Is Sweden’s Covid Strategy Working?

The answer: only if life there is very cheap.

Credit: Mando Gomez

Compared to its Nordic neighbors as well as to other wealthy countries, Sweden is a bit of an oddball when it comes to the coronavirus. For various reasons, Sweden relied far more heavily on voluntary restrictions than did most other “advanced” countries to manage their response to the Covid-19 pandemic.

The result? More dead Swedes, almost certainly. But — at least from a policy perspective — the two biggest questions are, 1) how many more people died because of their approach, and 2) what did Sweden get in exchange for this increased mortality?

A rough estimate of Sweden’s increased mortality

The following chart shows how the total number of Covid-19 deaths per capita have grown over time for Sweden, Norway, and Denmark.¹

As of this writing, Sweden has had about 1,500 confirmed Covid-19 deaths per million people, Denmark about 500 deaths per million, and Norway almost 200 per million people. If we assume that these three countries are identical other than their approaches to managing the pandemic (admittedly, they’re not), then Sweden’s policies so far have led to at least an additional 1,000 deaths per million people (i.e., 1,500 minus 500 = 1,000).

The population of Sweden is about 10 million, so the absolute number of excess Covid-19 deaths attributable to Sweden’s looser policies is about 10,000.²

The economic benefits of Sweden’s approach

If Sweden only had a bunch of excess deaths and no associated benefits, we’d probably all agree that it’s not such a compelling policy. In fact, earlier on in the pandemic it appeared as though Sweden’s economy wasn’t faring much better than that of its Nordic neighbors.

But a more recent analysis suggests that Sweden’s economy has done slightly better than the economies of Denmark and Norway. Specifically, economists estimate that Sweden’s GDP growth is one-quarter to three-quarters of a percent better than it would have been had it implemented stricter policies to manage the pandemic.³

We can use this estimate to get a sense of the likely economic benefit associated with Sweden’s looser approach. Sweden’s GDP (in US dollars, assuming purchasing power parity) is about $55,000 per person.⁴ Given a population of 10 million, Sweden’s GDP is therefore about $550 billion. If we assume that Sweden’s policies reduced the impact of the pandemic on their GDP by one-half of one percent (i.e., the midpoint of the estimates), Sweden prevented about $2.8 billion in losses (i.e., one-half of a percent of their $550 billion GDP).

A key question: how much is a human life worth?

Sweden lost more lives but they did better in economic terms. For policy makers, the question is whether that was a good tradeoff.

To answer this question, analysts use something called the value of a statistical life. The “statistical” part means that no one really knows which specific lives were lost or gained as the result of a policy intervention; those details are left up to chance. An example might help to show how this works.

Let’s say that the value of a statistical life (or VSL) is set to $5 million, and we are looking at an intervention that will save lives but incur economic losses. If we can gain additional lives for, say, $1 million each, the policy is cost effective: the cost of saving is life is a bargain relative to the $5 million value of a statistical life. On the other hand, if the same policy costs $20 million per each life gained, analysis will conclude that it’s too expensive relative to the benefits.

This makes me feel yucky

At this point, some of you may be thinking that this whole exercise is distasteful because human lives are of infinite value. This point of view resonates with a lot of folks, me included.

The problem is that if our governmental decision makers pursue policies that assume that lives have infinite value, we will soon run out of money to pay for them all. In other words, it’s fine to say that we hold human life as infinitely valuable. That sentiment isn’t the problem. The trouble arises because societies don’t have infinite economic resources to match. Setting a numeric bar for the value of a statistical life is one element of making sure that we’re saving as many lives as possible given a fixed set of resources.

So… did the Swedish approach work?

Now, let’s get back to Sweden’s policy. Here’s a way to think about it. Imagine that you are a policy maker in Sweden. You can take the Danish approach and implement stricter restrictions in the hopes of preventing deaths, or you can adopt a looser management strategy.

The tighter approach will save about 10,000 lives. Sadly, there’s no free lunch; implementing these tighter restrictions will trim about one-half of a percentage point on GDP growth, costing you about $3 billion. This means a Danish-like approach will cost about $300,000 per statistical life saved.

The official Swedish stance is to use $3 million as the value of a statistical life for policy making.⁵ This means that policies that save lives at a cost of less that $3 million are a good deal. And that means that the Swedish policy isn’t working. By adopting a stricter approach, Sweden probably could have saved lives for about 10 cents on the dollar.

And if you’re an American policy maker, the Swedish approach looks even worse. That’s because officials there use a VSL of about $10 million.⁶ In other words, from an American perspective, the Swedish policy looks even more out of whack.

What would have to be true for the Swedish approach to be working?

To achieve the target of $3 million per life, the GDP savings from Sweden’s policies must be 10 times greater than currently estimated, or the number of lives lost to Covid-19 from the Swedish policy must be ten times less than currently estimated, or some combination of the two. None of these things seem particularly likely, but admittedly this is a pretty rough analysis.

If we assume that Sweden’s Covid-19 policies generated 0.75% savings to GDP instead of 0.5%, the absolute savings to GDP would be roughly $4.5 billion (instead of $3 billion). The implied value of a statistical life then rises from $300,000 to $450,000. That’s an increase for sure, but one that still falls far short of the $3 million benchmark set forth by Swedish policymakers.

What about those excess deaths? Might they reflect something other than just Sweden’s looser approach to managing Covid-19? That is absolutely possible. For example, Sweden is a bit older than Denmark, and we know that (all other things being equal), as age goes up so does Covid-19 related mortality. That means that some of the difference in the per capita deaths between the two countries isn’t due to Sweden’s more lax approach to managing the pandemic.

Let’s noodle this through. About 20.6% of Sweden’s population is over the age of 65; for Denmark the comparable figure is 19.9%.⁷ If we are super aggressive and assume that all Covid-19 confirmed deaths occurred among people over the age of 65, we can get a sense of how important this age difference is.

The difference in the proportion of the population over 65 is 0.7% (20.6% for Sweden minus 19.9% for Denmark). Not surprisingly, this 0.7% difference is only alters our estimate slightly. We would expect that only 3% (0.7% divided by 19.9%) of Sweden’s excess deaths are related to age demographic differences with Denmark. In other words, the vast majority (97%) of the difference in deaths per capita has nothing to do with age differences.

If we’re off by a factor of five in terms of estimating Sweden’s excess deaths per capita attributable to their looser policies, the implied value of a statistical life rises from $300,000 to $1.5 million. Closer to $3 million, but still far short of the target $3 million needed for the policy to make sense.

It’s possible that the effect of Sweden’s policies on GDP will last for more than one year; assuming that the benefit persists for 5 years, the implied value of a statistical life rises roughly fivefold, from $300,000 to $1.5 million… still short of the target.

It’s possible to get to the $3 million per life threshold, but only by altering multiple assumptions at the same time. For example, if the savings to GDP from the looser Covid-19 policy are 50% greater (i.e., 0.75% instead of 0.5%), and if this advantage persists for five years, the target of $3 million per life is achieved, but just barely.

Caveats

There are potential limitations to this back-of-the-envelope estimate, so I’d like to mention some of them before y’all have a chance to do so.

• Potential limitation 1: There could be an enduring benefit to the Swedish economy. I agree in theory, and if this is the case, that one-half percent bump in GDP would apply for more than one year. The analysis cited earlier looks at differences in GDP growth through the end of 2020. I looked at figures from the OECD from the fourth quarter of 2019 (pre-pandemic) through the third quarter of 2021 (the latest available figures). Denmark was actually doing slightly better than Sweden, which raises some doubts about a longer-lasting economic advantage for the Swedish approach. Anyway, enduring effects would have to magnify the benefits by a factor of ten to get close to the Swedish approach cost effective. There is little evidence that this is the case.
• Potential limitation 2: The benefit to the economy could be greater than one-half of a percent. Again, I agree, but the true benefit has to be something like 5% (instead of one-half of a percent) for the conclusions to change. This would seem to be an impossibly large figure.
• Potential limitation 3: The difference in per capita deaths between Sweden and Denmark might have little to do with how the former managed the pandemic. I agree that this is a possibility. However, the implication is that the incremental management strategies implemented by the Danes were mostly ineffective. This seems unlikely.
• Potential limitation 4: There are other social benefits that aren’t captured by effects on the GDP. For example, Swedes probably derived pleasure from avoiding some of the harsher interventions that were imposed on the Danes, and this enjoyment is probably not completely captured by the change in GDP. Okay, fine, but these things have to be about 9 times greater in value than the direct effect on the GDP, and to me this seems like a pretty tall order.
• Potential limitation 5: Although the Swedish policy in itself might not be cost effective, it prevents the government from sliding down a slippery slope in which they gain too much control over individual behavior. Okay, maybe that’s true and maybe it’s not. One problem is that anyone can use this argument almost any time to critique a strategy that they don’t like.
• Potential limitation 6: This whole approach ignores the violation of fundamental individual rights attached to policies such as those employed by the Danes. I agree that this framework focuses only on the ends and ignores the means. That said, if we want policies that actually work, then we have to pay attention to the actual results.

There are also probably other warts of which I haven’t thought. If that’s the case, please leave me some feedback.

The takeaway

This quick (and dirty) estimate strongly suggests that Sweden’s policies for managing Covid-19 undershoot the target significantly. Specifically, the policy has led to more deaths per capita than comparator countries, but has not saved enough money to meet the government’s own targets for the value of a statistical life.

1. Our World in Data, downloaded 28 Nov 2021
2. Google is really fast at finding this kind of information.
3. Here’s their analysis. Note that despite their looser policies, Sweden’s GDP was about 4% or 5% lower than expected due to the pandemic. This suggests that the overwhelming effect on their economy was due to the pandemic regardless of how it was managed.
4. Data from the World Bank, downloaded 29 Nov 2021
5. In 2012, the Swedes recommended using 2.4 million euros as the value of a statistical life. This translates to about $3 million in todays US dollars.
6. An economic analysis published in 2019 estimates that in the US, the VSL is about $10 million.
7. Here are some data on age demographics in Sweden; similar data for Denmark can be found here.