On age and vote shares

Bringing some data to a vibes fight

Since the “Biden is old therefore Biden can’t win” discourse doesn’t seem to have abated much lately, I thought I’d check to see if we had much evidence that being old actually hurts candidate vote shares. And it turns out we have a rather old legislature in Washington, DC with very high reelection rates, so that might be a good place to start.

Now, I’ll note that there are a number of survey-based studies showing that, all things being equal, voters prefer younger (under 70) candidates, although this effect is mitigated when two older candidates are close in age. While this is an important set of findings, all things are not usually equal. A Republican-leaning voter, for example, confronted with a choice between a 45-year old Democrat and a 75-year old Republican, may prefer a younger candidate in his heart of hearts, but will still vote for the elderly Republican ten times out of ten.

To get a sense of how age and vote shares interact, I obtained data on the current U.S. House of Representatives. I was interested in each member’s age, party affiliation, and vote share in 2022 (specifically, their share of the two-party vote). Just plotting out those two important variables gives us the scatterplot below. Each red dot is a Republican House member, and each blue dot a Democratic House member. The trend line is a moving average, giving us an idea of how votes shares change with age.

Officeholder Age and Vote Share in the 118th US House

Okay, a couple of important things to see here:

• The trend line moves upward with age. That is, older officeholders tend to have somewhat higher vote shares. That doesn’t trend downward or flatten as the officeholders surpass 70 or even 80.

• There’s an interesting gap in the data on the lower right, where there just aren’t many old officeholders winning within less than 55% of the vote. My read on this is that those are very competitive districts, and those are not districts one grows old in; they tend to flip a fair amount and toss people into the private sector.

• There are a number of officeholders across the top of graph who were unopposed in 2022, and that seems to occur at a variety of ages.

This isn’t the most rigorous analysis so far, but one thing it doesn’t show is that voters turn against officeholders as they age. Yes, it’s possible that the most vulnerable elderly House members are nudged into retirement by their parties, but we know that parties very rarely push out their incumbents. Indeed, there are 11 House members in their 80s in the graph above.

Okay, let’s add a bit more data here. Below is a regression analysis that predicts the share of the two-party vote for each House member. I’ve included their age, their party (“dem”, equalling one if they’re a Democrat and zero if a Republican), the extremism of their district (“dist_ext”, measured by how far the presidential vote share in their district was from 50%), and the extremism of their House voting behavior (“mem_ext”, measured by the absolute value of their DW-NOMINATE score). I’m posting the data here if you’d like to tinker. (A few districts had more than one member over the past two years, which is why the number of cases is above 435.)

So there are a few things we learn from this. One is that Democrats tend to have somewhat lower vote shares (about six points) on average, while members from more extreme districts tend, unsurprisingly, to have somewhat higher vote shares. The ideological extremism of a member’s own voting record doesn’t seem to be related to their share of the vote. And, importantly, age is also unrelated to the vote. Age has a slightly positive correlation here, but it is far below statistical significance.

I ran the same regression but omitting those who ran unopposed. In this case, the age coefficient (0.03) is positive and near statistical significance (p = .088), suggesting that each additional year is associated with three hundredths of an additional percent of the vote. That’s not much.

This, again, isn’t the most rigorous look at this question, and it would be helpful to look at the challengers that these officeholders beat. But if being older definitely turned voters against officeholders, we should see some evidence of it in the U.S. House. We don’t.

Update (2/25/24)

A few readers correctly noted that the age variable here may be capturing some of the effects of incumbency. I’ve added a variable “frosh” to the dataset that equals 1 if the member was a freshman in 2023 and zero otherwise. And when I add that to the regression equation, it shows us that freshmen tend to win by about 4 points less than veterans do, and it also washes out any effect of the “age” variable. So, age still has no negative effect, but also no positive one.