• First published in 2004 as Bartholomew’s Notes on Religion (BNOR).

    Previously at:

    Email me
    (Non-commercial only)

  • Archives

  • Twitter

  • Supporting

  • Recent comments

A Note on a Media Version of Neil Ferguson’s “Track Record”

From a column by Matt Ridley and David Davis MP in the Daily Telegraph (and reproduced here):

Is the chilling truth that the decision to impose lockdown was based on crude mathematical guesswork?

…It is not as if [Neil] Ferguson’s track record is good. In 2001 the Imperial College team’s modelling led to the culling of 6 million livestock and was criticised by epidemiological experts as severely flawed. In various years in the early 2000s Ferguson predicted up to 136,000 deaths from mad cow disease, 200 million from bird flu and 65,000 from swine flu. The final death toll in each case was in the hundreds.

The link there is to a previous article, which outlines criticisms by Michael Thrusfield, professor of veterinary epidemiology at Edinburgh University, and Alex Donaldson, head of the Pirbright Laboratory at the Institute for Animal Health, both in relation to the foot-and-mouth disease modelling. The various figures quoted above, though, derive from a short list that has appeared on the Telegraph‘s “Politics Live” feed a few days before.

However, I’m sceptical that we can extrapolate a general theory of Ferguson’s incompetence based on one high-profile controversy, and the other examples of Ferguson’s “track record” – a short selective list chosen by the two authors based on UK media familiarity, so nothing about his work on influenza, Ebola or ZIka – are not fairly presented. A computational biologist named Nicolas Bray has pointed to the sources:

On BSE/vCJD: As quoted by Bray, this comes from a 2000 article in Nature: “We show here that the current mortality data are consistent with between 63 and 136,000 cases”

On bird flu: Bray describes this as a “context-free quote which appears to be discussing a *potential* bird flu pandemic”. Source is a Guardian article.

On swine flu: “the worst that might realistically happen”. Source is paragraph 82 of a Science and Technology Committee Parliamentary transcript.

On vCJD, the team of which Ferguson was a part revised its estimates as new evidence came in. As described in by Philip Yam’s book The Pathological Protein: Mad Cow, Chronic Wasting, and Other Deadly Prion Diseases:

Assuming an incubation period of less than 20 years, the U.K. will probably experience only several hundred vCJD cases, according to an August 2000 estimate by [Roy M.] Anderson and his colleagues. Their estimate for the maximum number of cases was 136,000 — and that number assumes an incubation period that can be as long as 60 years. “This would make it unusual, but it cannot be ruled out,” remarked Anderson’s colleague Neil M. Ferguson, especially considering that kuru can incubate for more than 40 years. In 2002, the team lowered its estimates, giving a range of 50 to 50,000 vCJD deaths between 2001 and 2080; in February 2003, it dropped its estimate further, to 10 to 7,000 deaths.

According to Guardian profile of Ferguson, the team

came up with an estimate that was incredibly broad for the likely number of human deaths – between 50 and 50,000 – but that was at a time when some were predicting 2 million people would be infected. There were calls for the sort of NHS resources now going into Covid-19 to be directed towards vCJD. Ferguson and [Christl] Donnelly’s modelling helped defuse that. In the end the UK had about 170 cases.

This works against the media narrative Ferguson as a man whose “track record” consists of sensationalist scaremongering predictions that are then proved to be wildly off-base.  Ferguson’s work, like anyone’s, is open to criticism, but why resort to misrepresentation?

The focus of Ridley and Davis’s article is on the alleged shortcomings of computer coding used in the Imperial College model of Covid-19 infection. It is difficult for non-scientists to make informed assessments here, although I will note informal reactions from a Berkeley biologist named Michael Eisen that the critique amounts to a “ridiculous attack”. Eisen writes that “Nobody thinks the Imperial model is flawless, but its top-line COVID predictions are the result of basic math – the model just fleshes them out”, while Bray adds that Ridley “doesn’t even realize that, as a stochastic model, it’s *supposed* to produce different outputs”.

A defence of Ferguson’s work on Covid-19 by Bryan Appleyard appeared in the most recent Sunday Times.

Donnelly and Ferguson’s 1999 monograph Statistical Aspects of BSE and vCJD: Models for Epidemics can be browsed on Google Books.

2 Responses

  1. The Free Zone of the Republic of China has been exemplary in how it has dealt with the plague from Wuhan on Mainland China: it has closed borders to those likely infected and isolated them so that the Free Zone is truly free, by and large, of the infection. Taiwan has only had about 400 instances of the plague and only about six deaths, without having to shut-down its economy.

  2. Not an entirely convincing rebuttal of Professor Ferguson’s critics. In fact, very weak indeed. Should ‘worst case’ scenarios become the model for government response? I happen also to have been a mathematical population biology modeller in my misspent youth, and know full well that these are simply there to suggest possible factors that need to be better scrutinized. I would have never dreamt that Ferguson’s flimsy models would become the basis for a wholesale lockdown of a country, with disastrous consequences. How ignorant can you get?!
    But then I should have known better, after the disproportionate and quite unnecessary cull of animals in the foot and mouth outbreak of 2001, which so distressed the farming community. I believe Ferguson was even awarded an OBE on the strength of his inept and misguided recommendations on that. Quite shocking!

Leave a Reply to Tom Ehrman Cancel reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.