This year's model (UK remix)

I have posted several times about the COVID-19 modelling that governments in Canada (specifically my home Province of Ontario) have been relying on to guide policies aimed at combatting the pandemic. As each successive wave of infections has started to build, governments have released modelling that supposedly shows catastrophic rises in infections lying just over the horizon. Without fail, these warnings seem to have been followed by a prompt reversal in the course of the disease, with new infections starting a steady decline.

I have tried to make the point that there is or at least should be a distinction between modelling, which takes account of a range of causative factors, and extrapolation, which simply assumes that the most recently observed growth rate will continue indefinitely. Just about every scary model here in Ontario has been extrapolation, sometimes from a very flimsy base indeed -- in one case, back in the winter months,  it appears to have been a projection based solely on the worst single day of the pandemic!

It's no surprise, but also no consolation, to find that the same thing may be happening in the UK. Boris Johnson's decision to proceed with his fatuous "Freedom Day" on July 19 led to an outcry from scientists and physicians. With cases of the delta variant rising sharply, they predicted an apocalyptic outcome, with cases soaring from the then-current rate of under 50,000 per day to 100,000 or even 200,000 in a matter of mere weeks.

Less than a week on from Freedom Day, what has actually happened?  Cases have begun to fall, and not just marginally. The daily infection count briefly moved above 50,000, but in recent days the figure has been around 30 percent below prior week levels -- and the number published today (July 25) was a startling 40 percent lower. Epidemiologists are lost for an explanation and are warning that this may be a false dawn, not least because Freedom Day itself may trigger a rise in new infections, which might not show up for another week or so. Fair enough, but when your forecast is wrong not just as to magnitude but as to the actual direction of change, you surely have some explaining to do.

I don't write these posts because I'm anti-vax or anti-lockdown -- very far from it. Nor am I anti-statistician -- that was part of my professional education. But the fact is, forecasts that consistently appear to overstate the threat run the very obvious risk of triggering a backlash: you were dead wrong the last time, and the time before that, and the time before that, so why should we believe you now?  Brits were on the streets of London yesterday protesting against the COVID restrictions, even though Freedom Day saw most of those restrictions removed across England. That's dumb -- although the English do love a good ruck --  but politicians need to ask themselves just what that kind of dumbness might be telling them. 

UPDATE, 27 July: This quote from the BBC website, attributed to one Professor Adam Kucharski, may inadvertently tell us more about the "modelling" than the Professor intended: "One of things you've got to remember is there is a lot of infection still out there. If behaviour changes, then you're only two doublings away from 100,000 (25,000 cases to 50,000 cases to 100,000 cases)."  Right. And if my granny had wheels she'd be a streetcar.