Photo credit: MikeEdwards/Depositphotos
The American political system is polarized--increasingly so in the last few decades. Such polarization has gridlocked national politics in the past few years. In contrast, U.S. states and counties are passing legislation, and increasingly, this legislation is reflecting polarized local politics.
A new study decided to look at whether these local political differences were linked to actual health outcomes. The researchers showed that county by county, the average life expectancy in a particular U.S. county is associated with that county's presidential voting patterns. Specifically, the study found that
how a state votes in presidential elections helps predict life expectancy for people living in that [county].
The study was summarized by journalists at NPR, and the story takes the form of a conversation between two journalists. It reported:
...political affiliation is one factor shaping the life span of people around the country according to this new study. Researchers went back about 20 years. They analyzed a CDC database that collects death data on, essentially, every individual in the U.S. They linked this data with federal election data, looking specifically at how counties voted in presidential elections and also in state governor's races.
a) So far, the journalist has mentioned two major variables in this study. What are they?
b) Were these two variables measured or manipulated?
Now here is some more detail:
Now, overall, mortality rates in the U.S. had been declining in the early 2000s. People were generally living longer. But the researchers found some of those gains have faded away. And when they looked at areas losing ground, they found Republican counties were losing more. Bottom line, people in Republican-leaning counties appear to be more likely to die prematurely.
In many correlational studies in psychology, the person is the unit of analysis (the unit of focus). But in this study, the county is the unit of analysis. So for each county, the researchers recorded the death rates and the voting patterns.
c) Take a moment to sketch a scatterplot of the relationship described above. Put "average age at death" on one axis and "percentage of votes for Republicans" on the other. Should your scatterplot show a positive or a negative relationship?
d) What does one dot in your scatterplot represent?
The researchers used regression to rule out several possible alternative explanations for this relationship, as you'll read about in this next passage:
Now, it's not something as simple as Republican counties being older. The researchers adjusted for age. Nor is it just an urban-rural divide in the country. They found the mortality gap held up in suburban and urban counties that vote Republican, too.
Note that you can see the results of these analyses in the empirical journal article.
e) What were the variables they controlled for in their analysis?
The rest of the journalist's discussion is about mediators--they discuss WHY there might be a link between Republican voting and death rates. Here's one example:
[researcher Dr. Steven Woolf argues that] if you look at policies such as the expansion of Medicaid, access to health care ... tobacco control, gun legislation, drug addiction - a whole range of policies have an impact on health and mortality rates. Democratic states have supported more of these. But Republican states have gone the other direction.
Researcher Woolf mentioned the pandemic, as well:
States with Republican leaders were pushing back on COVID-19 vaccination and enforcement of public health policies. And what we basically had was a controlled experiment where, you know, some states adopted this proactive approach, other states didn't. And we had an outcome that could be measured within weeks. And we saw massive differences in death rates.
f) After reading the passage above, you should be able to sketch the different mediators they are discussing, using the following template:
Republican voting pattern. ---> __________ --> Higher death rate
g) What other potential mediators could you propose for this link?
h) Are there potential third variables that you think the researchers might wish to control for in their regression analysis?