Using Data to push more Political Division

A critique of the recent "unvaxxed Republican excess death" study.

Image from fortmorgantimes.com

If you want to stir the political division pot just bring up anything associated with vaccines and the messages about the evil, dangerous Republicans who don’t want them.

As someone who isn’t affiliated to any political party I’ll be the first to tell my readers that the intermingling of science and politics to the degree we are witnessing is a very dangerous game, including the use of vaccination rates as an own on any specific party.

But of course, that’s not how the world works, and just recently a study came out arguing that excess death rates among Republicans are caused by the vaccines stoking even more of this divisive rhetoric.

One account, as reported in Slate, provides a rather, let’s say heavily biased account of this study. I encourage people to read it and notice that this is 20% study assessment and 80% narrative and divisive rhetoric a la “Republicans should be held accountable for killing of their own constituents!”

Click on the image above for a link to the article.

Again, I’m not a Republican so this type of angle doesn’t work on me.

It certainly didn’t work for self-professed progressives such as Bret Weinstein and Heather Heying who themselves have taken a stance against the vaccines along with other Democrats such as Pierre Kory of the FLCCC, as well as Jimmy Dore who has staunchly criticized the underreporting and obfuscation of adverse reactions.

And heaven forbid people make up their own minds and think for themselves outside of needing to associate with a political party in order to do so!

With that being said, the real meat of the article is looking at exactly what this study says and whether vaccines are the only factor at play here (hint: it’s not).

Excess Deaths and Political Debauchery

This study¹ was published in The National Bureau of Economic Research (NEBR)— a strange place to see a study on excess deaths, although we can’t ascribe anything specific to where a study gets published (at least not on the surface).

Click on the image for a link to the article.

This study was conducted by a group of Yale researchers, and interestingly the article first starts off with this little bit of information:

We gratefully acknowledge support from the Tobin Center for Economic Policy at Yale University and the Yale School of Public Health COVID-19 Rapid Response Research Fund. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.

Interesting. Make of that what you will.

Anyways, the gist of the study is that researchers looked at excess death rates and stratified the data based on political affiliation to look for differences between the two groups.

Excess death rates were defined by the researchers as follows:

Our approach estimates “excess death rates” as the percent increase in deaths above expected deaths that are due to seasonality, geographic location, party affiliation, and age.

Simple enough. Essentially, the researchers looked at death rates from 2020-2021 and compared them to death rates in 2019 matching for month to see if a higher percentage of deaths have occurred post-vaccination rollout.

The results of this first assessment can be seen below in the following graphs:

From Wallace, et al. Figure 1. The first chart shows aggregate excess deaths within the given timeframes and based on the measured parameters. The second chart shows stratification between political party affiliation as measured by voter registration data from 2017. The third chart shows difference in excess mortality between Republicans and Democrats.

The first graph looks at the aggregate excess death percentage without any stratification along party lines. It’s a straightforward graph, but note that it shows 3 large waves starting spring 2021 and going on into the winter 2021 season, indicating the seasonality of the disease as well as the emergence of the Delta wave in the summer of 2021.

However, you’ll notice that in the 2nd graph, which does stratify along party lines, that there appears to be a higher level of excess deaths among Republicans as compared to Democrats.

Even more important, the gap around August/September 2021 onward is rather large, although there appears to be a gap around the winter months of late 2020/early 2021.

This is also reflected in the 3rd panel which shows the percentage difference in excess deaths between Republicans and Democrats.

Taken together, the evidence appears—note, appears— to suggest that the vaccines are the reason for this discrepancy.

The researchers make this remark in regards to these results (emphasis mine):

In the second panel of Figure 1, we plot excess death rates for Republicans and Democrats separately. In 2018 and the early parts of 2020, excess death rates for Republicans and Democrats are similar, and centered around zero. Both groups experienced a similar large spike in excess deaths in the winter of 2020-2021. However, in the summer of 2021 — after vaccines were widely available — the Republican excess death rate rose to nearly double that of Democrats, and this gap widened further in the winter of 2021.

[…]

This sharp contrast in the excess death rate gap before and after vaccines were available suggests that vaccine take-up likely played an important role. Data on vaccine take-up by party is limited and unavailable in our dataset, but there is evidence of differences in vaccination attitudes and reported uptake based on political party affiliation [13, 10, 7]. Using county-level vaccination rates, we find evidence that vaccination contributes to explaining differences in excess deaths by political party affiliation, even after controlling for location and age differences.

If I was a journalist, this is all I would need in order to publish a hit-piece².

However, you may have noticed a little notation on the right of the 1st panel/graph. If not, I boxed it in black below:

That’s right. This study that supposedly looked at Republicans and Democrats only looked at voters in two states.³

Even more alarming, this study used voter registration data from voters in 2017:

To calculate excess deaths, we use 577,659 deaths of individuals linked to their 2017 voting records in Ohio and Florida who died at age 25 or older between January 2018 and December 2021.

This is very dangerous, as the researchers are essentially tying voter registration data from before COVID was even known to the public to argue the fact that there’s some association between party affiliation and excess death in 2021.

Wouldn’t a better measure be, I don’t know, voter registration in 2020 amidst the pandemic and at a time in which the early release of the vaccines were going to occur in the coming months, likely making this an issue for voters?⁴

The issue here is that the researchers have essentially produced a mismatch in their dataset. They argue in the excerpt above that attitudes about vaccines fall along party lines. But again, this is based on evidence and attitudes from during the pandemic, and would clearly not be reflected in voter attitudes in 2017 to the same degree as it would amidst a pandemic.

Who’s to say that voters in 2017 who went red would have being against the vaccine in 2021? Who’s to say that many blue voters would be in support of it? How many voters would vote for candidates not affiliated with their party given that 2016 was considered one of the most divisive elections to date? And how many voters may have changed their political party in 2020 based on their beliefs on the vaccines, which wouldn’t be reflected in 2017 voter registration data?

The researchers never took into account the chance that voters may have changed their political affiliation within the 4 years since that voter registration data, so how reliable is the use of 2017 voter registration data in making arguments for 2021?⁵ How likely is a voter to change their political affiliation given that we had one midterm and one general/presidential election in the time since, along with another midterm on the way (hint, hint)?

The researchers even note that the use of two states is a limitation in their study and may not generalize to the country:

Third, our study is based on data from the only states where we could obtain voter registration information (Florida and Ohio); hence, our results may not generalize to other states.

Both Florida and Ohio are battleground/swing states, so it’s likely that this would make record-keeping of voter registration even more meticulous compared to other states.

But again, how true to life is the evidence when only two states are used to generalize to the country?

The last figure in the article looks at county vaccination data between Democrats and Republicans in each county and plotted them against their excess death rates. Essentially, they took one county, looked at vaccination rates among party lines, and mapped them to come up with a curve.

One point of contention here is how vaccination was measured. As indicated by the researchers county vaccination data was collected on June 2021 and based on evidence of at least one vaccine dose:

We obtained additional information from the CDC. From the CDC, we also accessed data on county-level vaccination rates as of June 23, 2022. That data can be accessed here: https://data.cdc.gov/Vaccinations/ COVID-19-Vaccinations-in-the-United-States-County/8xkx-amqh. From that data, we obtained information on what share of a county’s population had received at least one dose of a Covid-19 vaccine as of June 6, 2021. We selected this date—two months after vaccines became available to all adults in our study states—because it represented the approximate time when all adults had had the opportunity (if they so desired) to complete the two-dose vaccine series as well as the additional 14 days following the second dose needed to be considered “fully vaccinated” according to the CDC.

Sort of a bold assumption to make here, as I would argue that this study is probably a better measure of the people who were “all in” on the vaccines from the get-go versus the people who were hesitant with the early rollout.

They’re also assuming that a large subset of these participants received their second dose within the given timeframe, and didn’t even include who got what. So we don’t even any any data outside of the initial rollout and receiving the first dose only.

Confounding variables galore going on here, and I would be concerned at the level of extrapolating occurring given these factors which the researchers are making assumptions about.

However, I would like to include some context here— something that the researchers appeared to have glossed over.

If vaccination was the only variable being measured here, we should at least expect the graph on the right to match the graph on the left (which I have questions for since there would be no vaccination data to stratify and therefore no numbers to create a curve for the left graph).

In essence, low vaccination rates should show higher excess deaths when compared to the higher vaccination rates.

Instead we see that, if we follow the curves on the right, that even among those with low vaccination rates there is still a discrepancy seen between Republicans and Democrats, and this discrepancy is pretty wide.

Wouldn’t these curves at least indicate that something else aside from the vaccine is at play?

The timepoint measured here for vaccination came right before the surge from the Delta wave, and it’s reported that Delta hit harder than any of the prior variants, especially for the most vulnerable groups such as the elderly and those with comorbidities.

At the same time the researchers are arguing over vaccination rates, this discrepancy may be due to differences in obesity rates, access to healthcare facilities, rates of diabetes, cancer, and differences in age given that the vaccination rate is what’s being controlled for in the construction of the curve—essentially many other variables besides the vaccines.

Now, this will present with a highly controversial comment, but we should remember that COVID is far more deadly for the elderly, and given the fact that the average life expectancy in the US has declined since the onset of COVID we may infer that the elderly are the largest proportion of deaths due to COVID.

Of course, we’ve known this for a long time. It’s not a secret that the elderly are the most vulnerable against a novel pathogen.

Given this fact, and the fact that the elderly are likely to have many comorbidities, vaccination for the elderly may be more beneficial (note: may be) than for younger individuals who may fare better with COVID and are at risk for more adverse reactions from the vaccines.

So it could very well be that this discrepancy may be a measure of excess deaths among the most vulnerable who, at then onset of the more deadly Delta variant, succumbed to the illness and may have had some protection from the vaccines.

This is what I would infer based on Figure 3, but again we are working with data that has not accounted for any other variables, and the researchers even note within their limitations that they couldn’t match vaccination rates with actual voter registration data, meaning they actually do not, in fact, have data on individuals to work with even though the study uses “individual-level dataset” quite often:

Second, because we did not have information on an individual’s vaccination status, analyses of the association between vaccination rates and excess deaths relied on county-level vaccination rates.

Using data to push Narratives

There’s more I can state, but instead I’ll argue that this study is one that is done not for the sake of providing proper evidence, but done more as a way to drive a narrative and create more divisive rhetoric—the Slate article is proof of that.

It’s interesting that the writer of the Slate article, Don Moynihan, is a political scientist at Georgetown University.

Let’s just say a quick peruse of his Twitter profile shows that the wears his politics on his sleeves (NOTE: DO NOT HARASS HIM OR OTHERS!!), although a read of the Slate article should be enough to show that this was pretty obvious.

All of this comes at the time where Florida’s Surgeon General is arguing against COVID vaccination for men under 40, that bivalent booster reception appears to be quite low (although low access may account for this figure), and with midterms coming next month and political division at a high.

Therefore, this would be a perfect study to put out to really stoke the flames of divisiveness and create a supposed boon for one side of the political aisle , if not given the fact that this study was horrendously flawed.

It lacks any context, and hardly even measures the variables it was set out to measure.

It’s far and away from being “the most definitive and direct evidence” as remarked by Moynihan.

Instead, it’s publication, weaponized by those who didn’t provide a more critical assessment of the study, may be used to push an agenda against those nefarious “anti-vaxxers” of which I and other non-Republicans are being lumped into even though this study would never include us in its dataset.

The evidence here is tenuous at best, and quite questionable at worst. It’s a reminder that data can tell you what you want it to given the right circumstances and parameters.

There’s already enough divisive rhetoric running around. We certainly don’t need more, especially ones that rely on questionable data in order to create such division.

Anyways, let me know your thoughts in the comments below including some criticisms of my assessment.

---

If you enjoyed this post and other works please consider supporting me through a paid Substack subscription or through my Ko-fi. Any bit helps, and it encourages independent creators and journalists outside the mainstream.

Click on the image above or visit the website at https://ko-fi.com/moderndiscontent

1. Wallace, J., Goldsmith-Pinkham, P., & Schwartz, J. L. (n.d.). Excess death rates for Republicans and Democrats during the COVID-19 Pandemic. Retrieved from https://www.nber.org/system/files/working_papers/w30512/w30512.pdf

2. Note: I am not saying the above writer for Slate is doing just that.

3. The researchers note that this is how they measured party affiliation:

This data was linked at the individual level to 2017 Florida and Ohio voter registration files [15]. For each record, the linked data include month of death, age of deceased, county of residence, and 2017 political party registration. Political party affiliation in Florida is actively registered by voters. Political party affiliation in Ohio is defined by whether they voted in a party’s primary election within the preceding two calendar years.

Note the discrepancy even between how Ohio and Florida measures party affiliation. Also, this study didn’t include any independents so this could be missing out on a large number of voters.

4. Although the information referenced does not specifically state vaccines, it does show a partisan gap in coronavirus as a top issue between Republicans and Democrats.

5. One can argue that using pre-COVID data could allow for consistency in the time before and the time after, and that voter registration data may not be collected routinely, but these would be factors that should be included in assessment of the data.