Wednesday, October 12, 2016

How do gender *and* age shape the electoral college?

Yesterday, Nate Silver posted a widely-disseminated article aspiring to show how the 2016 election would play out if only men or women voted. Predictably, the election swings deep blue when only women vote, and bright red when only men vote. Out of curiosity, I used Reuter's new States of the Nation tool to take a quick look at this year's age gap, and while the results were similarly predictable, this approach encounters three significant problems. First, the tool only allows one to drill down to the 18-30 year old bracket, which excludes Millennials ages 31-34. Second, a lot of the sample sizes are too small to produce reliable data, which means that even states like South Carolina (with 9 electoral votes) can't be called. And third, 538 used a completely different methodology to produce its maps:
Here’s a quick way to estimate it. In the polls I cited above, Clinton is doing 10 points better among women than among the electorate overall. So we’ll add 10 points to her current polls-only margin in every state to forecast her performance if women were the only ones who could vote.
This is a pretty fast-and-loose approach that I don't have much confidence in, but it at least makes it easy to make some apples-to-apples comparisons. So relying on the 538 approach, here's what the election looks like if only Millennials voted:




In a youth election, Trump is held to single digit electoral votes, and Clinton wins every state except two: Wyoming, and Nebraska's third district. Compare this to what happens if no Millennials vote:


Here, Clinton still manages a win, but Arizona, Iowa, North Carolina, and Maine's second district all swing into Trump's column. More likely, this could threaten Democratic hopes of capturing the Senate by allowing Republicans to keep control of seats in Indiana and Wisconsin - both states Democrats hope to flip.

The conclusions here are pretty straightforward:
  • In the electoral college, at least, the gender gap is bigger than the age gap: there is a 270 vote swing from men to women, compared to a 213 vote swing from olds to millennials.
  • These gaps create significantly different outcomes, however, because they're centered at different points: women are more reactionary than millennials, and men are more reactionary than olds.
These points raise an interesting intersectional question: how do age and gender combine to influence electoral outcomes? To answer this, the first thing we need to do is break down the vote margins for every combination of old and young men and women, which gives us a table like this:


Already, it should be completely obvious that age ends up playing a far more important role in predicting preference than gender. Trump has a slight advantage among old men, and because of the sheer number of olds voting, it's easy to look at the electoral college and conclude that this pattern holds for all age groups. But in fact, not only does this pattern not hold among millennials - it actually reverses, so that Clinton has an 8.4 point advantage among millennial men compared to millennial women. Even among millennial women, however, Clinton's advantage is astronomically larger than it is among old women, who actually do .2 points worse than the national average.

If we look at how these differences would play out in the electoral college, it's clear where Trump's support comes from and where Clinton's support comes from. Among older women, Clinton would actually lose an electoral vote in Maine's second district. And among young men, Clinton wins every electoral vote except one, in Nebraska's third district. It's youth that most powerfully predicts support for Clinton, with gender proving unreliable at best.