This is perhaps interesting. You can do a pretty good job of predicting recent R in each state based on Apple's mobility data and urbanization. No surprise, but cases are falling more where people are moving around less. 
But controlling for mobility, they're also falling more in rural states than urban ones. Montana isn't having many COVID-19 cases even though mobility is high there, for instance.
R's are estimated using the technique described here and attempt to account for the number of tests. They reflect a median of the previous 4 weeks, which should capture the post-lockdown period fairly well.
fivethirtyeight.com/features/coron…
fivethirtyeight.com/features/coron…
FWIW, there may also be some effects from partial herd immunity in edge cases, e.g. cases are falling faster than the model thinks they "should" be falling in NY. This effect is likely to be negligible for most of the country, though.
p.s. This would suggest the decline in social distancing from where it was just after lockdown to where it is now will raise R by something like 0.1. That's not *that* much. But since we were barely (if at all) reducing cases before, it could be enough to matter.
Plus, distancing will likely erode further. I don't think people should *necessarily* expect to see an *explosion* of new cases. It may lead to a long plateau at high numbers, though.
Maybe we get lucky and weather offsets some of this, although then there are problems in fall.
Maybe we get lucky and weather offsets some of this, although then there are problems in fall.
Another spin on this: if we had ways OTHER than social distancing to lower R (test/trace/isolate/etc), we could perhaps reduce social distancing quite a bit, so long as people were generally being prudent and we tackled some of the low-hanging fruit (e.g. avoid mass gatherings).
