Are Trump voters dying from COVID at a higher rate?
In the never-ending quest to politicize everything and grow division among Americans, NPR has published this Pro-Trump counties now have far higher COVID death rates. Misinformation is to blame. The premise of the piece is that Trump voters are misinformed about COVID & as a part of that, have lower vaccination rates, leading to more COVID deaths. They compare % of vote for Trump with the COVID Deaths per 100K by county. They choose a start date of May 2021 with the explanation that this is the time at which vaccines were readily available to everyone. Here is the plot that summarizes their support for the article’s headline:
The blue + is the average of Biden heavy counties, the black + is the national average, and the red + is average of Trump heavy counties. Wow, Trump counties are at ~100 deaths per 100k since May, while Biden counties are <50. I guess those knuckle dragging Trump counties’ misinformed citizens are paying for their ignorance in deaths.
Well, just to be a pain I decided to see if I could replicate this analysis. Using my normal set of data + the election results published from MIT, I can basically recreate the above plot (48 continental states, minus FL, NE, NJ as the deaths reporting is inconsistent in those states):
Rather than trying to come up with averages for Trump heavy, or Biden heavy, it seems a simple linear regression makes sense to try:
The R2 of 0.16 indicates that 16% of the variance in Deaths per 100K can be explained by the % of Trump voters within a given county. The slope of 1.5 means that for every additional 1% of Trump voters, we can expect 1.5 more COVID deaths per 100K of population. This would seem to indicate a strong relationship between a county voting for Trump and the COVID death rate.
But is there possibly more to the story? It seems I possibly have heard that there is some degree of seasonality with COVID and different regions are hit at different times of the year (sarcasm alert). And isn’t it the case that voting patterns are also very regional. Perhaps we can drill down into the data a little further to see if seasonality (i.e. time dimension) plays a role in this. First, let’s look at quarters:
That’s interesting. There is essentially no relationship in April and May, and then there is a relationship in Q3 & Q4, although those relationships are weaker than if we look at May-Dec as a whole? 🤔
We can drill down even further to look at months:
Hmm. May, June, and July show no relationship. Were Trump voters better informed in those months? Then in Aug-Nov they suffered from misinformation, only to get better informed in December again? This is all very confusing.
Just throwing this out there, but could it possibly be that July-Nov is when seasonal patterns hit the South the worst and since a large part of Trump’s support was in the South, that’s what we are seeing in those months? Let’s look at % of deaths in the North vs. South:
Not surprising that it would appear Trump counties had higher deaths in July-Nov when the South was making up a disproportionate amount of the deaths. By the time December is over, it may be that Biden counties experienced a higher COVID death rate in December. I’m sure NPR will do a follow up piece at that time.
Please excuse the snark in this post, this kind of sh*t just drives me nuts.
Had this conversation with a friend. Another point might be to look at variance attributable to urban (Democrat) vs rural (Republican) and health care access and quality.
But the question is how long have they sat on this analysis before releasing it? Did they have the model 3 months ago and waited for the seasons to change? My money says yes. Unfortunately I don't have that much money because the people capable of such shenanigans have probably already taken it from me.