Sweden

Keeping Schools Open Had Only a Minor Impact on the Spread of COVID-19 in Sweden

Sweden was one of the few Western countries that kept schools open in the spring of 2020. Pre-schools, primary schools and lower-secondary schools (for those up to age 16) continued with in-person teaching, whereas upper-secondary schools switched to online instruction on March 18th.

Despite this, zero Swedish children died of COVID-19 up to the end of June. In fact, only 15 were admitted to the ICU, and four of these children had a serious underlying health condition. 

So keeping schools open didn’t cause any deaths among Swedish children. But did it increase the spread of COVID-19? Although evidence suggests that children are less infectious than adults, their level of infectiousness is not zero. In addition, teachers could transmit the virus to one another in the staff room, and parents could do so when picking their children up from school.

In a paper published in Proceedings of the National Academy of Sciences, researchers from Stockholm and Upsala University examined the impact of keeping schools open on the spread of COVID-19 in Sweden. Their analysis focused on the period from March 25th to June 30th. 

The authors used rigorous methods to estimate the causal impact of keeping schools open on COVID-19 outcomes among parents, and among teachers. Specifically, they compared parents whose youngest child was in the last year of lower-secondary school (Year 9) to those whose youngest child was in the first year of upper-secondary school (Year 10). 

This method ensured that the two groups of parents were as similar as possible with respect to other possible causes of COVID-19 outcomes. But to be safe, the authors controlled statistically for characteristics like the age, occupation and region of the parents.

They found that there was only one additional positive PCR test per 1,000 parents among those whose youngest child was in the last year of lower-secondary school. They also looked at the number of diagnosed cases of COVID-19, but found this did not differ significantly between the two groups of parents. 

When the authors compared teachers from lower-secondary schools with those from upper-secondary schools, the differences were somewhat larger. However, the overall impact of keeping schools open on the spread of COVID-19 was small. The authors estimate that keeping schools open resulted in 620 more cases in a country that saw more than 53,000 up to mid June. 

They conclude that closing schools “is a costly measure with potential long-run detrimental effects for students”. And their results are “are in line with theoretical work indicating that school closure is not an effective way to contain SARS-CoV-2”.

Sweden’s Mortality Rate Last Year Was Lower Than in 2015

As I’ve mentioned several times, when you calculate mortality the correct way – as the age-standardised mortality rate, or as life expectancy – the year 2020 in England doesn’t look that unusual. Last year’s rate was a fair bit higher than 2019’s, but that was a year of unusually low mortality. 

Plotting the age-standardised mortality rate over time (as the ONS has been doing each month since July of 2020) shows that mortality last year rose to a level last seen in 2008. So while the year-on-year change was large, the level wasn’t particularly high – at least by historical standards. 

Interestingly, this point even found its way into a BBC article last September. The author noted:

And if you look at the age-adjusted mortality rates, which take into account the size and age of the population, you can see that while 2020 has undoubtedly been a bad year compared to recent years, what has been seen in terms of people dying is not completely out of sync with recent history. It is actually comparable with what happened in the 2000s.

Given that 2008 – which, to repeat, saw a higher level of morality than last year – wasn’t that long ago, one might argue the pandemic’s lethality has been overhyped. Of course, others would contend that, if we hadn’t taken the drastic measures we did take, mortality would have risen to a far higher level.

But I’m not convinced the UK’s lockdowns did do much to curb mortality, over and above the effect of restrictions on large gatherings and voluntary social distancing. And I’d argue that we could have saved more lives with a well-executed focused protection strategy.

However, many people continue to insist that mortality would have been far higher in the absence of lockdowns. It’s therefore worth looking once again at Sweden – the only major European country that didn’t lock down.

We already know that Sweden’s age-adjusted excess mortality up to week 51 was only 1.7% – below the European average. But when was the last time its mortality rate was as high as last year?

Going up to the end of week 52, the rate for 2020 – based on the European Standard Population – comes out as 16.4 per 100,000 (which is actually lower than in Denmark). And the last time Sweden saw this level of morality was in 2015 – just five years ago.

So despite taking the least restrictive approach of any major Western country, Sweden’s mortality rate only returned to the level of 2015. This casts doubt on the claim that mortality in the UK would have been much higher in the absence of lockdowns.

Taking the Average of 2019 and 2020, Sweden Had Lower Mortality Than Both Denmark and Finland

Faced with mounting evidence that lockdowns did not substantially reduce COVID-19 deaths in most of the countries where they were implemented, lockdown proponents have fallen back on what Paul Yowell calls the “neighbour argument” – i.e., the argument that comparing Sweden to its neighbours shows that lockdowns really do work.

On May 10th, a tweet plotting cumulative COVID-19 deaths per million in Sweden, Norway and Finland – which referred to the “Nordic natural experiment” – garnered over 6,000 likes. 

However, this argument isn’t convincing for a whole number of reasons, as I’ve outlined in two previous posts. For example: the other Nordics had a head start on Sweden; border controls – not lockdowns – made the difference in the first wave; and once you include the Baltics, Sweden no longer stands out.

However, suppose we just look at the mortality figures. Do they show that Sweden had an exceptionally bad year? Far from it. As I’ve noted before, the country saw age-adjusted excess mortality up to week 51 of just 1.7% – below the European average. 

Now, it’s true that all three other Nordics saw negative excess mortality (of up to –5% in Norway’s case). Because mortality rates declined gradually from 2015 to 2019, no change from 2019 to 2020 yields a negative value for excess mortality. In addition, there may have been fewer flu deaths and car accidents, thanks to social distancing. 

However, one reason why Sweden’s excess mortality figure isn’t lower is that the country saw particularly low mortality in 2019 (which brings down the average of the last five years). In that year, Sweden had the lowest mortality of all four Nordics – its rate was 4% lower even than Norway’s.

As several commentators have pointed out, this meant that there were more frail elderly people alive at the beginning of 2020 than there otherwise would have been. So even in the absence of a pandemic, you’d have expected to see a slight rise in mortality – owing to the “dry tinder” effect.

If we take the average of 2019 and 2020, then Sweden’s age-standardised mortality rate was 15.8 per 100,000, Denmark’s was 17.6, Finland’s was 16.4 and Norway’s was 15.5. In other words, Sweden’s was lower than both Denmark’s and Finland’s, and was only slightly higher than Norway’s. 

Of course, the average of the last two years isn’t a measure of the impact of the pandemic (and other relevant events). For that, we can need to compute the excess mortality for 2019–20, by comparing the average mortality rate in those two years to the average over the preceding four years. When we do that, the numbers come out as follows: –3.3% in Sweden, –4.4% in Demark, –4.8% in Finland and –4.9% in Norway. 

Although Sweden still saw the least favourable change (i.e., the smallest decline in mortality), the disparity with respect to its neighbours is much reduced. 

This exercise is not meant to obscure the fact that Sweden saw a moderate rise in mortality last year, unlike the other Nordics. It’s simply meant to put that rise in mortality into perspective. After all, having a sense of perspective is very important when trying to evaluate the measures that were taken during the pandemic.  

A Response to Dominic Cummings’ Pro-Lockdown Twitter Thread

Dominic Cummings – director of the Vote Leave campaign and former chief adviser to Boris Johnson – has written a pro-lockdown Twitter thread. However, I don’t find his arguments very convincing. What follows is a point-by-point response.

1/ Covid… Summary evidence on lockdowns. For UK political pundits obsessed with spreading nonsense on Sweden/lockdowns, cf. SW econ did a bit WORSE than Denmark which locked down, AND far more deaths in Sweden:

Not all sources indicate that Sweden did worse than Denmark in terms of GDP growth last year. For example, the IMF gives Sweden’s growth as –2.8% and Denmark’s as –3.3%. In fact, according to the IMF, only a handful of European countries had higher growth than Sweden last year.

It’s true that Denmark has had fewer COVID-19 deaths. However, it’s unlikely that lockdowns account for this difference. During the first wave, Denmark had zero days of mandatory stay-at-home orders, and did not introduce mandatory business closures until March 18th. But the country did introduce border screening on March 4th, followed by a total border closure on March 14th. Hence its success during the first wave is more plausibly due to border controls.

During the second wave, Denmark had about the same level of restrictions as Sweden, and in any case saw a moderate number of deaths. 

More importantly, the argument that “we have to compare Sweden to its neighbours” isn’t very convincing. Sweden’s age-adjusted excess mortality up to week 51 of 2020 was just 1.7% – below the European average. 

The epidemic in Sweden was already more advanced by the time its neighbours locked down. And since lockdowns don’t have much impact unless case numbers are low, locking down probably wouldn’t have made a big difference. What’s more, the Baltics are similar to the Nordics in terms of climate and population density, and once you include them in the comparison, Sweden no longer stands out.

Cumming’s tweet also links to an article by the economist Noah Smith, which argues that “lockdowns were good”. However, Smith doesn’t discuss any of the evidence contradicting his thesis, of which there is plenty. See herehereherehereherehere and here.

One of the biggest misunderstandings, spread by political pundits even now, is the ‘tradeoff’ argument. Fact: evidence clear that fast hard effective action best policy for economy AND for reducing deaths/suffering

The argument that lockdowns are good for both public health and economic output – that there’s no trade-off – only works if locking down enables you to completely suppress the virus. 

Once complete suppression has been achieved, the lockdown must be combined with a well-functioning system of contact tracing, and a well-functioning system of border controls. In the absence of these measures, a new epidemic will almost certainly emerge once the lockdown is lifted.

There is strong evidence that the UK’s lockdowns were bad for the economy. Indeed, the UK had the second lowest GDP growth in 2020 out of all the major countries in Europe, and its worst recession for 300 years. 

One could argue that the UK should have locked down earlier, but this is a bit like arguing China should have acted earlier to contain the epidemic in Wuhan. In other words, that ship sailed a long time ago.

What’s more, it’s doubtful whether the UK – which is much denser and more connected than, say, Australia – would have been able to contain the virus through measures like contact tracing and border controls. 

Dominic Cummings Blasts Boris for not Imitating China’s Policy of “Welding People Inside Homes” in Fact-Free Twitter Rant

Dominic Cummings has fired off his latest salvo against his former boss ahead of his appearance before MPs to give evidence on May 26th, laying into Boris Johnson and the Government for not locking down sooner, among other complaints.

The disgruntled former chief adviser to the Prime Minister wrote a series of posts on Twitter that began by criticising Sweden’s response before ranging over other issues including human challenge vaccine trials and the transparency of SAGE.

Those of us “obsessed with spreading nonsense on Sweden/lockdowns” are treated to Dom’s “summary evidence on lockdowns”. Unfortunately for him, however, he seems to get his facts from somewhere other than the real world.

Dom takes a shot at the “trade-off argument” – the argument that lockdowns intended to control disease have a lot of downsides. He argues that Taiwan shows how “fast hard effective action [is the] best policy for [the] economy AND for reducing deaths/suffering”, and that “if you REALLY get your act together not only is [the] econ[omy] largely unscathed but life is [close to] normal”. He claims the Government is “totally hostile to learning from East Asia” because they and their advisers believe “Asians all do as they’re told it won’t work here”.

It’s true that East Asian countries have suffered considerably fewer deaths during the pandemic than the countries of Europe and the Americas. But the idea that that is because they imposed lockdowns hard and fast is palpable nonsense. Japan has not imposed a strict lockdown and neither has Taiwan or South Korea (see below). Worth recalling that South Korea has more commonly been lauded for avoiding hard lockdown by being so good at contact tracing, not for being fast to lock down hard. Contact tracing is also very unlikely to be the main reason for South Korea’s epidemic remaining small, but either way there is no basis to Dom’s claim that East Asia’s success is due to hard and fast lockdowns. As for Taiwan’s current “normal”, that involves very tight border restrictions that have been in place since February 6th 2020, and the country has just imposed new restrictions on the capital region Taipei.

Did Lockdown Shift the Burden of COVID-19 Onto the Working Class?

One of the claims put forward by the authors of The Great Barrington Declaration is that lockdowns unfairly shifted the burden of COVID-19 onto the working class. As Martin Kulldorff and Sunetra Gupta argued in a piece for the Toronto Sun last November:

Low-risk college students and young professionals are protected; such as lawyers, government employees, journalists, and scientists who can work from home; while older high-risk working-class people must work, risking their lives generating the population immunity that will eventually help protect everyone.

The same idea was captured in a viral tweet by the art critic J.J. Charlesworth:

To evaluate this claim, let’s begin by looking at some of the data from Britain. Last July, the ONS attempted to quantify the extent to which different jobs can be done from home. Unsurprisingly, they found that higher-paying jobs in the professional and managerial classes are much easier to do from home, whereas lower-paying jobs in the skilled and unskilled working class are much harder to do from home. (‘Front-line doctor’ is an exception.)

While “key workers” are drawn from all income deciles, a relatively large percentage are drawn from the 2nd, 3rd and 4th deciles – particularly in the food and necessary goods sector. And according to the ONS, 15% of such workers were at an increased risk of COVID-19 because of a pre-existing health condition.

In January of 2021, the ONS computed age-standardised mortality rates for COVID-19 in different occupations. They found that men in professional and managerial occupations were substantially less likely to die of COVID-19 than those in service and elementary occupations:

The pattern among women was similar, although somewhat less pronounced. (The highest age-standardised mortality rate was for women working as plant or machine operators.)

What Second Wave? Total Deaths in UK and Sweden Now Average for 2021

New figures from the ONS released yesterday show that deaths in England and Wales are running 7.3% below the five-year average for the week ending April 30th. This is the eighth consecutive week that registered deaths have been below the five-year average.

While the UK’s winter epidemic has been over for some months now, Sweden, like much of the continent, has seen a spring wave.

ICUs have been busier in spring than they were in winter.

Border Controls, Not Lockdowns, Explain the Success of Denmark, Norway and Finland

I’ve previously explained why “we have to compare Sweden to its neighbours” isn’t a convincing argument for lockdowns. However, the argument keeps cropping up on social media. So I’ll have another go.

As I noted in my previous post, Sweden has had more deaths than the other Nordic countries – whether you use ‘confirmed COVID-19 deaths per million people’ or age-adjusted excess mortality. 

However, this doesn’t mean that lockdowns are what account for the divergent mortality trends. In other words, it doesn’t follow that if Sweden had locked down at the same time as its neighbours, then it would have seen many fewer deaths from COVID-19.

Even if you believe that lockdowns were the main factor behind the other Nordics’ low death rates (and they probably weren’t), the epidemic was already more advanced in Sweden by the time its neighbours locked down. And since lockdowns don’t have much impact unless case numbers are low (as in Australia and New Zealand), locking down probably wouldn’t have made a big difference. 

Moreover, there’s good reason to believe that lockdowns weren’t the main factor behind the other Nordic’s low death tolls. Rather, the main factor was probably border controls.

Let’s examine what each country did during the first wave, using the Oxford Blavatnik School’s COVID-19 Government Response Tracker. (I will ignore Iceland, since it’s a small island in the middle of the Atlantic ocean, and its geographic advantages are obvious.) 

Recall that the Blavatnik School’s database includes several measures of government restrictions. I will focus on mandatory workplace closures, mandatory stay-at-home orders, and restrictions on international travel (i.e., border controls). 

Let’s start with mandatory stay-at-home orders. None of the Nordics had any days of mandatory stay-at-home orders during the first wave. (This is in contrast to the U.K., which was hit much harder than all four Nordics, and had a mandatory stay-at-home order in place between March 23rd and May 12th.)

Now mandatory workplace closures. Norway did introduce these quite early on March 12th. However, Denmark did not introduce them until March 18th – just five days before the U.K. And Finland did not introduce them until April 14th – more than three weeks after the U.K.

These comparisons reveal that the other Nordics did not lock down particularly hard or particularly early. Indeed, all three had less strict lockdowns than the U.K. (which saw many more deaths during the first wave). Finland’s success is particularly difficult to explain with reference to lockdowns since the country did not introduce any real measures until after the peak of infections.

Imperial College’s Modelling is Even Worse Than We Thought

When Professor Neil Ferguson and his team at Imperial College London have been challenged on their model’s miserable failure to predict the pandemic death toll in Sweden they have always pushed back saying they didn’t model Sweden, disavowing the work of the team at Uppsala University which adapted their modelling to the Swedish context. But it turns out this is not exactly accurate. Phillip W. Magness explains on AIER:

In the House of Lords hearing from last year, Conservative member Viscount Ridley grilled Ferguson over the Swedish adaptation of his model: “Uppsala University took the Imperial College model – or one of them – and adapted it to Sweden and forecasted deaths in Sweden of over 90,000 by the end of May if there was no lockdown and 40,000 if a full lockdown was enforced.” With such extreme disparities between the projections and reality, how could the Imperial team continue to guide policy through their modelling?

Ferguson snapped back, disavowing any connection to the Swedish results: “First of all, they did not use our model. They developed a model of their own. We had no role in parameterising it. Generally, the key aspect of modelling is how well you parameterise it against the available data. But to be absolutely clear they did not use our model, they didn’t adapt our model.”

The Imperial College modeller offered no evidence that the Uppsala team had erred in their application of his approach. The since-published version from the Uppsala team makes it absolutely clear that they constructed the Swedish adaptation directly from Imperial’s UK model. “We used an individual agent-based model based on the framework published by Ferguson and co-workers that we have reimplemented” for Sweden, the authors explain. They also acknowledged that their modelled projections far exceeded observed outcomes, although they attribute the differences somewhat questionably to voluntary behavioural changes rather than a fault in the model design.

Ferguson’s team has nonetheless aggressively attempted to dissociate itself from the Uppsala adaptation of their work. After the UK Spectator called attention to the Swedish results last spring, Imperial College tweeted out that “Professor Ferguson and the Imperial COVID-19 response team never estimated 40,000 or 100,000 Swedish deaths. Imperial’s work is being conflated with that of an entirely separate group of researchers.” It’s a deflection that Ferguson and his defenders have repeated many times since.

In fact, though, as Phillip points out, it is not true to say that the Imperial team never estimated 40,000 or 100,000 Swedish deaths. Hidden away in a spreadsheet in the appendix to Report 12, published on March 26th 2020, are the team’s estimates for other countries including Sweden. The projections are expressly intended to encourage those countries to follow suit with social restrictions. They write:

To help inform country strategies in the coming weeks, we provide here summary statistics of the potential impact of mitigation and suppression strategies in all countries across the world. These illustrate the need to act early, and the impact that failure to do so is likely to have on local health systems.

The predictions for Sweden are up to 90,157 deaths under “unmitigated” spread (Uppsala projected 96,000) and, under “population-level social distancing” (lockdowns), 42,473 deaths (compared to Uppsala’s 40,000). So, contrary to their repeated denials, Ferguson’s team did make predictions for Sweden very close to those made by the Uppsala team who adapted their model, and those predictions were just as way off. Sweden’s Covid death toll at the end of the first wave, on August 31st, was 5,821.

Phillip summarises further failures of the Imperial modelling in a table showing four non-lockdown countries (Sweden, Taiwan, South Korea, Japan) and the United States (most of whose states imposed a lockdown in the spring) with their one-year death toll and how it compares to Imperial’s projections.

Performance of Imperial College Modelling in Four Non-Lockdown Countries and the United States (AIER)

It’s worth saying, though, that the models for the ‘unmitigated’ scenarios predicted the deaths to occur over the course of a few months, not a whole year including another winter flu season. There will be another ‘wave’ of deaths every winter, possibly from (or with) COVID-19 if it remains the dominant respiratory virus (and if we keep on testing for it). If we keep on adding the deaths over several seasons then of course they will eventually reach the predicted figures. But that wasn’t what the models were claiming to show and would be a case of making the evidence fit the model.

The AIER article is worth reading in full.

“We Have to Compare Sweden to Its Neighbours” Isn’t a Convincing Argument

In a recent post on Lockdown Sceptics, I argued that the case for lockdown basically collapsed in May of 2020, when Sweden’s epidemic began to retreat. Sweden was the only major Western country that didn’t lockdown in 2020, yet it saw age-adjusted excess mortality up to week 51 of just 1.7% – below the European average.

A common reply is that, although Sweden did better than the European average, it did worse than its neighbours. Here its neighbours are taken to be the other Nordic countries: Denmark, Norway, Finland and Iceland. Looking at age-adjusted excess mortality, it’s true that the other Nordics did better than Sweden. All four saw negative excess mortality up to week 51.

Does this mean lockdown sceptics are wrong to cite Sweden as evidence that the benefits of lockdowns are vastly overstated? No, I don’t believe it does.

First, the economist Daniel Klein and his colleagues have identified 15 different factors that may account for the higher death toll in Sweden as compared to the other Nordics. These include the greater number of frail elderly people alive at the start of 2020 (the ‘dry tinder’ effect); the larger immigrant population; and the lack of adequate protection for care home residents in the early weeks of the pandemic. 

Second, as the researcher Philippe Lemoine has pointed out, the epidemic was already more advanced in Sweden by the time most European countries introduced lockdowns and social distancing. The other Nordic countries therefore had a head start in responding to the deadly first wave. This is particularly important because, when the first wave struck, the best ways of treating COVID-19 were not yet well understood.

I would add that, with the exception of Denmark (which saw a moderate second wave), the other Nordics are small, geographically peripheral countries for which a containment strategy was actually workable. As I’ve noted in Quillette, all the Western countries that have managed to keep their COVID-19 death rates low (Norway, Cyprus, Australia, etc.) benefited from pre-existing geographical advantages. And all imposed strict border controls at the start (something the UK Government’s scientific advisers cautioned against).

Third, as the legal scholar Paul Yowell has argued, the Baltics (Estonia, Latvia and Lithuania) are similar to the Nordics in terms of climate and population density, and once you include them in the comparison, Sweden no longer stands out. Lithuania actually had higher age-adjusted excess mortality than Sweden last year, despite imposing a strict winter lockdown.

Finally, as Yowell also points out, the ratio of Sweden’s COVID-19 death rate to Denmark’s isn’t that much higher than the ratio of Denmark’s to Finland’s. And this is despite the fact that Denmark has taken a more restrictive approach than Finland. One could therefore take the comparison between those two countries as evidence against the efficacy of lockdowns.

What’s more, this exercise could be repeated with other pairs or trios. For example, despite taking a slightly less restrictive approach than Spain and Italy, France has reported fewer deaths from COVID-19 (as well as lower excess mortality). Of course, these kinds of comparisons don’t tell us very much. But that’s the point. We shouldn’t only compare a country to its immediate neighbours.

And when researchers have analysed European countries and US states in a systematic way, they haven’t found evidence that lockdowns substantially reduce deaths from COVID-19.