Sweden has been the world’s Covid-19 outlier, pursuing social distancing but rejecting mandatory lockdown. Schools, bars and restaurants are open – albeit with strong voluntary social distancing compliance and streets that often look almost as empty as Britain’s. Has this been enough? Sweden’s public health agency has now published a study of its R number, a metric which the UK is using to judge the success of the lockdown. The UK objective is to push R below one, by which it means it wants the number of new cases to fall. Last week, the UK’s R number was estimated at 0.8 (± 0.2 points), a figure described as an achievement of lockdown. But Sweden’s reading is 0.85, with a smaller error margin of ±0.02pts.
This raises an interesting question: might voluntary lockdowns work just as well? And might they keep the virus at a manageable level with lower social and economic costs?
The UK government has used modelling from Imperial College London, which makes some clear assumptions about lockdown. Imperial’s graph, below, shows its argument: shielding, voluntary social distancing, even school closures are shown to make very little difference to the spread of the virus (ie, the R number). But lockdown, by contrast, is shown to be a game-changer with “the R” sinking immediately. This graph below, if taken at face value, makes an open-and-shut case for lockdown.
But is it true? We don’t know because “the R” is notoriously difficult to pin down and not published in Britain. But Imperial also applied its UK assumptions to Sweden, warning that its rejection of lockdown was likely to leave the virus rampant with an R possibly as high as 3 or 4. That is to say: every person infected was giving it to three or four others. Here’s the graph Imperial published on 30 March for Sweden.
The virus was shown to be spreading far faster in Sweden, said Imperial, “not because the mortality trends are significantly different from any other country” but because “no full lockdown has been ordered so far”. Nor was a full lockdown ever ordered. So what did this imply for Sweden? Imperial didn’t translate the above graph into deaths, but when its assumptions were published others joined the dots. A Lund university academic warned that it meant 85,000 deaths for Sweden. An Uppsala team, feeding Imperial’s parameters into its own study, agreed. The modelling envisaged Sweden paying a heavy price for its rejection of lockdown, with 40,000 Covid deaths by 1 May and almost 100,000 by June.
The latest figure for Sweden is 2,680 deaths, with daily deaths peaking a fortnight ago. The virus, it turns out, has been spreading at a fraction of the speed suggested. So Imperial College’s modelling – the same modelling used to inform the UK response – was wrong, by an order of magnitude. Of course, as the saying goes, all models are wrong but some models are helpful. To be helpful, Sweden has now published its own graph saying its R was never near the 4 that Imperial imagined and has, in fact, been below the safe level of 1 for the last few weeks.
As Johan Norberg wrote , Imperial’s model ‘could only handle two scenarios: an enforced national lockdown or zero change in behaviour. It had no way of computing Swedes who decided to socially distance voluntarily. But we did.’ Anders Tegnell, Sweden’s state epidemiologist, has seen his trust ratings soar. Some Swedes are even having his face tattooed on their arm.
When Imperial first made its models, everyone was guessing. We know more now. Every day, in The Spectator‘s Covid-19 email, we bring new studies that add more detail to our understanding of the virus. At present, Britain is considering the South Korean model: an ambitious combination of tech, surveillance, track and trace. But given that Sweden achieved what Imperial College had thought undoable, without the surveillance or the tech or the loss of liberty, its lessons are also worthy of consideration.
Sweden’s Prime Minister has said he is relying on ‘Folkvett’ – people’s wit, or common sense. As Boris Johnson considers his options, he should also ask whether a folkvett option – described in a recent Spectator leading article as a ‘trust the public’ approach – might work for Britain.
PS For all of its prominence in virus modelling, ‘the R’ is not a known number. It can only be guessed at, because the actual number of infections can only be guessed at. It won’t be uniform around any country. This is why Sweden has not targeted “the R” – its calculation was part of its general analysis of the problem. Sweden has simply sought to keep the virus at manageable levels (ie, so hospitals have spare capacity) which – as it points out – has had the side-effect of keeping a low “R”. But the UK’s approach is more influenced by models, and No10 now says keeping an R below 1 is its main policy. Its policy was explained in this briefing last week:-
PPS Imperial College has been in touch to stress (as I should have done in more than a link) that it did not go so far as to translate its figures into death figures: this was done by other academics in Sweden. I’ve updated this article to make that clear and included, in addition to the Uppsala link, an English-language interview with Prof Paul Franks of Lund University who warned that Imperial’s figures meant an impending caseload for the healthcare system of up to 370,000 Covid patients. Imperial also pointed me to a Sweden page on its website where the data has been changed. The 30 March claim that Sweden had spent weeks with an R near 4 now seems to have vanished without explanation. I asked why: better data, it says, became available. Quite.
Now hear the analysis on the Prime Minister’s latest address from Katy Balls, James Forsyth and Fraser Nelson on the Coffee House Shots podcast: