Covid Herd Immunity vs Interventions

What matters more? Herd immunity or behavior/interventions? Louisiana provides an interesting case study.

On the surface, it seems like all of Louisiana was heavily impacted after reopening despite already experiencing a significant first wave.

But if you break it down by counties, the data tells a different story. New Orleans and Jefferson Parish, which were heavily impacted in March/April have not seen the same dramatic resurgence in July. Most new cases came from counties that were largely spared early on.

There's a very negative correlation (-0.8) between the number of deaths in Apr/March and the number of deaths from Jun-Aug.

Why might that be? 

It could be that people in the counties became more aware after the 1st wave and changed their behavior accordingly. However, Louisiana did not issue a state-wide mandate for mask-wearing or close bars until July 13th and cases had already plateaued by then.

Or it could be that a high level of immunity in a population helps further curb the spread. Youyang Gu of MIT, from whom I stole these fabulous charts, has estimated that about 20% of New Orleans had been infected by May 1, which some new mathematical models, which take into account the heterogeneity of disease susceptibility, are suggesting might be high enough to achieve herd immunity and that might explain the fall.

I think that after you take into account what we have seen in Sweden, which has not significantly adjusted their intentional lack of interventions, but who has seen a similar trajectory, that gives me a little more conviction that immunity rather than interventions is the key feature mitigating a resurgence in cases/deaths. It further highlights that the threshold for herd immunity, estimated at perhaps 60% for a vaccine, might be as low as 20% with respect to natural infection because the disease initially propagates at a higher rate among the most susceptible in a population, thereby acting like a targetted vaccine, leaving a residual population that has a very low R0.