FULL Fact is one of those over-confident websites that claims, unlike all those other websites which peddle mere opinion, to be supplying readers with the pure, unfiltered facts. It has now brought the weight of its authoritative wisdom to bear on lockdown sceptics, with one Leo Benedictus penning a ‘fact check’ entitled ‘Can we believe the lockdown sceptics?‘ The London Economic ‘digital newspaper’claims this piece ‘definitively discredits lockdown sceptics‘.
Unusually, it actually makes some effort to be kind to sceptics and says ‘we must not dismiss the lockdown sceptics’ claims out of hand . . . Mr Hitchens, Mr Cummins and Dr Yeadon all say that we should base our views on evidence, and on that we agree with them. However, on the overwhelming balance of the evidence, their claims are wrong’.
Benedictus claims there is much published evidence that lockdowns work.
He writes: ‘A research paper published in June in Nature, one of the most prestigious scientific journals in the world, concludes: “Our results show that major non-pharmaceutical interventions – and lockdowns in particular – have had a large effect on reducing transmission.”
‘Another in Science in July said: “Focusing on COVID-19 spread in Germany, we detected change points in the effective growth rate that correlate well with the times of publicly announced interventions.”
‘Another in the BMJ said: “Earlier implementation of lockdown was associated with a larger reduction in the incidence of COVID-19.”
‘Mr Cummins may have been referring to a research paper in one of the Lancet’s online journals in July, which said that “full lockdowns… were not associated with COVID-19 mortality per million people”.
‘But so much research is published, especially on Covid, that it is often possible to find at least some evidence in support of almost any view.
‘For instance, he might instead have referred to a different paper in the same journal eight days earlier, which said: “With a tighter lockdown, mobility decreased enough to bring down transmission promptly below the level needed to sustain the epidemic.”
‘Some science supports Mr Cummins on some points, in other words, but a great deal – which he does not mention – contradicts him.’
Let’s take a closer look at these.
The June Nature paper is by Flaxman et al and is a model-based paper by the same team from Imperial College whose modelling inspired the lockdowns in the first place, so they are not exactly an impartial team. The June paper is notorious for making many dubious assumptions, as documented here by Nicholas Lewis. In fact, so troubling is the paper that last week Nature itself published a ‘matters arising’ paper by Soltesz et al pointing out many of its flaws.
It says: ‘We conclude that the model is in effect too flexible, and therefore allows the data to be explained in various ways. This has led the authors to go beyond the data in reporting that particular interventions are especially effective. This kind of error – mistaking assumptions for conclusions – is easy to make, and not especially easy to catch, in Bayesian analysis.’
You can read our own critique of the Flaxman paper here.
The July Science paper, by Dehning et al, focuses only on one country, Germany, so cannot be used to counter international comparison data. In addition, Germany had an atypical spring as it did not experience any excess deaths.
The July BMJ paper is by Islam et al and is also model-based. It suffers from only using ‘case’ data (which is unreliable as it depends on who is being tested and how many tests are being carried out) and does not look at deaths. To measure lockdown strictness it relies on the Oxford Covid-19 Government Response Tracker rather than using mobility data, and only considers five broad types of intervention rather than any finer measure of severity. This means it counts Sweden as implementing as many restrictions as many of the other countries because it did something in four out of five categories, despite Sweden’s restrictions being much lighter than others. Then the results are not actually very favourable to lockdowns at all. They show huge variation between countries, with all those with fewer restrictions seeing a reduction in cases over the period at a rate greater than many of those with more restrictions. Overall the study finds restrictions reducing ‘cases’ by just 13 per cent on average, a paltry impact for the immense cost.
On the other hand, the July Lancet paper by Chaudhry et al that Benedictus dismisses as an example of it being ‘often possible to find at least some evidence in support of almost any view’ is in fact the main peer-reviewed article which carries out an international comparison of Covid mortality. It is notable that Benedictus doesn’t even attempt to counter the point that Covid mortality was not associated with lockdown strictness and timing, because he can’t. It is a fact, as Chaudhry et al show.
The Lancet paper by Vinceti et al that he then cites by way of riposte only looks at Italy and only at case numbers, not deaths. It concludes that one lockdown was ineffective while the second one worked, which it attributes to strictness. However, the context to keep in mind is that Italy had one of the worst Covid mortality rates in the world during the spring, and the later lockdown may have ‘worked’ only because it coincided with the change of the season, a possibility the study does not consider. The conclusion that the first lockdown did not work supports the sceptics’ position.
The lack of consideration of ‘organic’ alternatives as explanations for epidemic decline such as warmer weather and rising population immunity is a general failing of many of the coronavirus studies, made more obvious by the fact that the epidemics declined across all contexts regardless of how many or few restrictions were in place. In each region, the lifecycle of the epidemic followed a Gompertz curve.
To back-up his ‘fact check’, Benedictus claims that the epidemic in all regions in England declined at the same time, suggesting it was a result of lockdown. However, this is not true. As the Government dashboard shows, London deaths peaked around April 5, North West deaths peaked around April 13, and Yorkshire and Humber deaths peaked almost two weeks after London on April 18; London’s decline was also steeper. The seasonal effect of course provides an alternative explanation as to why the epidemic declined across the country at around the same time during the warm April weather.
Benedictus makes the mistake of assuming that everyone who’s been infected will test positive for antibodies in order to ‘prove’ the Covid Infection Fatality Rate is higher than many sceptics claim. However, numerous studies have shown that not all people who are infected develop or retain antibodies. For instance, a large Spanish study found that less than 20 per cent of symptomatic cases later had (IgG) antibodies, and a large Italian study similarly found only 25 per cent of symptomatic cases later had (IgG) antibodies. (See here for more on this.) The role of T-cells and pre-existing immunity in the immune response has also been shown by a number of studies. (See this in the BMJ and this pre-print.)
Benedictus says it is ‘hard to see how banning human contact could fail to reduce the spread’, not recognising that lots of contact isn’t banned – many workplaces continue, supermarkets and other shops remain open, often schools are open, and there are hospitals and care homes. Furthermore, confining people to homes can increase spread within homes.
Benedictus disputes Peter Hitchens’s claim that it was ‘not possible’ for the lockdown announced on March 23 to have caused the subsequent decline in daily infections and deaths, saying it is not ‘generally accepted’. But Chief Medical Officer Chris Whitty himself stated to MPs in July that new infections were falling before lockdown. Oxford’s Professor Carl Heneghan has pointed out that GP data shows suspected Covid referrals dropped off before lockdown. The standard estimate for the average lag between infection and death is 23-26 days, which is the figure used for example by the Imperial College team in Flaxman et al, which puts peak infection around March 15, 23 days prior to peak deaths on April 8 and nine days before the lockdown came into effect. London death peak was three days earlier, putting its infection peak on March 12.
Benedictus claims that the second waves in the autumn disprove sceptics’ claims about herd immunity. However, it’s worth bearing in mind that some of the worst ‘second waves’ have been in countries not badly affected in spring, such as Switzerland, Germany, Slovenia and Czechia (Czech Republic).
It’s true that the UK did experience excess mortality in November and December, but only peaking at around 20 per cent and declining during December. Other causes of death are running below average, suggesting there is some misattribution going on; the other obvious cause of extra deaths this year is lockdown. It’s not true to say that lockdown sceptics necessarily expected an easy ride for hospitals in the autumn, with many anticipating further Covid deaths during the colder months, particularly in areas not strongly affected in spring. The point about there being no second wave is that this autumn/winter Covid epidemic is much more like an ordinary seasonal viral epidemic than something similar to the spring when the disease was new, and this time round infections seem to be caused by local outbreaks rather than a national tsunami. The point about false positives and pseudo-epidemics is not that there is no Covid around any more but that problems with the testing regime turn something eminently manageable and basically normal into a crisis that appears much bigger and more disastrous than it really is.
Benedictus claims that lifting the lockdowns explains the resurgence of the virus, but this is untrue as lockdowns were lifted much earlier in the year yet there was no new surge until the autumn. The resurgences, insofar as they are real and not an artefact of increased testing, are much better explained by the onset of colder weather.
A more general criticism of the ‘fact check’ is that it fails to look at the countries and states such as Sweden, Belarus, Tanzania, North and South Dakota and others in the autumn such as Switzerland and Spain which did not impose strong restrictions but did not experience higher rates of Covid mortality than those that did. This means it neglects to consider the key question: what would happen without lockdown? Benedictus claims ‘we will never know the effects of things we did not try’, but of course we can look at other places that didn’t lock down. He quotes a Government report claiming lockdowns saved thousands of lives by preventing hospitals being overwhelmed, but fails to spot that hospitals were not overwhelmed in those regions that forwent lockdowns. Worth remembering that hospital services in the UK and other countries are routinely stretched in the winter, with people often being treated on trolleys in corridors.
In sum, Leo Benedictus’s attempt to ‘fact check’ the lockdown sceptics falls short. He fails to engage our most important points about the countries which show epidemics declining without interventions, dismisses the significance of the main international study comparing different rates of Covid mortality, and makes factually dubious claims himself, such as that the epidemic in England began to decline in all regions simultaneously.
Must try harder.
This article was first published on Lockdown Sceptics on December 30, 2020, and is republished by kind permission.