A map of Covid data corruption -- and the Covid approach that worked!
Interesting analysis, thanks for looking into it. I am not sure the amount of lifetime gained / lost as an individual is quite the right way to look at the positive side of the ledger, however. You also have to consider the loss of expected lifespan for your friends, parents, children, favourite celebrities, etc.
"would you have preferred to live through a total travel ban and total lockdowns, like Australia’s, to save yourself 10-to-15 days of life? " is a different question than "Here’s a question: would you have preferred to live through a total travel ban and total lockdowns, like Australia’s, to save EVERYONE 10-to-15 days of life? ".
Very great work here.
Is the data available to do an "under 50 excess death"?
I suspect any "Lockdown Related" deaths of despair will be most visible there.
1) It's a disservice to leave out the well-publicized and much maligned control group: Sweden.
2) The number of recorded suicides was lower, but deaths of despair (including overdoses) *skyrocketed* in the US.
It would be amazing if you could control for age directly in the charts somehow (per million people over 65?) and if you could do US states. California vs Florida??
You ask: "Here’s a question: would you have preferred to live through a total travel ban and total lockdowns, like Australia’s, to save yourself 10-to-15 days of life?" However, going by PeeDee's account elsewhere in this comment section, New Zealand, which had a very low COVID death rate, didn't intrude that much into day-to-day life, instead having only a more-or-less total international travel ban along with brief lockdowns. I see no reason that Australia, which is the example you mention, couldn't have had a repsonse like that of New Zealand, which was very successful at preventing deaths but didn't involve long total lockdowns - I think that Australia's response was simply poorly executed. So that question seems to have a false premise: it overstates significantly the government intrusion necessary to have very low COVID deaths.
(IIRC, China was able to have a similar response, with very few deaths but not that much intrusion into day-to-day life outside the border, but I'm not confident that I have an accurate picture.)
I think countries fall into four groups, broadly. There were countries like Canada and the United States (or western Europe, though I'm not very familiar with them), that didn't have significant border closures, and tried to control spread through lockdowns, to partial success. There were countries like Russia, that didn't attempt to control spread at all, and had about one death per 1200 people. There were equatorial African countries, like Kenya and and Togo which you mention, which seem to have some sort of magic that nobody else has been able to replicate. And there is New Zealand (or Australia, if they'd had a better-executed response, or China, from what I understand), which endured brief lockdowns and a more-or-less total border closure, but had very few COVID deaths.
Given those options, the question now becomes: if you can't replicate the magic that equatorial African countries seem to have, would you prefer to be in Russia, the US or Canada, or New Zealand? Given that New Zealand's response doesn't seem to have cost that much, I'd prefer New Zealand's response, but I can understand wanting to roll the dice on the Russian response, even if I wouldn't do it myself. I think the US's and Canada's responses were the worst of both worlds, and it's hard to justify unless it's somehow entirely impossible to close the border.
It also seems like a good idea to figure out why the countries in equatorial Africa had such a low death rate and if it's possible to replicate that elsewhere.
I've been wondering about the comparison with excess deaths. Thanks.
About the tradeoff between life-days and freedoms; I was in New Zealand through this and yes, travelling internationally became almost impossible. However, aside from a short-sharp lockdown the first few weeks to stamp out the only community cases, and periodic alarms when something slipped through the border, we all lived pretty much normal lives until the last months when Omicron got away (was released into the wild). Our total effective lockdown domestically was several months at most.
Rawls is rolling in his grave because you assumed risk neutrality by using E[days]. Didn’t he use the maximin decision rule in his veil of ignorance thought experiment? That would lead to a very different conclusion. Maximin is extreme, but so is risk neutrality.
Brilliant, persuasive, comprehensive.
You might want to make some mention - without delving into - the political aspects of it all, as some might imagine you are avoiding it all.
Apart from being an island, Australia is also has a population distribution that is particularly amenable for effective lockdowns. The vast majority of the population live in urban centres that are many hundreds of kilometres from each other, and only few people regularly cross state borders. Closing off state borders was therefore a much more practical option in Australia than it could have been in the US or in Europe.
I am pleased to see somebody looking for the fudge factors. Thank you for doing this. But, especially when it comes to Sweden where I live and am familiar with the data there are other fudge factors that may greatly vary between countries and make it hard to make comparisons
First, remember that covid deaths are counted here as 'anybody who tests positive for covid' and dies. No attempt was made to distinguish between 'death from covid' and 'death from something else, including things such as trauma from a motorcycle accident, in a person who tested positive for covid' and 'death in a person who was dying of many things, one of which was covid'. Many other places reported like this, while other places tried to distinguish between people who were dying 'with' and dying 'of' covid. Once you start relying on somebody making a human judgement as to what caused the death, you are immediately up against 'and was there an incentive for people to over-report or under-report deaths due to covid?' This is a great reason to take the subjective factor out of the data, as long as everybody is aware that you are going to be systematically overcounting covid deaths. But by how much? When anybody, including Swedish epidemiologists and the statistic bureau talks about 'deaths from covid vs deaths with covid' you have to know they are guessing. Informed guessing, but still.
So 'all cause mortality' begins to look better and better as a metric. But there are problems for the data analyst here, too. In the Northern Hemisphere, death is seasonal. Winter is the time when most of the dying happens, because this is when most elderly people die of age related things, and respiratory infections. But the calendar year, unfortunately, turns over in the middle of the death season. You really would like to compare years with a start date in the summer or September. If a disease comes by and kills off all the people it is going to Nov-Dec and then doesn't kill people the rest of the winter, it makes the November year look particularly bad while the January one looks particularly good. Since it is common here in Sweden for infectious diseases to do most of their killing in a 2 month period, but sometimes this starts early and sometimes this starts late it can make for a lot of fudge factor in your analysis.
My friend Tommy Lennhamn played around with the Swedish data, to see how much this matters.
see: https://softwaredevelopmentperestroika.wordpress.com/2021/06/20/sweden-mortality-the-period-of-balancing-the-books-matters/ where picking the month to start matters. Swedish “Excess Deaths” go from 1300 excess to a deficit of 1500 (though this is YTD May 2021, since the calculation was done June 2021).
Picking your year to start in the fall works well in the Northern Hemisphere. It doesn't work well in the Southern Hemisphere, and a simple change (so pick it in the beginning of _their _ fall) doesn't work either, as a lot of the southern hemisphere doesn't have a single yearly death hump, but two. (And while theories have abounded about this since it was first noticed, we still do not know why this is so, though the monsoon seems to be important for places that have one.)
The largest elephant, however, I see in a Swedish fudge factor is what to do with the year 2019? It is a hard problem. In Sweden the great data outlier in mortality calculations is not the large number of excess in 2020. It's the even larger deficit the year before. Never, ever since we have been keeping records have so few people who were expected to die held off the grim reaper for another winter season. This means that when covid struck in March there was an exceptionally large group of old people who were likely to die if they caught it still alive. Now, if your immediate thought was 'if only granddad had died before last Christmas like he was supposed to ..' then you need to get your head out of the data for a while, and rediscover your humanity but when it comes to modelling the baseline of expected deaths you do have a problem. Not only will more people die in 2020 than is usual, you will have undercounted 'as usual' by including 2019.
Most places where I have seen analysis have decided to just average the last 5 years before covid and assume that they were normal. They will have both problems. For statistical purposes, the corner of the world where I work and use this data (and we want to know if we need to start building new elderly homes and if so, how many, and should we open another hospital) the usual thing to do is to throw away that year (2019) altogether when considering what is 'normal'. 2020 wasn't normal either.
One of the things we would dearly like to know, but don't now, is whether those people who 'ought to' have died in 2019 but didn't were statistically any different from a random selection of people of their age in the country in 2020 when it came to being at risk for a death. Some of them we know were sicker than average ('it is a miracle he has lived so long!') but was this statistically significant? When calculating 'how many people are expected to die this year' should we stick a powerful fudge factor into our expectations '+lots because so few died last year but we expect most of them to die this year instead' or a minor one of longer duration 'over the next five years we expect more than the usual number of people to die as the excess survivorship is trimmed'? It will be a while before we know what would have been the correct thing to do. It looks like 2 winters of covid were enough to kill off the surplus, but if people start dying in droves this winter, we will rethink that. But we are erring on the side of being better to open a care home you don't really need than to really need one you don't have.
Right now, in Sweden the new figures for August 2022 showed the first excess mortality for 18 months. It was tiny, but we worried that this might be a trend. But it looks as if the numbers are down again for September. So maybe it's a case of 'nothing to see here, within expectations'. Or maybe something else is going on. But it is worth looking at what excess mortality looks like, with a year end in August and mortality bars for each month. Note that this is using the 'let us not use 2019' calculated baseline.
Here is Tommy Lennhamn again:
Eurostat is using 2019 this and so they are showing Sweden as having low excess mortality (instead of none or surplus) in places you may have read such as here:
where you can make the interactive display show excess deaths for the European country of your choice.
Nobody is being dishonest here. Figuring out what to use as the baseline is hard. But it is a huge fudge factor on its own.
I'm not sure what the general experience was, but is getting Covid while unvaccinated much more unpleasant compared to getting it while vaccinated?
Because in that case it's not just a loss of 10-15 days expected days of life, but somewhat more. (For example mild symptoms for 1 week compared to severe symptoms for 2-3 weeks.)
Also to what extent was it necessary to have at least temporary shutdowns/restrictions during the early surges of Covid to prevent the health care system from shutting down? I remember a lot of fear was related to fatality rates potentially skyrocketing if hospitals run out of equipment and personnel. Once most people were vaccinated this risk seemed minor at most.
What a superb piece of analysis! As one who followed the national per capita death rates obsessively for the first couple of years of the pandemic, it's refreshing to see an analysis that understands the basic concept of "per capita" (as much mainstream analysis did not). The careful "excess deaths" analysis here adds another important level of understanding.
One minor quibble. I don't think you can attribute the difference in Australian and US death rates solely to the differences in the government-mandated lockdown policies of the two countries. Vaccinations were not mandatory in either country, but Australians chose to get vaccinated at significantly higher rates than Americans. (The World in Data website indicates that today 84% of Australians are fully vaccinated, while only 67% of Americans are fully vaccinated.) Surely, some of the differences between the US and Australian death rates are attributable to these individual choices and not to the differing government mandates.
Do you have a color key somewhere that shows what each represents?
I like the work the author has done. I feel it still leaves out some things like deaths due to the jab (which are also baked into the "excess deaths" figures); the economic impacts "compared to what?" (i.e. what would things have looked like with no Covid, a not inconsequential question); and some long term things we really can't quantify yet like long term suppression of the human immune system some say the jab causes. But, sanity has to start somewhere and this is a good start.
I wholeheartedly agree that the lock downs were never worthwhile to anyone other than those who lust for the exercise of raw power. Historically, totalitarian measures never benefit anyone but the totalitarians.
India's low death rate is more likely true than not. Unable to afford vaccines, the govt turned to use Ivermectin. The lifesaving results were spectacular. Some regions of the country saw a reduction in Covid deaths as high as 87%! Conversely, we saw 100s of thousands, probably millions, die due to our so-called health experts banning the use of Ivermectin and threatening medical doctors if they dared to use the drug off-label. My physician has treated more than 700 Covid patients with Ivermectin, Zinc, and Vitamin D3, and has had only one hospitalized and ZERO deaths. Despite such a sterling record, the Texas Medical Board has threatened his licensing if he continued to prescribe Ivermectin. Made ZERO sense. Why would that be happening? Shall we follow the money? Ivermectin is pennies on the dollar when compared to current treatments. From Remdesivir to plasma or new pills, all these treatments are expensive! Vaccines alone have made companies multi-billions! Covid has been a cash cow for hospitals. Reports were that Medicaid and Medicare were paid up to an additional $40K per patient. Why would these entities want to allow the use of low-cost medicines when they are killing it financially? They didn't.
It seems like a simple deaths per 1M population would provide a baseline of mortality in various countries. That could reveal some interesting clues that can be hidden when using ratios. For one thing, did African have better outcomes using ivermectin or HCQ without vaccines? That would be interesting to see, but data isn't available for most of those countries. What's up with that?
I like the excess death approach. I'm disappointed that you don't provide a key or legend to share the meaning of each color that you used in the maps. Without a key to decode the colors these maps remain ambiguous. For example what does white mean? What does green or cream mean? What does pink, light red, red, dark red mean? In one place you mentioned black. However, all the colors need to be associated with what they mean. Please update this post and provide information that will make your maps useful and understandable.