A Tale of 39 Counties: Re-opening Washington and Political Orientation

A Tale of 39 Counties: Re-opening Washington and Political Orientation
A Tale of 39 Counties: Re-opening Washington and Political Orientation
[Co-author of this article is Eric Tyberg, a retired IT executive and consultant residing in Lincoln, California. Originally from Falun, Wisconsin, he rose through the ranks at IBM and formed his own consulting business when IBM downsized.]
In concluding our article comparing the increase in cases between Skagit and Whatcom counties as the initial phases of re-opening take place, we noted that it appeared that higher increases in COVID-19 cases were found when they were aggregated for the 27 Republican-leaning counties than when aggregated for the 12 Democratic-leaning counties. This piece follows up on this observation.
Both our earlier piece and this one are motivated by research suggesting “…that Republicans are about three times more likely than Democrats to say their state is moving too slowly to reopen business and ease restrictions, and Democrats are more likely than Republicans to report taking preventative measures like wearing masks in public” .
We know a wide range of factors affects transmission, infection, and reporting, and we acknowledge we are not able to cover them all here. They include: testing and its accuracy; data suppression; potential “super-spreader” venues such as group quarters, e.g., assisted care centers, college dormitories, prisons, military barracks, naval vessels, and prisons; food processing plants; protest and other rallies and marches; blood type; and demographic factors including population density, rates of interaction, population age, race and ethnicity, and gender (more here). These and other variables have the capacity to produce a lot of “noise” in attempting to identify and isolate factors that affect the transmission of the COVID-19 virus.

In the process of examining the four demographic variables just listed, we found that in spite of the “noise,” that “race/ethnicity” stands out in terms of its correlation with per capita case rates, which we operationalized as the percent of the total population in a county that is White/Non-Hispanic. Looking at the state’s 39 counties, we found there is a strong inverse relationship between the percent White/Non-Hispanic and per-capita case rates. That is, as the percent of the population that is White/Non-Hispanic increases, the per capita case rate decreases. The relationship is shown the graph at left. The model generating the (dashed) linear trendline providing the “best fit” to the data is given in the small box in the upper right of the graph. A guide to understanding this model can be found at Wikipedia.
Because of the nature of the distribution of the state’s population, one needs to use caution in looking at correlations across the state’s 39 counties: the 2019 state population is 7,546,410, but 27 of the state’s 39 counties have populations less than 100,000. King County is the largest with a 2019 population of 2,226,360, while Garfield is the smallest with a 2019 population of 2,200. However, looking at per capita case rates relative to the percent of the population that is White/non-Hispanic provides us with a means of controlling for the asymmetrical population distribution. Continuing with our analysis, however, we decided that the effect of political orientation on the increase of COVID-19 cases would best be analyzed by comparing the results of: (1) aggregating the data across the 12 Democratic-leaning counties; and (2) aggregating the data across the 27 Republican-leaning counties.
As you can see in the table below, the increases in cases since April 1st are not only far higher for the 27 counties that voted for Trump (rose-colored background), but the gap is widening. Between April 1st and May 1st, the overall increase in the 27 Trump-voting counties was 394%, by June 1st it was 887%, and as of June 10th it was 1,086%. For the 12 Clinton-voting counties (blue-colored background), the respective increases are far lower, 237%, 309%, and 324%.
Importantly, the percent of the total population that is White/non-Hispanic in the aggregate of the 12 counties carried by Clinton in 2016 is lower than found for the aggregate of the 27 counties carried by Trump. All else being equal, this difference suggests that the counties carried by Trump should have a lower rate of increase in cases, which is not the situation we see in the table.
Here, it is worthwhile to note that seven counties approved to move from phase 2 to phase 3 of the re-opening process (Columbia, Ferry, Garfield, Lincoln, Pend Oreille, Stevens and Wahkiakum) are all Republican-leaning counties with very small populations, each of which has a high percent that is White/Non-Hispanic (ranging from a low of 73 percent in Ferry County to a high of 93 percent in Garfield County). They also tend to be sparsely populated, very rural and isolated. In addition, they had very low or zero case counts as of April 1st (Garfield still reports no confirmed cases) and have largely remained so.

In spite of the low case counts found in these seven Republican-leaning counties, when we look at the 28 Republican-leaning counties as a whole, they have noticeably higher case increases than the 12 Democratic-leaning counties as a whole. (see table at left) Is it because, as suggested by the research described earlier, that the former are less likely to be observant about social distancing, wearing masks, hand washing, and other pandemic containment guidelines than the 11 Democratic-leaning counties as a whole? It appears that if this were not the situation, then the latter would not be exhibiting noticeably lower increases in COVID-19 cases since April 1st than the former, even in the face of all of the factors and “noise” associated with this pandemic.
Data Sources. The COVID-19 data are from The Coronavirus Resource Center maintained by Johns Hopkins University) and are slightly different than the summary provided in the earlier article because of how we handled cases unallocated to specific counties. These differences affect neither the results nor our conclusions. The demographic data are for 2019 and taken from Small Area Demographic Estimates (SADE) by Age, Sex, Race and Hispanic Origin. Voting data are taken from Wikipedia).
12 Comments, most recent 4 years ago