This article is taken from the February 2022 issue of The Critic. To get the full magazine why not subscribe? Right now we’re offering five issue for just £10.

When their predictions of doom fell apart spectacularly last summer, sticking the boot into Covid modellers seemed unnecessarily cruel. Cruel because they had tried their best, and unnecessary because the government would surely never take their prognostications seriously again. 

But then winter arrived and with it the Omicron variant and a new set of “projections” that put us within a hair’s breadth of a Christmas lockdown. Our lives have been in the hands of these glorified numerologists for too long. It must never be allowed to happen again.

Some people suspect the modellers of wanting to always ramp up restrictions. SAGE can certainly be accused of being overly cautious and the modellers seem to suffer no financial or professional repercussions from being too pessimistic. 

But there is one source of bias that is often overlooked. If the government introduces new restrictions, predictions based on the government doing nothing cannot be tested. If they are not tested, they cannot be shown to be wrong. It is therefore in the interests of the modellers for the government to do something.

The government has called their bluff on three occasions in the last year and each time it has ended badly for the number-jugglers. In July, a team from the London School of Hygiene & Tropical Medicine produced a model containing dozens of different scenarios. All of them projected cases, hospital admissions and deaths rising through July and peaking in August or September “if roadmap Step 4 is enacted on 19th of July 2021.” Step 4 was enacted and the number of cases fell by half.

In September, SAGE published three scenarios for the autumn with a peak in hospital admissions ranging from 2,000 to 7,000 a day in England “if no further changes in behaviour or policy take place”. Without any change in policy and with no obvious change in behaviour, they peaked at 955 on 27 October. 

When Omicron emerged in December, Neil Ferguson’s team at Imperial College produced a model for “a high-income country setting with substantial prior transmission and high vaccine access” which was obviously England. In every scenario, there was a daily peak of at least 40 deaths per million in early 2022, equivalent to more than 2,500 deaths a day in England. 

Not to be outdone, Warwick University projected 2,890 deaths and 13,600 admissions a day in a relatively optimistic scenario. Their “realistic worst case scenario” projected over 5,000 deaths and 27,000 hospital admissions per day. Since either of these situations would have rapidly overwhelmed the NHS, the modellers helpfully provided some less depressing scenarios involving varying degrees of lockdown should the government be minded to avert disaster.

Summarising the evidence, SAGE gave themselves a huge margin of error and concluded that Covid deaths in England this winter would reach between 600 and 6,000 a day. Anything less was apparently unthinkable. The pressure on Boris Johnson to “follow the science” and put the country into lockdown must have been immense. It is to his credit that he resisted. 

I write this in mid-January when there is still time for a twist in the tale, but there are currently 200 Covid deaths a day and little sign that this number will quadruple, let alone rise by a factor of 30. 

Moreover, there are signs that the Omicron wave peaked on New Year’s Eve, which may come as a surprise to academic mathematicians but is less of a surprise to those of us who remember the previous wave peaking on the day England played Italy in the European Championship final. If the modellers had a better understanding of the nation’s social calendar, they might have a better hit rate. 

“To be completely honest, I don’t think any of the modellers have a good explanation for what happened,” said Graham Medley, the chair of SAGE’s SPI-M group, when England’s wave of infections mysteriously subsided after the end of an international football tournament that was largely hosted in England. 

People who cannot interpret recent trends have little chance of correctly predicting the near future. The plain fact is that the self-proclaimed experts have spent much of the pandemic as bemused observers, endlessly startled by events.

In April, after the predicted surge of infections from reopening schools failed to occur, Mike Tildesley from the Warwick team declared himself to be “really pleasantly surprised”. In late July, when it was clear that “Freedom Day” had not led to 100,000 infections a day, Professor Neil Ferguson — who had previously said such an outcome was “almost inevitable” — cheerfully admitted that he was “happy to be wrong if it’s wrong in the right direction.” 

Something has gone badly wrong when likely outcomes are routinely treated as black swan events

This flippancy ignores the social and economic costs that would have been incurred had the government been panicked into maintaining restrictions, but at least Ferguson admitted he was wrong. The lamest defence of the modelling is that it can neither be right nor wrong since it does not offer predictions, merely “projections” and “scenarios”. Where to begin with this sophistry? Leaving aside the dictionary definition of a projection is “a prediction based on known data”, all scenarios are conditional, but once the conditions have been met they become forecasts. If you say that Manchester United will beat West Brom 3-0 if Fernandes plays and it doesn’t rain, and West Brom go on to win despite Fernandes playing and the clouds not gathering, you were wrong. You don’t get to weasel out of it with some gaslighting wordplay. 

It is not as if the models only seem flawed in hindsight. The idea that Covid could kill 5,000 people a day in heavily vaccinated England was always insane. It would require an infection rate so enormous that the whole thing would burn itself out within days. 

In an interview with The Times, Graham Medley admitted that the chances of there being 7,000 hospital admissions a day in October was as unlikely as Burnley winning the Premiership, but it was included in the model because it was theoretically possible. By implication, admissions never reaching 1,000 a day — which is what actually happened — was not included in the model because it was considered to be essentially impossible.

Something has gone badly wrong when likely outcomes are routinely treated as black swan events. It is not true, as some believe, that the modellers only look at worst case scenarios. Perhaps the most pitiful aspect of this pseudo-science is that the models include such a huge range of scenarios that dumb luck should ensure that one or two of them hit the mark. In reality, with only a handful of exceptions, a blindfolded chimpanzee could do better. This suggests something more than mere incompetence. It suggests a systematic bias towards pessimism and a hankering for restrictions. 

Whatever the reason, there can be no way back for the modellers after this. They have made astrology look like physics. Although I would never advocate tarring and feathering anyone, they should never be heard from again.