The Government’s Latest Scary Modelling is Already Wrong

There follows a guest post by Daily Sceptic reader Graham Williams (a pseudonym), a maths graduate and by profession an analyser of business plans, models, forecasts and funding requests. He is not impressed with the latest Government pandemic modelling.

I have just read the SPI-M consensus statement paper of September 8th, which appears to be at the heart of the recent stories about possible future lockdowns etc. This paper seems to be as big a load of negative, hyperbolic scaremongering as all the ones they have issued so far this year (February at the start of the roadmap, March, April, June and July).

In paragraph two they state: “SPI-M-O groups have reflected on their modelling of Step 4 of the Roadmap, and despite unexpected falls in cases in mid-July 2021, these scenarios can still be used to consider the future autumn and winter trajectory.”

They appear however not to have reflected that were it not for the unforecast Delta variant their modelling since February would have overstated the position of deaths, cases, and hospitalisations by June 21st by around 1,000%. Even with the rise caused by the variant, their forecasts remained hugely overblown, but they still continue to model with the same flawed methodology.

After paragraph two there follow about 18 paragraphs of largely unsubstantiated waffle with a few facts thrown in.

One of the facts is that R is currently (i.e., at the date of the paper) between 0.9 and 1.1, so broadly flat. The covering page to the report says: “These are not forecasts or predictions… They are based only on the observable trends and data available at the time the projections were produced.”

Had the modelling actually done what it said on the tin, project observable trends, then it would have been in line with their own medium-term projection of September 8th, which shows a fairly flat trend for September, even if arguably the base they have used is a bit low.

Instead of doing this, SPI-M’s modellers produced three scenarios based on different R numbers. Of course, it appears to be hardwired into their models that R can never be less than or equal to one, so despite suggesting the modelling follows the recent trajectory of 0.9 to 1.1, the numbers they use are 1.1, 1.5 and 2.0.

Lo and behold, when you run those through the models you get a rapid increase in numbers, leading to the 7,000 hospital admissions a day clickbait headline numbers on the 2.0 scenario.

Even discounting that clickbait modelling, SPI-M’s refusal to countenance an R of less than one and only consider a range of 1.1 to 1.5 means that in only a week since they prepared the paper the numbers are already significantly adrift (forecast hospital admissions 850-950, compared to an actual seven day rolling average of 724) and the discrepancy will likely widen as the recent drop in case numbers feeds through.

It may be that some of their predictions come true, but the document reads as an attempt to scare the largely innumerate members of the Cabinet into imposing further restrictions rather than a balanced piece of advice.

Notify of
Newest Most Voted
Inline Feedbacks
View all comments
Would love your thoughts, please comment.x