Thanks for posting this.
I believe one thing they are missing is the issue of people's actual behavior in one region or another or at one time or another. I would imagine that the effectiveness of a mask requirement in Asia is different than a mask requirement in the US. In the US, I think a mask requirement issued on March 15 would have a different effectiveness than a mask requirement on Dec. 15.
I think one can see a possibly unconscious bias by those that wrote this paper by the fact that in Figure 2 they did a delta ("Additional effect...") for their last category but didn't do that on the other categories (ie, they didn't show "Additional effect of limiting gatherings to 100" after showing the effect of limiting gatherings to 1000, which looks to be a 10% drop). Had they similarly shown "Effect of stay at home including the above", you would see a much bigger number. They do show cumulative effects in other figures.
Also, I believe there is a negative effect of doing nothing. That is, many of the effects time out as people grow restless.
I expect there is a large amount of slop in these numbers due to effects of the perceived direness of the situation by the public, where the public adjusts their behavior in other ways. For instance, the first ban on meetings of 1000 in the US probably also decreased out-of-region spread, exposure in airports, hotel stays. But it probably also caused fewer meetings of <1000 and possibly impacted other optional gatherings.
While it is a convenient way to treat it, I don't buy as proven that these effects are best treated as per cent reduction on R. An alternative way would be as a percent reduction on R-1, which isn't necessarily right either. I notice the article doesn't mention super-spreading events. I still don't understand why at least some versions of R seem to always be increasing or decreasing but rarely nearly static. (It suggests that the first derivative of R is roughly constant over some periods of time.)
This particular study focused on European countries. It would be worth comparing to the earlier study on Canadian regions, but the earlier study used different metrics for the effectiveness so I couldn't do it quickly.
I believe one thing they are missing is the issue of people's actual behavior in one region or another or at one time or another. I would imagine that the effectiveness of a mask requirement in Asia is different than a mask requirement in the US. In the US, I think a mask requirement issued on March 15 would have a different effectiveness than a mask requirement on Dec. 15.
I think one can see a possibly unconscious bias by those that wrote this paper by the fact that in Figure 2 they did a delta ("Additional effect...") for their last category but didn't do that on the other categories (ie, they didn't show "Additional effect of limiting gatherings to 100" after showing the effect of limiting gatherings to 1000, which looks to be a 10% drop). Had they similarly shown "Effect of stay at home including the above", you would see a much bigger number. They do show cumulative effects in other figures.
Also, I believe there is a negative effect of doing nothing. That is, many of the effects time out as people grow restless.
I expect there is a large amount of slop in these numbers due to effects of the perceived direness of the situation by the public, where the public adjusts their behavior in other ways. For instance, the first ban on meetings of 1000 in the US probably also decreased out-of-region spread, exposure in airports, hotel stays. But it probably also caused fewer meetings of <1000 and possibly impacted other optional gatherings.
While it is a convenient way to treat it, I don't buy as proven that these effects are best treated as per cent reduction on R. An alternative way would be as a percent reduction on R-1, which isn't necessarily right either. I notice the article doesn't mention super-spreading events. I still don't understand why at least some versions of R seem to always be increasing or decreasing but rarely nearly static. (It suggests that the first derivative of R is roughly constant over some periods of time.)
This particular study focused on European countries. It would be worth comparing to the earlier study on Canadian regions, but the earlier study used different metrics for the effectiveness so I couldn't do it quickly.