There's a new paper by John Ioannidis and co-authors that's intended to push their anti-lockdown message by performing a flimsy empirical analysis. Adding to @GidMK's thread, I will just highlight one flaw that should have prevented this paper from being published but hasn't.

What they basically do is regressing case growth in Spring (terrible data) on a bunch of NPIs. Then they're puzzled bc their results indicate a positive "effect" of some NPIs on case growth. Well, that's not puzzling at all, that's because NPIs and case growth are endogenous!

NPIs tend to get tougher the worse the case growth gets. But case growth might already be taking off when you've just tightened your NPIs, making cases further grow exponentially before the NPIs do anything. This results in a positive correlation between NPIs and case growth.

But obviously, this correlation is not causal; it's not the NPIs driving cases higher, it's the worsening epidemic. For the paper, this means their estimates are biased and worthless for what they want to investigate bc they don't come close to causal effects of NPIs.

The funny part is: They admit that in the paper; they're saying that as long as NPIs aren't randomly allocated, which they clearly aren't, there's bias. Their excuse: 'Everyone else is having the same problem!'

I'm fairly certain that in economics, this paper would have been rejected with some very unfriendly comments. But the journal has already published several highly criticized Ioannidis papers in 2020. Note that this type of analysis is quite far from the journal's aims and scope.

Link to Gideon's thread: https://twitter.com/GidMK/status/1349164532627693570 and a bit more background on endogeneity and COVID-19: https://www.intereconomics.eu/contents/year/2020/number/3/article/common-pitfalls-in-the-interpretation-of-covid-19-data-and-statistics.html