June 2, 2020•465 words
I'm interested in how many people will die because of what Dominic Cummings and Boris Johnson have done: Cummings by treating the lockdown rules as something that apply to other, lesser people, and Johnson by demonstrating that doing that is just fine. The result will be that people take lockdown and social distancing less seriously, and some of them die as a result.
So I wrote an epidemic simulator. It's fairly simple-minded, but it does the susceptible / infected / immune-or-dead thing. And I modelled three scenarios:
- a mitigated epidemic where lockdown followed by some weakening happens;
- a version of the previous model where, after the initial lockdown, things are weakened by about 5% more, with this declining over time to 2%
- a version with 10% weakening, declining to 4%.
So this is meant to model the case where, because of what Cummings and Johnson have done, people are about 5% or 10% less likely to obey the restrictions.
The numbers below are worth what you paid for them: I hope that someone with a serious model will run equivalent scenarios, and I'm sure they have, but I could not find any when I looked a few days ago.
The model runs used a population of a million: I've scaled the results to 70 million. The figures are for deaths: it assumes a death rate of 1% of infected people. Figures are averaged over 5 runs, and lasted for 3 years.
The initial results indicated that the actions of Cummings & Johnson will cause tens of thousands of deaths. This seems far too high to me. However my epidemics are slower than the real one seems to be, so I also did some experiments moving the effect later in time as well, to see if I could reduce the effect.
|when||unmitigated||mitigated||mitigated + 5%||mitigated + 10%|
The figures without the Cummings-Johnson effect vary because the model is statistical and there is some variation.
The late figures indicate between a few thousand extra deaths and perhaps ten thousand. Everything is to three significant figures, which is less than the variation between runs: bigger populations have less variation but take longer to simulate, and I got bored with my machine being hot.
I really want to see someone with a proper model publish estimates. But I am sure that Cummings and Johnson have been privately told the probable results of what they have done, and I'm sure Johnson knew them before he chose not to sack Cummings. Johnson may not fully understand them ('numbers, what? jolly boffins deal with those'): Cummings may do.
What neither of them do, of course, is care: so they've killed some little people, does it matter?