Emily Badger, a reporter at the New York Times, wrote an article for The Upshot called What’s Your Ideal Community? The Answer Is Political. In it, she dug into a survey from Pew Research Center about political differences in American’s preferences over the type of community where they prefer to live. Conservatives prefer rural areas whereas liberals prefer the city.
In this post, we look at how political differences in community preferences vary by age group. We find that older adults are less politically polarized about where they prefer to live, especially compared to Americans aged 30-49.
Here’s what we found:
We looked at the percentage of Americans in each age group and for each party-leaning who preferred to live in one of four types of community: city, suburb, small town, or rural area. Sample sizes get quite small (less than 500) once you slice the data by both age group and party leaning. For this reason, we used some fancy statistical methods (see the methods section at the end) to estimate the margin of error.
Political polarization is strongest for the 30-49 age group. In this group, community preferences are the exact opposite for each party. Democratic-leaning Americans aged 30-49 prefer more urban settings, whereas Republican-leaning Americans of the same age prefer more rural areas.
Above, we estimated the percentage of Americans who preferred a given type of community for each age group and party leaning. To understand political polarization in community preferences, we must dig even deeper. One way to measure polarization in community preferences is to take the absolute difference between political leanings in the percent of Americans preferring a given type of community. The greater the absolute difference, the more highly polarized Americans of a given age are about that type of community. It turns out that Americans across age groups are most politically polarized about cities and rural areas.
In the last section, we looked at political polarization by community type. Yet to understand how politically polarized an age group is as a whole, we need to take the average polarization across community types for each age group. Adults age 65 and older are the least polarized group on average. Due to sample size constraints, we’re most certain about differences in polarization between older adults (less polarized) and those aged 30-49 (more polarized). Yet looking back at the last chart that shows polarization by community type, we’re most certain of higher polarization about cities when comparing the 30-49 year age group to older adults.
Political polarization differs by age group, at least when it comes to the type of community where you prefer to live. Today, older adults are slightly less polarized than other age groups, especially those aged 30-49, and especially when it comes to cities. Are these differences generational (i.e., based on the year you were born), or does the type of community you prefer change as you age? Both birth year and age-related changes are likely important.
If age-related changes are more important, then given these patterns we’d expect younger generations to become less polarized as they age. Yet if older Millennials and Gen Xers as a birth cohort are more politically polarized than their grandparents, and if they will stay that way, what does it say about how older adults will choose where to live in the future?
Raw data on community preference by age and political leaning came from Table 3.2 of the Pew Research Center survey analysis. To estimate population-level preferences from this data, we first extrapolated the number of respondents for each community type, age group, and political leaning by multiplying the reported percentage by the reported total sample size for that age and political leaning. We then used independent Dirichlet-multinomial models to estimate the posterior distribution of the community-type-specific percentage, one model for each age group and political leaning. The Dirichlet-multinomial models were approximated using 100,000 draws from the posterior distribution under a uniform prior (as implemented in the rdirichlet function of the gtools package for R, the statistical computing environment). We used these approximate posteriors to compute all figures in the plots.