Distance rose quite rapidly. (While nursing homes have been a non-stop tragedy…)
[I live in suburban Seattle metro, which is in King County, WA. The above is direct observation. The nursing home where the pandemic blew up in WA is about 2 miles from my house.]
My understanding is that NYC is hit worst in the Bronx and now Queens – surely among the two highest density areas of that size in the US. Surely more multi-generational families crammed in small spaces. And it would not be physically possible for Seattle to have anything like the transit use that NYC has.
And so even if all of those people (who don’t all speak English!!!) instantly got as socially distant as they could, what effect would that have given the large family units, and the heavy congregation in mass transit?
That is, is it just the case that NYC, northern Italy, Spain, etc., just couldn’t get enough Effective Distance (ever), and couldn’t get what was possible fast enough?
(I’ve seen claims that Spain’s issues are mostly around Madrid….)
I have no idea how this applies to Belgium, the Netherlands, or France….