Review whether politics or economics impact a county’s vaccination rate?


by Michael Bitzer

In the middle of summer, I posted a simple analysis which looked at North Carolina’s 100 counties and their full vaccination rates (second shot) versus whether a county voted for Trump or whether other factors could affect the county’s vaccination rate.

In that previous analysis, Trump County’s vote rate was statistically significant, which means we can confidently say that it impacts and helps explain a county’s vaccination rate. But it appeared to be in combination with a county’s per capita income, along with a few other “variables” that ultimately helped explain a county’s 62 percent complete vaccination rate.

Recently, the New York Times reported an analysis that showed that in the counties that voted overwhelmingly for Trump (70 percent or more) “the virus has killed about 47 in 100,000 people since the end of June,” compared with 10 in 100,000 in those counties that voted the opposite (less than 32 percent for Trump).

I decided to update the data for the number of full vaccinations in each county in North Carolina, so I created the percentage of those aged 15 and over for a county’s full vaccination rate to replicate the analysis of July.

First, a scatter chart of the county’s full vaccination rate against the Trump County vote:

back inside July, the scatter plot showed a linear r-squared value of 0.164; now, that number is 0.338, showing an improvement in explaining a county’s vaccination rate using Trump’s vote rate in 2020.

I decided to do a linear regression to see the corrected r-squared rate, which turned out to be 0.332; which means that a third of a North Carolina county’s vaccination rate could only be explained by the county’s percentage of votes for Trump. Previously, the same adjusted squared r number was 0.155.

Next, I ran a scatterplot analysis of county per capita income versus vaccination rate. In July, the r-squared for that scatter chart was 0.427. Now, the value had dropped to 0.394.

Again, running a linear regression with per capita income only, the adjusted r-squared value came in at 0.387, up from 0.422 in July.

In July, combining the Trump County percentage, the county’s per capita income, and other factors such as a county’s non-white population percentage, COVID case percentage, and mannequin variable for urban or rural counties, the overall explanation for a county’s vaccination rate was 0.625.

In re-running the same model, several dynamics have changed. The county’s non-white percentage, COVID case rate, and the dummy variable for rural counties all became statistically insignificant. The urban variable was just significant (at level 0.5). But the two main variables – Trump County percentage and per capita income – remained statistically significant.

In July, the explanatory power (corrected r-squared) of these six variables was 0.625; now, the corrected r-square was 0.609, a slight decrease.

However, when I eliminated all but the two main variables (Trump and per capita income), the corrected r-squared was 0.595. Just a further slight drop in explanatory power, but with only two variables now, a noticeable difference.

To summarize what we’re seeing in terms of explanatory power (the corrected r-squared values) for the July and end-September time analyzes for all 100 NC counties and their vaccination rates:

Using the reported analysis in the New York Times (those counties with over 70 percent support for Trump and less than 32 percent), I looked again at those counties’ scatter charts.

Now, a word of statistical caution: for this analysis, the number of NC counties drops from 100 to 26, something that most statistical analyzes would say “you should really be careful about hiring such a small ‘n‘(number of observation units) with this analysis. ”

With that caveat, here’s the scatter plot of per capita income versus 26 counties:

The r-squared value rose to 0.676, meaning potentially two-thirds of the full vaccination rate in those 26 counties could be explained by per capita income.

But when it came to the scatter chart of the Trump County vote for the 26 counties:

The r-squared value rose to 0.785, which means that nearly 80% of the full vaccination rate in the 26 counties could be explained by Trump’s share of the county in the 2020 election.

Indeed, when linear regression is performed with just the two variables, the per capita income rate loses statistical significance (at 0.089), while using the Trump variable alone had an adjusted r-squared rate of 0.776.

Thus, in those 26 counties, the single variable of Trump’s 2020 vote rate explains three-quarters of a county’s full vaccination rate.

Of course, there are likely other explanations for the dynamics taking place at this point in the pandemic. We will need to continue observing the dynamics of COVID vaccinations in counties, but nearly three months after the initial analysis, it appears that overall, North Carolina County’s Trump vote rate is becoming a better “explainer” than the percentage. of vaccination of a county.



Please enter your comment!
Please enter your name here