What's the Gappiest Gap in Current U.S. Politics?
My nominee for the biggest, cleanest single split in basic voter demographics
Here’s the task. Considering the usual basic demographic stuff – age, race, gender, LGBT, marital status, education, religion, church attendance, urban/rural, income, employment status, union membership, and military service – what’s the best yes/no question you can ask these days to correctly guess the most voters’ selections between Democrats and Republicans? And so things don’t get out-of-control complicated, let’s also stipulate that the question can’t exceed 50 words.
So, for example, let’s say you think that racial minorities, college-educated women, and Millennial and Gen Z women are really at the heart of the demographic split between Democrats and Republicans. We could ask voters the question:
Are you any one (or more) of the following: Black/Hispanic/Asian; a woman with a 4-year college degree; a woman who was born after 1980? ___ Yes ___ No
When I check this against the very large dataset that I’ve stitched together from publicly available data from CES, GSS, ANES, PRRI, and others released in the past few years, limiting the sample to only folks who voted for one of the two major parties in the last two presidential elections, here’s what I find. We’re guessing that those who would answer “yes” to the question above voted for Democrats and those who would answer “no” voted for Republicans. It turns out that these guesses are correct around 65% of the time looking at both the 2020 and 2016 elections.
How can we improve on this question? A nice resource here is the Cooperative Election Study’s (CES) very cool interactive display, which allows you to drill down on various demographic splits in their surveys of the 2024 potential electorate so far. If you spend some time with it, you’ll discover there are a few gaps that stand out more than the others: If you look at the CES vote breakdowns by race, LGBTQ, and religion (among white respondents), you’ll see a gap of around 30 points between Blacks and Whites, a gap of around 32 points between LGBTQ and non-LGBTQ, and gaps around the 30 to 50 point range between Christian groups and non-Christian groups.
In contrast, some widely discussed gaps are real and interesting, but not huge: If you pull up the CES vote breakdowns by age, education, and gender, you’ll see gaps of around 14 points between the youngest and oldest groups, around 13 points between college and non-college folks, and only around 4 points between women and men. You can get bigger gaps by subdividing into smaller and smaller groups and including racial splits, but then you’re headed into specifics that can’t be put simply into two yes/no groups (and can’t be captured in 50 words or fewer).
So, on coming up with our single yes/no question on demographics, the best versions, I think, will rely heavily on race, LGBT, and religion, and will sprinkle in a bit of gender/education/age to the extent there’s room.
OK, without further ado, here’s a question that I think would be hard to beat in predicting the two-party vote (if limited to the demographic features I listed initially, and to no more than 50 words), given the publicly available data on the 2020 and 2016 elections:
Are you any one (or more) of the following: Black, Hispanic, or Asian; lesbian, gay, bisexual, or transgender; atheist, agnostic, Jewish, Buddhist, or some other non-Christian religion; a woman with a graduate degree (e.g., MD, PhD, JD, MA, MBA, et al.)? ___ Yes ___ No
According to the very large dataset, this yes/no question would split the adult citizen population more or less in half and produces correct guesses around 75% of the time for votes in the 2020 and 2016 elections. The graphs below show the voting stats in a format similar to CES’s – the gap here is around 45 points.
Are there ways to improve on this question using only the standard demographic items that I listed up top? Probably, but I’d be very impressed if someone were able to get more than maybe another percent or two while keeping the yes/no question at a maximum of 50 words.
As to why this list so efficiently captures the fundamentals of the current party coalitions – why, that is, the parties have primarily become organized into, on one side, those who are racial minorities, and/or LGBT, and/or non-Christian, and/or professional women, and, on the other side, those who are none of these things – I’ll save that discussion for another day.