An outsider’s perspective on the Swedish Experiment

A colleague (and friend) of mine from Deloitte Sweden asked me for my opinion of how Sweden is approaching the COVID-19 pandemic. He knows I am not an epidemiologist, but does respect my work as someone who relies on data and evidence when looking at topics. He asked me a series of questions (in bold), which I answered. He reviewed and commented, and I had further responses (those second comments from my colleague and I are in square brackets.)

All of this is obviously my personal views, not official research.

Before I begin with his questions, I will add that I have seen at least a few articles that were unfair to Sweden and their approach. I have spent over 100 days in the country over the last decade, doing research and meeting with literally thousands of Swedes face to face. The idea that their COVID response is careless, is ignoring the value of human life and has too high a regard for profits and money is simply not true. As a society, they are largely non-impulsive, thoughtful, and actually motivated by money LESS than most people around the world.

Their approach to the coronavirus may (in hindsight) turn out to have been “wrong” (we will get to what that means later) but it was not stupid, foolhardy or reckless. In order to have a reasonable debate about what works and what doesn’t we need to not “blame Sweden.”

Sweden is often referred to as an experiment in the age of COVID-19, but what country is not experimenting in the sense that no one knows what the right action is?

There are two things going on.

The first is that different countries saw COVID-19 hit 100 deaths at different dates. Italy passed 100 deaths on March 4, Spain on March 13, France on March 15 while Sweden saw that happen on March 27, or over three weeks later than Italy. Since I am Canadian, I looked up when we passed that threshold too: March 31.

Italy was flying almost blind: the only data they had on the disease was from Asia (mainly China) and people in Europe were unsure how accurate that data was and if what happened in Asia would happen the same way in Europe. Spain and France had a little more data and Sweden more still. The point is that the state of knowledge/science/best practises changes and evolves over time. So public health response to the pandemic was different in Italy and Sweden, and some of that was due to different timing and knowledge levels. That is NOT “experimenting”.

The second factor is definitely experimenting. When two countries hit the 100 death level (or any other key benchmark) around the same date, they get to choose their response. Even two countries (Spain and France) who went into severe lockdown at the same time had subtly different solutions: Spain didn’t let people out for exercise, while France did (with a note, and only for a 1 km radius). Across Europe, Germans could still go to bike stores, Belgians could buy fries and books, the French wine, Italian print newspapers, and so on.

Those are all experiments: nobody really knows how opening or closing any of those really changes the spread of infection. It is also critical to note that these aren’t like science experiments: you can’t allow people to buy frites in Belgium and see what happens, then go back and do it again, only this time with friteries closed!

That said, Sweden made policy choices that were not slightly different from its neighbours, but materially so. Not closing all schools, stores, bars and restaurants was significantly different from most other countries in Western and Northern Europe, even those deciding about lockdown at the end of March, rather than the beginning or middle of the month.

[We did close schools, schools for student over 15 years old, but why only them? Because generally those student requires travelling on public transport. And why did we let the younger ones stay? Because then nurses/doctors/caretakers would have to stay home.

That’s true…but every other country also has a population of nurses/doctors/caregivers some of whom have children under 15. And almost all of them decided that closing all schools was the right thing to do, and that the medical system would be able to figure out the childcare issue for medical professionals. Nobody knows just how important closing all schools (including for those under 15) is in terms of viral spread, but Sweden was clearly a significant outlier.

Another thing to note it that Sweden may not have closed all schools, but they did close some. They did not force people to work from home, but did recommend it and about half of Swedes are doing WFH. Gatherings of more than 50 were prohibited. The government asked Swedes not to travel, and internal travel fell by 90%. Even though restaurants and bars were not forced to close, I have talked to friends in Sweden who say that many Swedes are choosing not to go out as much, and some bars and restaurants are closing…not because of a forced lockdown, but because business is too slow. No tourists either, which doesn’t help.

A key point is not that Sweden is experimenting with “everything is normal.” Their experiment is in going for “lockdown light” with suggestions rather than a bigger one mandated by the government.]

Everybody has been experimenting to some extent…but the Swedish Experiment was bigger.

I am often surprised by the amount of criticism we get but we are only compared to our Nordic neighbors. Is that the right comparison?

First off, I would say that about half the articles I read about Sweden are critical…but the other half are positive. It skews along partisan lines: Media in the English speaking world that leans right (and normally hates Sweden) tends to talk about how your approach is better while those on the left (who normally adore Sweden) are harshly critical. If you tend to read more from the Guardian and the NY Times, you will likely feel under attack. But the Daily Mail and Fox News are currently in a Sweden love-fest. Don’t worry – it won’t last.

Comparing between countries is always tough. The infection begins in different countries at different times, and with different levels of knowledge about how it spreads, etc. There are important differences in medical systems, long-term care systems (old age homes), and demographics (age, healthiness, underlying medical conditions, race and poverty.) Many other variables too!

One of my jobs is to do research on different countries. I generally find that trying to compare Sweden with France or Turkey or Singapore is not terribly helpful: there are so many differences that comparisons are often not useful. But, over time, across multiple topics (financial attitudes, media use, women in tech) it has been useful to compare Sweden with Denmark, Norway, Finland and (surprisingly) Canada. It always important to use per capita stats: Sweden is roughly twice as populous as each of its three neighbours (although Sweden doesn’t share a land border with Denmark, the ?resund/?resund bridge has about 25 million people crossing per year) and Canada is about 3.7 times bigger than Sweden.

Weather may also matter a lot. For one unsurprising reason, and for another (possibly) surprising one.

First, viruses spread faster or slower, either through the air or by contact, depending on things like temperature and humidity. It isn’t a simple thing like hot=safe, or we wouldn’t see spread all over the world, including Singapore: there seems to be a sweet spot where spread is fastest, and hotter/colder and wetter/drier than that sweet spot sees slower spreads.

If you glance at the countries in Europe with more than 300 deaths per million, you see a pattern: Italy and Spain first, then France, then Belgium and Netherlands, then UK and Ireland. All with temperate maritime climates in March, and the virus spreading faster further north as the month advances and weather warms locally. But (at least in March) the weather in non-maritime Central Europe and the Nordics (and Canada) may have been too cold for fastest spread. (As a quick sanity check, the death rate in England is 40% higher than in Scotland.)

The second factor is cottages/cabins/country homes/etc. We have them in Canada and the Nordics, just as they do in France, Spain and Italy. And when lockdowns hit in March in those three countries, hundreds of thousands (possibly millions) fled cities and decided to lockdown in the countryside where they had more room: this has been widely established as a serious contributor to the spread. Although some Canadians and Nordic city-dwellers went to their cottages on March 17…most didn’t because of the cold, snow, ice, possible loss of power, and general bad weather. I was in France the day the lockdown went into effect, and it was 17C and glorious. That same day in Toronto it was 2C.

Another reason why some countries may not be good comparisons for studying the effect of the Swedish Experiment is travel. I don’t have recent data, but according to a 2008 study, annual international visitors per capita: Spain 1416/1000, France 1277/1000, Denmark 821/1000, Italy 735/1000, Belgium 689/1000, Finland 683/1000. UK 495/1000 and Sweden 380/1000. These are annual averages: at a guess the Nordic numbers in March would all be much lower than the others, so making them less good comparisons.

There are important differences between the four Nordic countries, but in many ways (demographics, medical system, weather, etc.) they are good comparisons.

[The government in Sweden did not have the legal means to decide and enforce a lockdown.

I know nothing about Swedish law and communicable diseases, so perhaps this is possible. But I do know that around the world governments have always had sweeping powers around war, insurrection, natural disasters and pandemics. And that during this crisis, no other country seemed to have legal problems in enforcing the first lockdown. (There have been legal challenges around coming out of lockdown, but not the initial closures.) I find it remarkable (and unlikely) that only Sweden lacked the power to do this. I am going to be blunt, and say this sounds like a pretext: Sweden didn’t want to do a lockdown, so didn’t, and the lack of powers angle was a post-hoc justification.]

Canada is less good, but not bad. All five countries measure cases and deaths slightly differently, but largely those differences are minor. The chart below is from a Canadian national news network. It shows COVID-19 deaths per 100K population, on a rolling seven day basis. Fascinatingly, this is NOT a custom chart I created for this answer. It is the second chart they show on the website. (My point being that the “let’s compare the four Nordics and Canada” is not some weird and wild thing to do…it is a standard chart.) As a note, the chart shows since the 100th case, not 100th death.

A lot of the lock down countries have done much worse.

Yes, they have. But (see top of article) they hit 100 deaths much earlier than Sweden, and have very significant differences in demographics, long term care, and so on. In my view, the fairer comparison is the Nordics plus Canada.

[What’s the difference between the lock down in Norway and in Italy apart from the fact that Norway had more data to rely on?

As far as I can tell, how early a country went into lockdown is in fact the biggest difference.

Italy went into national lockdown on March 9, when they had 463 deaths and 9,172 confirmed cases. Spain was March 14 with 196 deaths and 6,391 cases. France staggered lockdown, but March 14 is about right, at which time they had 91 deaths and 3,661 cases. UK went into lockdown on March 23, with deaths at 359 and 6,650 cases.

In sharp contrast, Norway went into national lockdown on March 12, when they had 1 death and 798 confirmed cases. Denmark went on March 11, when they had 0 (zero) deaths and 514 cases. Finland locked down March 16, with 0 (zero) deaths and 278 cases. Canada went into lockdown March 12-18 (it varied a bit by province) and on March 17 had 8 deaths and 598 cases.

For reference, on March 11 (the day Denmark locked down), Sweden had 1 death and 500 cases.

The difference between Italy/Spain/France/UK and Norway/Denmark/Finland is the difference between someone who hits the brakes right away when they see the tail lights of the car in front of them go off…and someone who waits a second or two to see what happens. I think that’s a fair analogy: as drivers there are times when we hit the brakes right away, and times when we wait a second or two. Neither behaviour is “right” or “wrong”….except in hindsight.

Virtually everybody agrees that the reasons the Italian/Spanish/French/UK lockdowns didn’t work better was that they left them too just a little too late. Even a week or two earlier might have made a big difference.]

Sure Norway and Finland have less deaths but time will tell if it stays that way when time/science catches up and statistics become a trustworthy source that in all fairness is not the case right now.

I don’t agree. First, although the Swedish statistics collection and reporting agencies (for health, but other stuff too) is excellent…the same is true of Denmark, Norway, Finland and Canada. None is perfect, but globally they are all seen as high quality non-politicised agencies. I have seen no evidence that Sweden is over-reporting COVID-19 deaths, or that the others are under-reporting them.

If anything, it looks like Sweden is undercounting very slightly. From the New York Times analysis of excess deaths, Sweden has reported 2,996 COVID-19 deaths between March 16-May 3, while the number of excess deaths is 3,400…for a slight undercount of about 500. (To be clear, that’s not a big deal: the UK undercount is 16,000, Italy is 11,000 and France is 4,900.) But the Times did do analysis for Denmark, Norway and Finland, and in each case the number of COVID-19 deaths are LOWER than excess mortality. They might not be over counting (there are factors which are lowering deaths from other causes) but there is no way they are under-reporting C-19 deaths. Putting it together, I want to show the chart below, which shows total deaths so far per 100K population for the five countries.

[You answer my statistics question but not what will be the right answer when Covid-19 has gone through all its rounds.

That is true. We don’t know what will happen over the next 1-2 years. There will almost certainly be further waves. Based in the 1918 pandemic, there were three waves, and there is no reason to think this will be any different. That said, I think it is possible to discuss how the various strategies are going so far.]

What’s your thoughts on this? You always refer to data I know but still, we are so early into this so acting based on historic experience from previous pandemics, how can that be wrong?

1.  We are early: the pandemic will likely last in some form for at least another 12-18 months, and Sweden will only hit three months since its 100th case (not death) around June 7.

2.  That said, we are nearly three months into this, which is enough time to draw some conclusions.

3.  Based on the chart above, the Swedish response will result in deaths per 100K population about 2.3x higher than Canada, quadruple Denmark and 10x higher than Norway. If it was just one of those countries, I might ignore it as random or some reporting/measurement issue. But not when I compare with four other countries that are decent comparables.

4.  To switch from per capita numbers to raw, Sweden has 4,266 reported deaths as of today, May 28. (The real death totals are likely about 4,700, per the NY Times under-reporting estimate.)

5.  Had Sweden followed some version of the Nordic or Canadian lockdown, my almost certain conclusion is that number would be as low as 500, or as high as 2,000, with a midpoint of 1,250. There have been about 3,500 more dead Swedes as a result of the "lockdown light" policy. (Yes, two thirds of all deaths in Sweden were those over 80 years old. That still means over a thousand of the additional deaths were under 80.)

 [Totally agree but I wonder whether totals will be so different once the pandemic is over?

 We don’t know that yet, of course. It is possible. But in order for the gap to close, you have to assume that Sweden is gaining something as a result of not doing the full lockdown. It could be herd immunity, but the data doesn’t support that, as discussed below.]

 In my view, we can only have a rational conversation if we start with the statement:

“Sweden did "lockdown light", and that policy resulted in more deaths thus far than if they had chosen a more severe lockdown policy on March 11. Roughly 3,500 more deaths, at least so far.”

NOW the conversation can start, and it gets interesting (in my view.) There are those (especially in the media or political opponents) who will say “You can’t put a price on a human life” or “one extra death is too many.”

This is nonsense. 300 Swedes died on the roads in 2018. I could get that number to zero by banning motor vehicles, or setting the speed limit to 20 km/h. But that would cost too much. Not a lot, but a few kids die in swimming pools every year: let’s ban pools! We could eliminate deaths from fires by building only out of non-combustible materials…but we don’t, because the cost per death avoided is too high.

We ALWAYS, in ALL countries, put a price on human life or death avoided. We always have, and we always will. The question that must be asked, in Sweden and elsewhere, is “are the benefits and costs of locking down or not locking down worth it?”

If the Swedish Experiment had generated US$3 billion more in economic activity, then the policy choice would have had a $1 million benefit for every additional life lost. One might think that was a good choice or a bad choice, but we could have a debate about it. If it were $10 per life lost, then almost no one would think it was worthwhile, and if it were $1B per life, then almost everyone (except those who had lost relatives and friends) would think the choice was worthwhile.

Sadly, it appears that although Sweden did not do a hard lockdown, its GDP is likely to fall at about the same rate as countries that did. This, to be honest, surprised me…I would have expected some benefit, even if small. Therefore, it appears that the main benefit of keeping open was not economic, but only less restrictions on things like shopping, bars, restaurants and so on. Only Swedes can answer if that – to them – was worth the 3000 extra deaths.

[I would have a look at unemployment rates as well as getting people back to work after they have lost their jobs takes time and costs massive amounts of money. So it’s not just about GDP but also keeping jobs for when things start to turn around.

That’s a great point. Although Swedish GDP is declining in line with countries that have done harder lockdowns, there are other economic indicators. In contrast to other countries that have seen big jumps in unemployment, Swedish unemployment rose from 7.1% to 8.2%. Retail sales in Sweden were down 1.3% in April…which is (high) streets better than in the UK, which was down 18%. I am sure there are other benefits: likely better mental health and lower anxiety, etc.]

Some people have speculated that as a result of a lighter lockdown, Sweden is now closer to herd immunity, and will suffer fewer deaths if there are further waves of infections and deaths. There are several problems with that logic. We don’t know for sure if those who recover are immune, although that now looks likely. We don’t know how long that immunity (if any) lasts, so those who were infected in March 2020 might be susceptible again next year or the year after.

Most importantly, herd immunity is generally thought to kick in once 70-80% of the population has had the disease and recovered. Sadly, studies suggest that only 7.3% of those in Stockholm have had the virus as at the end of April. It is almost certainly higher than that now...but with the CDC warning that positive serology tests of previous exposure may be too high by as much as 50% due to false positives, it is clear that despite following a “lockdown light” experiment, Sweden is still far below levels needed for herd immunity.

There would need to be multiple further waves of infections for herd immunity to be reached. And presumably many more deaths.

On troubling thing, that I have noticed in the last week, is a chart from Our World In Data that shows the number of daily deaths per million people. It obviously bounces around day to day, but on a consistent basis Sweden is one of the worst performing countries in the world. As of May 28, Sweden had 9.41 deaths per million in a day, compared to 0.34 in Denmark, 0.18 in Finland and 0 (zero) in Norway. Even compared to hard hit countries in Western Europe (Spain 0.02, France 1.01, Italy 1.94 and UK 6.07) Sweden is doing worse…even worse than Brazil and Peru, which are still heading towards a peak.

I think "lockdown light"…so far…is not working well. To be clear, it is not a disaster. We are talking about a few thousand deaths in a country of over ten million people. But at the end of the day, there is a difference between 4,200 dead and 1,000 dead. At least so far.

[We will only know when this thing is under control, it’s not under control anywhere because we don’t know whether it will return and in what shape and form. We have the lock down measures still at our disposal, most other don’t as you cannot resort to it more than once.

Wow. What a completely fascinating argument! As far as I know, this notion (that lockdowns can only be used once) was not originally used as a justification for the Swedish Experiment, but correct me if I am wrong.

At one point, the opposite was assumed: the idea was we would do a hard lockdown now, reopen after a couple of months, then need to lockdown again in the fall (October/November time frame), again in spring 2021 and maybe even again in fall of 2021. Until we got a vaccine we’re going to need to do this more than once, right?

So far, we have seen a second lockdown (called a circuit breaker) in Singapore as they dealt with a second wave. South Korea is seeing a second wave sort-of lockdown response in Seoul. Could we see one in other countries?

I don’t know about Europe, but in North America (to some extent Canada, but especially the US) I cannot see a second wave of lockdowns of equal severity (eight weeks, closure of all public spaces except food and pharma, schools closed, etc.) succeeding. It might work in a few states or provinces, but not nearly as many as we saw in the last two months. In some (many?) states an attempt to do so could lead to civil insurrection, etc.

Most handguns have 5-6 bullets or more. The original Derringer pistol was a single shot. Having only one bullet changes strategy.

In 1918, the first wave of the influenza pandemic was small in terms of deaths, the third wave was worse, but the second wave was awful: most of the deaths in the US were from Wave 2.

What if history repeats itself, and those countries that have used lockdown in Wave 1, find that the political consensus to support another lockdown is gone? What if they have used up the only bullet they had?

In that case, and if they are able to pull the trigger on a lockdown at that time, the Swedish Experiment might turn out to be a success…with fewer overall deaths across the multiple waves.

For those of us in North America (and perhaps the UK, etc.) we might want to start thinking about what we will do if cases and deaths spike up again in the fall…and lockdowns are not a possible intervention. If that happens, we will end up finding out just what the effects of doing the Swedish Experiment ourselves are.]