The Volokh Conspiracy
Mostly law professors | Sometimes contrarian | Often libertarian | Always independent
No-Lockdown Sweden Seemingly Tied for Lowest All-Causes Mortality in OECD Since COVID Arrived
Focusing on all-causes mortality, and not just on COVID mortality, helps account for various potential indirect effects of lockdowns.
COVID lockdowns could have saved lives, by decreasing COVID deaths (and also incidentally decreasing some other deaths, for instance from auto accidents, from other communicable diseases, and the like). They also could have cost lives—for instance, through drug abuse deaths or suicides or homicides stemming from people being cooped up for months, drug abuse deaths or suicides or homicides stemming from economic damage and unemployment caused by the lockdown, cancers not caught early as people delayed early screening (even if such early screening would have been officially excluded from the lockdown), and so on.
What was the likely aggregate of all these effects? In particular, to look at just one data point (recognizing that it's indeed just one data point), how did Sweden, which basically didn't lock down, fare compared to other prosperous countries?
Sweden did have a higher COVID mortality rate than many other European countries, including the neighboring Norway, Denmark, and Finland. But what about what is sometimes called "all-cause excess mortality," which is to say total mortality in the country compared, on a percentage basis, to the pre-COVID mortality? Moreover, some of the effects of lockdowns (positive and negative) could have lasted well after the lockdowns, indeed for years past them. What then about the all-causes excess mortality from 2020 to the present, rather than just based on the 2020 data or the 2020-21 data?
A recent UK Office of National Statistics report says that Sweden and Norway were essentially tied for the lowest "[p]roportional all-cause excess-mortality scores" (which "measure[] the percentage change in the number of deaths compared to the expected number of deaths (based on the five-year average [from 2015 to 2019])" among the listed European countries, looking at data from Jan. 2020 to June 2022: Their excess mortality was up 2.7%, compared to, say, 5.2% for Denmark, 7.1% for Finland, and 11.8% for the Netherlands.
Nor is Sweden's success likely caused by Sweden's age distribution, as measured by the percentage of the population that's at least 65 and the percentage that's at least 80, which are in the middle to high middle among the OECD countries. The UK report provides information on "cumulative age-standardised mortality rates" as well, and Norway just barely inches ahead of Sweden, at -4.1% compared to -4.0%, well ahead of all the other listed countries. (The percentages aren't directly comparable to the excess mortality percentages as such, but they are still helpful for internal rankings within each kind of measure.)
I also tried to do a similar analysis myself, based on OECD data (which covers most of Europe, the U.S., Canada, Australia, New Zealand, and Israel), for Mar. 2020 to Oct. 2022 (the most recent data I could find). Based on that data, Sweden seems to have had the lowest rate, even below Norway. (Others have remarked on this as well.) You can check my data in this spreadsheet; I downloaded the 2020, 2021, and 2022 weekly data from the OECD site, deleted a few Latin American countries that didn't have data for all the years, calculated the average excess death percentages for each year (see row 8 of each sheet), and then calculated the average excess death percentages for all three years (see row 2 of the 2022 sheet).
I also asked our UCLA School of Law Empirical Research Group people to check into that, and they confirmed; here are the aggregate excess mortality percentages they reported (or see the Excel spreadsheet), though I think my data excluded the first 9 weeks of 2020 (as basically pre-COVID) and their analysis included it:
| United States | 20.90 |
| Poland | 20.13 |
| Slovak Republic | 18.89 |
| Czech Republic | 16.19 |
| Slovenia | 14.25 |
| Greece | 13.90 |
| Israel | 13.14 |
| Netherlands | 12.54 |
| Spain | 12.34 |
| Canada | 12.12 |
| Estonia | 11.94 |
| Portugal | 11.79 |
| Italy | 11.73 |
| Lithuania | 11.14 |
| Hungary | 10.81 |
| Austria | 10.80 |
| United Kingdom | 10.60 |
| France | 9.99 |
| Latvia | 9.66 |
| Switzerland | 9.42 |
| Germany | 8.86 |
| Luxembourg | 8.79 |
| Finland | 8.60 |
| Belgium | 8.21 |
| Australia | 8.07 |
| Iceland | 7.98 |
| New Zealand | 7.61 |
| Denmark | 6.12 |
| Norway | 4.28 |
| Sweden | 2.79 |
Of course, there are obvious limitations with such an analysis:
- As I mentioned, this is based just on comparing one no-lockdown country to other countries; perhaps Sweden was an outlier for other reasons (what if, for instance, it had also improved various unrelated health care measures in the last couple of years, more so than other countries had?).
- It would of course be helpful to also control for other factors, such as immunization rates over time, the fraction of the country that lives cheek-by-jowl in major population centers, more precise measures of age, and more.
- The excess mortality data may be imprecise in some respects, though the COVID mortality data might be as well.
- Total excess mortality is a crude measure in various ways; one might want to consider years of life lost (reasoning that, say, the death of a child with 70 years to live is even more tragic than the death of an adult with 7 years to live), or one might want to exclude suicide, or do something else.
- One might also want to try to look at more than mortality, and include serious but not fatal long-term health results (both from COVID and other sources), lost educational opportunities for children, and many other things.
- As one of the people in the Empirical Research Group noted to me with regard to my analysis (which averages the excess weekly mortality), "Looking at and aggregating percentage change in seasonally adjusted time series is something we typically want to be careful about. Mortality rates are seasonally cyclical—in most OECD countries, mortality rates are systematically higher in the winter. In the US pre-COVID, deaths in the winter averaged about 15-20% higher than deaths in the summer, although how much higher in any given year depended a great deal on how bad the flu season was. In principle, when baselines are different, averaging percent change (what I take it you did based on your email) can be misleading. Think about it like this: 30% higher mortality in the winter is a greater number of excess deaths than mortality being 30% higher in the summer. So, a country with 20% higher mortality than expected in the summer and 40% higher mortality than expected in the winter will over the course of the year have higher excess deaths that year than a country with 40% higher mortality in the summer and 20% higher mortality in the winter, even though averaging 20% and 40% gets you 30% either way. The seasonal differences in mortality rates are not huge, but the differences in weekly baselines are a key reason that if one were to average the average weekly excess mortality rate for a year, it wouldn't match the overall excess mortality rate across the year exactly."
This having been said, I thought the information on Sweden was worth noting, and of course I'd love to know whether there is more reliable information.
I should note that I have neither been a strong supporter or strong opponent of lockdowns, except insofar as I've taken the view that lockdowns are generally constitutionally permissible. Whether or not they are sound is an important empirical question, which deserves more study. And to the extent that Sweden's experiment seems to have been at least potentially successful (though contrary to the judgment of health authorities in other countries), I hope that it leads people to study the question more closely.
