COVID-19 has thrust the relationship between science and policy into the spotlight. Before midMarch, for most Canadians COVID-19 was a remote concern, with a single COVID-19 death in Canada, a B.C. man in his 80s with underlying health problems.
Then, suddenly, the world changed and, within weeks, practically the entire Canadian economy was shut down. What happened? A team led by Neil Ferguson at Imperial College London released a series of papers that took the world by storm, predicting tens of millions of deaths in a worldwide pandemic glibly compared to the Spanish Flu, the most deadly epidemic in modern times.
Ferguson predicted that, unmitigated, COVID-19 would kill 510,000 in the U.K., 2.2 million in the U. S., and 326,000 in Canada. The only option, warned the report, was radical physical distancing of the entire population for up to 18 months until a vaccine was invented. Less restrictive measures, such as isolating suspected cases and physical distancing of the elderly and at-risk, he predicted, would merely reduce that toll by half; what was needed was total isolation, which he predicted would bring deaths in Canada down to a still-catastrophic 46,000.
Even at the time, Ferguson’s numbers were at odds with other researchers, most prominently a much lower projection from an Oxford University team, which predicted pre-existing immunity would cut the rate of transmission. Meanwhile, Ferguson’s lockdown recommendations directly contradicted World Health Organization recommendations published only last year.
In the weeks after Ferguson’s lockdowns were implemented, it became clear his model was deeply flawed. Surprisingly, the code that generated Ferguson’s doomsday scenarios was neither publicly available nor peer reviewed. Thousands of lines of the code were not even documented, leaving doubt as to how he got his results.
One senior software engineer from Google found amateur errors in Ferguson’s code itself, including giving different answers depending on the processing speed of the computer running the model, rendering the results utterly unverifiable. The auditor noted that such code would be unacceptable in the private sector, for example, in car insurance.
Meanwhile, reporters uncovered evidence that Professor Ferguson himself has a history of hysterically over-predicting deaths from new diseases. In 2002, he predicted up to 150,000 deaths from CJD (“Mad Cow Disease”) — 55 times the actual death toll of 2,704. In 2005, he predicted that Bird Flu could kill up to 200 million people. The actual death toll was 455.
We now have data from countries that ignored Ferguson’s recommendations, where even restaurants and bars remained open. In Japan, for example, Ferguson predicted 1.4 million deaths, when the actual number as of June 18 was 935. In Korea, he predicted 381,000 deaths, compared with 280 on June 18. In Taiwan, he predicted 212,000 deaths, when the actual number of dead was seven. Even in hard-hit Sweden, he predicted 85,000 deaths, compared with the actual number on June 18 of 5,053.
Here in Canada, COVID-19 is tracking above the seasonal flu, with 8,300 deaths as of June 18 compared with 3,500 in a typical flu season. Given that daily deaths are down 75 per cent from the peak and falling fast, even Ferguson’s 46,000 projection with maximal restrictions is likely to be grossly exaggerated.
Several recent papers from Ferguson’s own Imperial College London and from the University of California have argued that lockdowns “worked” — essentially, better safe than sorry. However, both papers admit that the lion’s share of mitigation was likely from voluntary distancing. U. S. restaurant traffic had already fallen by half before any lockdowns were announced. One recent survey of European countries indeed concluded that generalized lockdowns may have had very little impact at all on deaths.
This matters because there are real health costs to lockdowns that we cannot glibly dismiss. Aside from concrete deaths from delayed surgeries and missed detection of cancer among the middle- aged, economists have long known that mass unemployment and bankruptcies carry enormous health consequences that are very real to the victims suffering joblessness, drained life savings, ruined businesses, mental and physical health deterioration, even suicide.
Had we known the true scale and parameters of the threat we might have therefore avoided these tragedies with better- targeted policy. For example, COVID has killed far fewer Canadians under age 35 than traffic accidents normally do, while 80 per cent of Canadian COVID deaths have occurred in seniors’ centres. It may have been more prudent, and even saved more lives, to focus resources on safeguarding the elderly and immunocompromised instead of reckless economywide lockdowns.
Further concerns involve early overestimation of fatality rates. American health officials have already revised COVID fatality estimates down by an order of magnitude, while one Stanford epidemiological survey estimated the median COVID fatality rate for those under age 70 at just 0.04 per cent — one in 2,500. Another paper just this week estimated that up to 8.7 million Americans had been infected by the end of March, 80 per cent undetected, suggesting the actual fatality rate for COVID has been grossly overestimated.
Going forward, we need to fortify our policy “immune system.” One former Indian bureaucrat put it well: “Emergency situations like this pandemic should require a far higher — and not lower — level of scrutiny” since policy choices have such tremendous impact.
At a minimum, the public has a right to expect that studies driving policy, from lockdowns to carbon taxes, are robust and transparent, using data and code open to public scrutiny and subject to rigorous scrutiny.
Peter St. Onge est chercheur associé senior à l’IEDM et l’auteur de « COVID-19 : le modèle défectueux qui a confiné le Canada ». Il signe ce texte à titre personnel.