Political Covid

Some thoughts as COVID spreads to different parts of the country and various maps and articles try to explain what's going on and its significance.  Focusing again on the NYT coverage because I think they've done a pretty good job with maps and analysis and I think there's more value in trying to improve the good stuff out there than simply highlighting the bad.

This July 2 piece tries to show the change based on region, density, race, and politics.  Their map:

Pretty good overall, showing a clear trend and the obvious fact that lots of counties in the south and west have lots of cases per capita - and that the northeast counties generally have much lower current infection per capita.  My main pet peeve, as with my last post, is that showing this by county on a per capita basis draws our attention to small counties with a few cases and perhaps ignoring the counties that are contributing the most to the problem. So again we have giant cones of red dotting Arkansas and Mississippi while LA county - which has more daily cases than both states combined - looks like a little bit.  So not bad if you're interested in the details - but you're liable to miss the big national picture. 

But on to politics.  Here's their summary:

Again, OK at highlighting a trend - counties won by Clinton are about where they were at the last peak and counties won by Trump have more than twice the infection rate as March and April.  But we can surely do better.  States and counties, after all, aren't Trump or Clinton; they're filled with voters for both.  It makes a big difference that New York City, which bore the brunt of the epidemic in April, went for Clinton with more than 80% of the vote.  To get a better understanding of how Trump and Clinton voters are encountering COVID, we need to look at more than which counties were "Red" or "Blue".

So let's start with one baby step forward.  Assume for now uniform distribution of COVID within counties so that, for example, a county with 60% Clinton voters, 40% Trump voters and 100 COVID cases leads to 60 Clinton Covid cases and 40 Trump COVID cases.  

OK, so we see the same general trend: bluer at the beginning and redder now.  But the degree of changed is dampened.  From the NYT analysis we have more than 20K (almost 23K) cases in blue and fewer than 10K (maybe 8K) in red; in ours we see the blue never reached 20K in the first peak and the red was about 12.5K at the beginning.  From the NYT analysis the current day blue cases are about at the earlier peak, but here we see they are much higher.  From the NYT analysis, red cases have more than doubled from the earlier peak, but here we see it's up by about 70%.

Maricopa county in AZ barely went for Trump in 2016, so all those Phoenix cases are goin to "Trump counties" in the NYT analysis; likewise, Houston cases will all go to "Clinton counties".  The reality is some of the large metro areas in the southwest are about evenly split and surging in cases - be wary of analysis that attributes all of them to one side of the political divide.

The long and short is that there does appear to be some political evening out of infections, but just not as dramatic as the NYT analysis might suggest.

Parting thoughts:

• There are some cases where declaring a locality red or blue might make sense.  One set of rules might be enacted for a certain place, and that might be driven largely by the majority.  For example, a 55% red county might open up faster, have less enforcement of mask wearing, etc.
• My analysis above is only a step forward.  Cases are certainly not uniformly distributed within counties.  Even within counties, democrat-leaners might live in denser areas with higher COVID rates; richer people within a county might lean republican, and be less likely to work jobs having COVID exposure, etc.