Young Unaffiliated Voters Aren’t Closeted Democrats

Recently I’ve been doing a bit of a deep-dive into the North Carolina voter file. Specifically, I wanted to look at how people’s party registration (which here acts as a proxy for voter preference, the thing we really care about) varies by geography, race and age. What I found was that young voters in North Carolina were not as Democrat as I had perhaps thought (or hoped).

This chart shows the % Democrat (two-way registration, so ignoring people who aren’t registered with either party) of white North Carolina voters by age.

Some patterns are familiar; the super-old are pretty Democrat thanks to North Carolina’s old Solid South days, the most Republican are around 55 (people who came of age in the Reagan years) and then you see people getting steadily more Democrat as they get younger. But then, there’s a sudden reversal around age 30 and now you have whites aged 20 being one of the most Republican age groups in the state. What gives??

A lot of people have very strong priors that young people are Democrat and getting more so. When I posted these charts on Twitter, I got a lot of replies saying this must be wrong or misleading somehow. The leading theory was that this data doesn’t include Unaffiliated voters, and young people who register as Unaffiliated are generally pretty liberal. I think this is in part drawn from the popular idea of the Bernie supporter as a 20-something who will reliably vote Democratic when they have to but is dissatisfied with the party overall and so won’t register as a Democrat. What this misses of course, is that the exact same kind of person exists on the Republican side too. Disillusionment with the major parties and willingness to register as Unaffiliated is not a feature of young liberals, it’s a feature of young people in general.

However I didn’t have a robust methodology to see how true this was, so I decided to build a model to test it. The basic idea is that we have a bunch of other variables in the voter file in addition to party registration that we can use to infer a voter’s affiliation. I test this on voters aged 18-34 who did choose Democrat or Republican, and then I use it to predict a party affiliation for those who chose Unaffiliated.

The most useful model type here is logistic regression, since we’re predicting a binary variable (D/R) but I personally have an easier time interpreting linear regression coefficients and they give pretty similar results overall, so here’s the linear regression output with the caveat that it’s not the most technically useful output:

Some variables explanation:

in_state = born in NC

lib_state = born in a Biden +10 or more state

overseas = born overseas

con_state = born in a Trump +10 or more state

lib_county = lives in a county that voted for gay marriage in 2012 (Mecklenburg, Wake, Chatham, Orange, Durham, Buncombe, Watauga)

dem_county_reg = the two-way party registration of that county among voters aged 18-34

drivers = whether or not the voter has a driver’s license on the file

There’s some pretty interesting stuff here , for example the fact that young people born in-state are likelier to be Republicans than young people who’ve moved from a more conservative state. This isn’t too surprising, I imagine that young people who move states generally skew liberal. We also see that having a driver’s license is a pretty strong negative predictor of being Democrat – possibly a proxy for wealth?

Now when I take this model (or rather the logistic version of this model) and use it to predict party affiliation for the young voters who didn’t choose a party, I get an average of 52.2% Democrat. Does this mean young Unaffiliated voters skew liberal? After all, that’s a Dem-leaning group. Not exactly. Voters aged 18-34 who chose either Democrat or Republican are actually 55.4% Democrat, so the unaffiliated voters are (modeled) 3.2% more Republican, or 6.4% in margin. So if anything, young unaffiliated voters skew conservative relative to their age group!

Some limitations of course are that this model doesn’t have an amazing R-squared (0.369) and there’s certainly more variables I could look at, such as density of the zip code the voters live in. But the upside is that this analysis covers literally every active registered voter in North Carolina which means we don’t have sampling error or anything. This is everyone.

Why am I not surprised by this? Well for one thing, Unaffiliated voters skew white. Also I think people generally conflated these two ideas: “Bernie’s primary support in 2016 and 2020 came largely from young people, many of whom are Unaffiliated” and “young Unaffiliated voters are generally Bernie supporters”.

Does this mean really young people (18-24) are actually conservative? Not exactly. Nationally I think there’s some evidence that the 18-24 group are more Republican than the 25-34 group, while still being more Democrat than virtually every other age group. Here’s the same graph as above, but for all races:

Yeah the trend line among the under-30s is not good, but they’re still likelier to register as Democrats than Republicans. In a state that is currently 6 points to the right of the country … I wouldn’t be freaking out if I were the Democratic Party.

Advertisement