Hi Joel, thank you for your thoughtful comments. I'll try to explain better something that I've said many times. After reading concerns like the one you reiterate here, I went back to my models and did some experiments changing contact patterns- proportional to social distancing.
I made those changes in a number of ways, either increasing or decreasing individual variation in connectivity during lockdown. Estimated herd immunity thresholds didn't change much with these kind of alterations.
I have actually put some thought into this but my explanations don't seem to be getting across. To me these results are saying that much of the selection that lowers HIT happens very early in epidemics, before social distancing.
So, however interventions change contact patterns doesn't seem to matter much for HIT estimation presuming that when they are removed completely, contacts will be reset to pre-pandemic patterns.
Having said this I totally agree that relaxing social distancing should be done gradually and vigilantly, but I am confident on our estimates for the 4 European countries we analysed.
My informal team and I have been addressing the criticisms presented to us and the results stand so far. To make this process more efficient we are now preparing a set of files with the main code to make it easier for others to play with the models if they wish.
I think it has been counter productive to have so many (well-intended) people raising concerns publicly without having actually worked out the maths. By the time we work it out and respond, the presumed criticisms have already spread widely as if they were established concerns.
This hampers the inclusion of our models in the set that policymakers look at and disproportionately affects public confidence on the relevance of our results. Generally, I think this is being detrimental on multiple fronts.
