Nowhere near as many people are dying in Sweden as the COVID death count would suggest. That may tell us something important about the impacts of the virus and the lockdowns.

by Will Boisvert

Throughout the COVID pandemic, the number of excess deaths in most countries has been a lot higher than the official number of deaths attributed to the coronavirus. Scientists blame that discrepancy mainly on a sizeable undercount of COVID cases among the dead.

But Sweden is a big exception to that story. In Sweden, excess deaths are running well below COVID deaths—up to 42 percent lower, depending on the baseline used to measure the excess. That anomaly may constitute evidence that many non-COVID deaths were misattributed to COVID, that the mortality impacts from COVID lockdown measures were larger than we thought, or that the impact of COVID in terms of years of life lost was lighter than we estimated.

Here’s a graph of the shortfall of excess deaths relative to COVID deaths. The gray line represents the shortfall when we subtract COVID deaths from excess deaths; the lower it goes, the bigger the shortfall.

It’s a big gap. Through December 31, 2021, Sweden’s public health agency logged 15,362 COVID deaths. But its preliminary all-cause mortality statistics indicate just 8,981 net excess deaths over the duration of the pandemic from March 1, 2020, to December 31, 2021. (The excess deaths are calculated by taking each month’s deaths in 2020-2021 and subtracting the average number of deaths in that month during the five-year period from 2015 through 2019.) That’s a shortfall of 6,381 excess deaths, or 42 percent of COVID mortality—a stark difference from the usual pattern of excess mortality running higher than the official COVID death toll.

What’s causing the anomalous shortfall? It could be misattribution of non-COVID deaths as COVID deaths. We know that this has happened in some cases, but not on the scale suggested by the Swedish shortfall. It’s also not clear why Sweden would be more likely than other countries to misattribute non-COVID deaths to COVID.

On the other hand, one could conjecture that those other countries had much higher mortality related to lockdowns, which Sweden was spared. In the United States, for example, there were substantial increases in deaths by drug overdoses, auto crashes and heart attacks in 2020-2021, all of them plausibly linked to lockdowns and other restrictions. These non-COVID excess deaths might have swamped the effects of misattribution in other countries, but not in Sweden.

A third possibility is that the shortfall is the natural consequence of the short life expectancies of the very old and sick people who typically die from the coronavirus.

Studies have estimated the average life expectancy of COVID victims at 10-12 years, but I’ve argued elsewhere that their real life expectancy is probably a lot shorter given the large proportion who are nursing-home residents with normal life expectancies of two or three years or less. Short life expectancies mean that many COVID victims would have died of something else within a few months to a few years had they not caught the virus. And when those short-timers are prematurely culled by COVID, there should be a noticeable shortfall of deaths during subsequent months. If a man who normally would have died in July is prematurely killed by COVID in April, then July’s all-cause death numbers will be lower than normal (after subtracting out any new deaths COVID itself causes in July). If COVID victims have very short life expectancies, as I’ve argued, we should see a larger shortfall, developing sooner, than if they have relatively long 10-to-12-year life expectancies as the literature suggests.

This table shows the monthly progression of the death shortfall. The numbers offer some support for the possibility that much of it is due to the prior culling of victims with short life expectancies.

The table shows that the shortfalls in excess deaths were modest during the first wave in the spring of 2020, then gradually increased through the summer and fall. The shortfalls then swelled dramatically in January through April of 2021, roughly one year after the first wave—and during the traditional flu season. This pattern of shortfalls suggests that many of those who died in the first wave of COVID had a life expectancy of about one year, and that the spring 2020 wave was prematurely culling people who would have perished anyway during the next winter’s dying season. When winter of 2021 came around, many of the frail, elderly people who would normally have died then had already been killed by COVID a year before. There was therefore a large deficit of non-COVID deaths that winter that was only partially made up by new COVID deaths.

But conflicting with my conjecture above is the pattern in the autumn of 2021 where we don’t see shortfalls in excess deaths, which we might have expected given the huge number of COVID deaths a year earlier in November 2020 through February 2021. So go figure.

There are other factors that might explain the shortfall, like a flu shortfall. Even Sweden’s lax COVID restrictions apparently eliminated the 2020-2021 flu season, so part of the death shortfall would be flu deaths that didn’t happen. Flu deaths in Sweden from 2015-16 to 2019-20 averaged about 545 per year, about 8.5 percent of the cumulative shortfall.

There could also be problems with Sweden’s use of average deaths in the period 2015-2019 as the baseline for calculating excess deaths in 2020-2021. One issue is that it doesn’t account for the fact that Sweden’s population was slightly larger and older in 2020-2021 compared to 2015-2019. Correcting for that would tend to push up the number of baseline deaths, reduce the estimate of excess deaths and swell the shortfall. A countervailing factor is that there is a pronounced downward trend in Sweden’s mortality rates over time that would have the opposite effect: lowering the baseline, raising the estimate of excess deaths and shrinking the excess-death shortfall relative to COVID deaths.

Here’s a graph showing the decline in crude Swedish mortality rates—deaths divided by population—over the decade from 2010 to 2019. I’ll use the trend line as a rough guide to what the mortality rates and baseline all-cause deaths should have been in 2020-2021. (The final two points for 2020 and 2021 are extrapolations calculated using the trend line, which was fitted to the 2010 through 2019 datapoints.)

The 2010-2019 data look like a decent fit for a downward linear trend line, which indicates a somewhat lower 2020-2021 baseline than the average of deaths in 2015-2019 that the Swedish government used. Going by the trend line, 2020 should have had a mortality rate of about 875 deaths per 100,000, which would give a baseline of about 90,592 total deaths. The 2021 trend-line mortality rate of about 865 per 100,000 would have given 90,084 deaths, for a two-year baseline total of 180,676 deaths. Actual deaths in 2020 and 2021 were a total of 190,082, for an excess of 9,406 above the baseline and a shortfall of 5,956 relative to COVID deaths. Subtracting the average flu-season deaths of 545 gives an estimate of 5,411 for the death shortfall, 35 percent of the COVID toll of 15,362. So, somewhat smaller than the shortfall I reckoned earlier, but not dramatically lower.

Okay, no more layman’s epidemiology. These are very rough estimates, with large error bars and probably just plain errors.

Still, the Swedish death shortfall is big enough that it’s hard to dismiss, and it has interesting implications. One is that the pandemic’s mortality impact on Sweden has been substantially lighter, perhaps 35 to 40 percent lighter, than the COVID death numbers indicate. That makes Sweden’s do-little approach to the pandemic look more reasonable and less death-cultish than critics have claimed. A second is that the mortality impact of the virus in general, in terms of years of life lost by its victims, may be significantly less than we thought. Most of its victims may have lost a few years or less at the very end, not a substantial chunk of their lives. A third is that we need to look harder at the mortality impacts of COVID restrictions and lockdowns: by ascribing jumps in all-cause excess deaths to COVID the virus rather than COVID the panic we may have seriously underestimated them.

Sweden is a useful datapoint because of its unusual combination of high COVID mortality with lax COVID restrictions; that produces a very visible mortality signal without much confounding from lockdowns. There aren’t many places like that, but I’ll be looking at some of them in the future to see if any patterns hold up.