Thread by @GidMK

Fascinating. The paper itself merely compares three different models of the reproductive number of COVID-19 in different places. For lockdowns:

Model 1 = big benefit
Model 2 = small benefit
Model 3 = mixed benefit

https://twitter.com/DrEliDavid/status/1343359007004774400

Model 1 was published to great acclaim in a paper in Nature nature.com/articles/s4158…

The crux of this new preprint appears to basically be saying that this previous publication was pretty shit

And look, it's hard to disagree that a lot of COVID-19 models are terrible

What's weird is how denialists are seizing on this paper as evidence that lockdowns didn't work

Some quotes from the paper:

"We do not claim that lockdown measures definitely had no impact"

"our results should not be interpreted...that NPIs are totally ineffective"

"substantial uncertainty may remain despite the best efforts of the modelers"

Oddly enough, I'm not sure the conclusion in the abstract actually agrees with the lengthy conclusions of the paper itself, which is a bit confusing

If nothing else, the paper demonstrates that lockdowns reduce R(t) by a reasonable amount, even if that is from 0.9 to 0.6, and this could reduce the overall death count in a pandemic wave by many thousands

Although worth noting that I've been saying this for months, from mine and other data I suspect that the actual implementation of restrictions at the start of the pandemic had less of an impact than the caution around the disease itself

Anyway, it's absurd cherry-picking to say that "lockdowns had little or no benefit" from this paper, what it appears to show is that models used to determine the benefits of lockdown are pretty crap