I have a distinct memory of watching the first of the pandemic press conferences; the first time Chris Whitty, Patrick Vallance and Boris Johnson stood up behind those podiums in the wood-lined room and told us what was going to happen. It was March 3, 2020. They unveiled the “flatten the curve” plan — to slow rather than stop the virus’s spread, so that we would slowly build up population immunity while preventing the NHS being overwhelmed.
It seemed reassuring. I know this because I messaged something like “It seems like they might know what they’re doing!” in a science-nerdy chat group that I’m part of. But my optimism was not echoed. “It seems like they’ve just killed all our grannies,” responded someone.
I recall this to make two points. Firstly, I can’t claim any sort of foresight. I didn’t see then, though perhaps I should have done, that the early response of the British government to Covid was disastrous and wrongheaded. Instead, I was falsely reassured by the confidence of the scientific advisers, Whitty and Vallance.
But secondly, other people did see it. The failings were predicted. And not just by my friend: a large number of smart, numerate generalists outperformed public health experts repeatedly in predicting the course of the pandemic in those early months.
On Monday night, two House of Commons select committees released the findings of their report into England’s Covid response. The headlines yesterday were stark: the “worst public health failure ever”, “big mistakes”, “damning”.
The individual mistakes — failure to protect care homes; failure to move quickly enough on lockdowns; failure to build testing infrastructure and more — are certainly worthy of criticism. But having read the report, I want to talk about what strikes me as the overarching theme: being too bloody clever by half.
Let’s return to early March 2020, when I was falsely reassured by the Whitty/Vallance/Johnson press conference. The “flatten the curve” strategy that the three men introduced involved slowly introducing public health measures one by one. The idea was to carefully control the R value – the number of people that each infected person would, on average, give the disease to. The idea was to keep it at just above one, so that the disease would slowly spread through the population rather than either burning through it or dwindling down.
In paragraph 87 of the report, the committee writes: “It is striking, looking back, that it was accepted that the level of Covid-19 infection in the UK could be controlled by turning on particular non-pharmaceutical interventions at particular times.”
“Striking” is certainly one word for it. This was a novel disease. We knew almost nothing about its behaviour. On March 6 2020, I went to a gig at Alexandra Palace. Looking back now, it should not have been allowed to go ahead — but because the consensus at the time was that the disease was spread by fomites, by touching surfaces, the only advice was to keep washing our hands. In fact, it is spread largely through the air.
So we had a completely wrong idea how the virus behaved, but we nevertheless had a confident plan to control it very precisely by now banning large gatherings, then closing schools, then encouraging working from home. We were trying to thread a needle blindfolded.
This sort of overconfidence in highly uncertain science crops up over and over again. As well as flattening the curve, the phased introduction of distancing measures had another purpose: it was not believed that the public would accept long-term restrictions on their freedom. “There is a risk,” Whitty told the public in one of the early press conferences, “that if we go too early, people will understandably get fatigued and it will be difficult to sustain this over time.”
This was also the official Sage position: “Sage agreed that a balance needs to be struck between interventions that theoretically have significant impacts and interventions which the public can feasibly and safely adopt in sufficient numbers over long periods.”
It’s not clear what this claim was based on. Other behavioural scientists questioned it, with 200 of them signing an open letter to the Government asking for the evidence. But what evidence were they hoping to receive? What comparable situation have we ever faced? A few small-scale experiments about people obeying instructions under laboratory conditions probably can’t tell us anything at all about what people would be willing to do in order to save hundreds of thousands of their fellow citizens, very possibly including their own elderly relatives. Nonetheless, we based highly counterintuitive policy decisions — not imposing strict lockdowns even as the disease was spreading and killing, despite seeing it being successful elsewhere — on this apparently unevidenced belief.
And let’s be clear: the coming hundreds of thousands of deaths were very much known about. The report rightly says that we over-relied on a plan for pandemic flu. That was a huge mistake, and it led us to strange things like moving to “delay” the virus rather than “contain” it, and stopping community testing.
That meant that we had no idea how widespread the virus was. “The UK was reduced to understanding the spread of Covid-19 by waiting for people to be so sick that they needed to be admitted to hospital,” says the report. “For a country with a world-class expertise in data analysis, to face the biggest health crisis in a hundred years with virtually no data to analyse was an almost unimaginable setback.”
So, yes, sticking to the flu plan was a disaster. But if you read the flu plan, it wouldn’t have worked any better for flu. The disaster that arrived was completely foreseen, yet it confidently said we couldn’t do anything about it, so we shouldn’t try. It said that planners should “aim to cope with a population mortality rate of up to 210,000–315,000 additional deaths” within 15 weeks of a virus arriving, and should “plan for a situation in which up to 2.5% of those with symptoms die”.
The lowest number there, 210,000 over 15 weeks, works out at 2,000 deaths a day. At the peak of the second wave, the seven-day rolling average of deaths briefly – for nine days, in fact — broke 1,200 a day.
That period was horrifying, but the pandemic flu plan was to “cope” with at least 2,000 deaths a day for 15 weeks. That appears to have been based on the idea that we, the British public, were so profoundly wedded to liberty that we would rather let the equivalent of the entire population of Norwich die in three months than have the nightclubs closed.
More than that, in fact, it was important to keep them open so that people didn’t get depressed: “Large public gatherings or crowded events where people may be in close proximity are an important indicator of ‘normality’ and may help maintain public morale during a pandemic.” The possibility that people might not want to attend large public gatherings if it might kill them or their family, or might value the lives of their elderly relatives highly enough to stay home for a few weeks rather than watch them die in their thousands, was not considered..
Over and over again, the British response was one of unwarranted confidence in highly uncertain outcomes. We were, of course, dealing with a largely unprecedented situation. But that meant that the evidence was necessarily weak.
So statements such as “There is very limited evidence that restrictions on mass gatherings will have any significant effect on influenza virus transmission” (from the pandemic flu plan) were on the one hand true, but also basically meaningless. There’s very limited evidence either way.
But the common sense attitude would be that being in a large group of people probably helps the virus to spread. And given that the alternative, which the plan explicitly endorses, is at least 210,000 people dying in 15 weeks, it seems incredible that the approach was “well, there’s no evidence that it works, so we might as well not try”.
There are other areas where this overconfident, clever-clever approach backfired. For instance, Matt Hancock told the committee that in the early months, they were advised that “testing people asymptomatically might lead to false negatives” so you shouldn’t do it. Of course, it would also lead to lots of true positives. But the fear of a hypothetical second-order effect — like Jenny Harries’s claim last March that wearing masks could increase the risk of infection, because it would make people more confident — overshadowed the very obvious and real risk of disease. It’s like the plan was based around not wanting to be obvious, but always being clever and counterintuitive.
And the reliance on the pandemic flu “flatten the curve” plan meant that there was no modelling of the impacts of a lockdown. Dominic Cummings told the committee: “You are going to have to lock down, but there is no lockdown plan. It doesn’t exist. Sage haven’t modelled it.”
Often, simply plotting an exponential curve on a graph might have done better than the more complex forecasts the Government used. Vallance told a press conference that we were “four weeks” behind Italy, when simply looking at the number of deaths showed that we were 14 days behind. And when other countries locked down, with some success, we didn’t learn from their example, because we had a plan, and we were confident in it.
And, of course, border closures. Naively, you’d think that stopping people coming into the country would reduce the likelihood of the disease coming here. But overconfident modelling told scientific advisers that it wouldn’t, so they trusted that, rather than thinking “OK, but if the model is wrong then we’ve completely messed up.”
This isn’t to place all of the blame on Whitty, Vallance and Sage — or to exonerate the politicians. There was plenty of cowardice and dithering on the part of the Government, not least (as the report shows) when they ignored scientific advice to impose a short “circuit breaker” lockdown in autumn 2020, presumably because they feared being unpopular.
But the valuable lesson, I think, is not “we should have done X” or “we should have done Y” specifically, but that we should be less confident in our ability to predict highly complex situations. And more than that: we should look at the possible results of being wrong in our attempts to predict those highly complex outcomes. If you’re 60% sure that lockdowns won’t work, then that implies there’s a 40% chance they will, and that implies that you think there’s a 40% chance of saving hundreds of thousands of lives. Instead, it seems like the whole way through, the thinking was “it probably won’t work so let’s not try”.
I didn’t foresee any of this. I should have: it was all there in the pandemic flu plan. Others, smarter and more conscientious than me, did, or at least foresaw that disaster was a realistic probability. But the takeaway should be not “in the next pandemic, we need to protect care homes” — although we should — but “in the next pandemic, we should stop trying to be so damn clever”.