Horse race update

Edgar Degas, Race Horses, 1885-88, Metropolitan Museum of Art.

I never got around to explaining what it was that impressed me so much about Rachel Bitecofer's work, in the interest of talking about the specific predictions she was making,  and I don't want to let it go, because the more I think about it the more important I think it is: it's that she isn't really making predictions in the normal horse race fashion at all, which aims at anticipating the result at all costs, but deliberately making predictions that could easily be wrong.

She's making falsifiable predictions, in fact, in the service of testing a hypothesis, so that there's something to learn from it. When we follow a fivethirtyeight.com prediction through the election, we learn how smart Nate is, and if we're gamblers we may collect some money, but we don't learn anything in particular about how the political system is working. When we follow Bitecofer's prediction, say that Democrats would pick up 43 House seats in the 2018 midterm, we learn whether her particular hypothesis is consistent with reality or not. Bitecofer is putting some science into political science.

The hypothesis is basically that the old arrangement in which elections were decided by the independents no longer applies; it's the polarization that has followed the sorting of liberals and conservatives into the Democratic and Republican parties, and the different degrees of emotional engagement that accompany it:

Although partisanship has always been an important driver of electoral behavior, the influence of partisanship on vote choice is immense in the polarized era. What matters most to the vote decision is party identification. For most voters, in most places, and in most elections, even judicial elections, this consideration overrides all others. Despite the rise of self-identified Independents over the past few decades split-ticket balloting, which refers to the decision a voter makes to vote for a presidential candidate of one party and at least one congressional candidate of another, as well as the number of states where the winning presidential candidate is from the opposite party of the winning Senate candidate, has collapsed since the mid 20th century. In 2016, 34 out of 34 states chose the same party for president and Senate and in 2016 only 35 House districts of the 435 total districts voted for a presidential candidate of one party and a House candidate of the other.

So she builds a remarkably simple model based on the assumption that degrees of Democratic vs. Republican partisanship will be decisive, as measured by the Cook Partisan Voter Index (PVI) score for each congressional district, alongside the factors of proportion of the population with college degrees, difference between Democratic and Republican campaign spending, and presence vs. absence of a Republican incumbent (racial diversity and urbanization are also implicitly considered, since they correlate strongly with PVI—i.e., more racially diverse and urbanized districts are generally more partisan on the Democratic side). When the prediction made by these factors comes out accurate, that corroborates the hypothesis that partisan attachment, not the behavior of independents, is driving the voting behavior.

As you could have observed without all the math in a more informal way, if you were looking for it—

Although Barack Obama won the majority of Independents in his 2008 presidential race (primarily because the economy was quite literally collapsing on Election Day) he did not win the majority of Independents in his 2012 reelection bid. Given the conventional wisdom of elections, such a thing should not be possible. And it’s not just that he failed to carry Independents nationally, he failed to carry Independents in critical swing states such as Ohio that he still won. In fact, Obama lost Independents in that decisive swing state by a staggering 10 points, but he still won the state because the impressive turnout operation established by the Obama campaign managed to produce an electorate that was 38% Democrat. And as I show in my unfortunately titled book The Unprecedented 2016 Presidential Election, Democrats lose Independents quite often, and in elections they win and they lose because they have a population advantage in many places and when their partisans turn out in high numbers, it trumps the combined loss of Republicans and Independents...

—but you wouldn't recognize how systematic it was.

Where this intersects with my preoccupations is in my endless irritation with the concept of a "white working class" represented by people who eat breakfast in diners, as a decisive factor in recent elections and something we must be constantly on the lookout for, as in the Biden campaign. I can't believe that a group whose partisanship is so weak they can vote for Obama in 2012 and Trump in 2016 is a significant factor in the current climate, and Bitecofer gives me scientific backing.

Martin, in a Twitter discussion of the Trump "base", linked to an old piece of his, from the 2016 campaign, that I hadn't really noticed at the time, of some interesting relevance to that post and also to my old idea that the "Obama-Trump voters" of 2016 were something of a bogus category; where Boo had been arguing, in point of fact, that the interesting thing about 2012 was some gigantic number up to six million of white people who didn't vote at all in 2012:

Just realized those missing white voters were the ones who falsely told 2016 exit polls they'd voted for Obama in 2012 and gave rise to the "Obama-to-Trump" myth.

— Purview From the Bridge (@Yastreblyansky) August 3, 2019

Source?

— Boswood (@Bosengood) August 3, 2019

The source was my own piece from August 2017, with data sources linked.

And Phillips, cited in my diary, but he doesn’t link to his sources and a commenter said there were 9% O to T voters. (Rebutted apparently by Bitecofer, which I have to read, based on liars). But why lie about voting for Romney? https://t.co/mWpfPrLB9W

— Boswood (@Bosengood) August 3, 2019

And I shouldn't have let it be said that Bitecofer was accusing anybody of lying, that was entirely my idea. I only meant her analysis (dismissing the importance of the Obama-to-Trump voters) was consistent with mine (that they didn't altogether exist as a group). So Martin apparently went and took a look at my old piece and it annoyed him

With respect, how do you do it? How do you establish the number of Obama-to-Trump voters? Do you cross-check against lists of those who didn't vote?

— Purview From the Bridge (@Yastreblyansky) August 3, 2019

And this is what happened in county after county in Pennsylvania and throughout the Rust Belt, and also in other places where it didn’t matter.

— Martin Longman (@BooMan23) August 3, 2019

Except there may have been no relative surge, but you know from other factors that those three million missing white guys were back!

So in those rural counties in PA and OH you don't think the "missing white voters" had ever been missing after all? I can see that, but not in less rural MI and WI, where they clearly returned to the polls while young people and minorities who had voted in 2012 sat it out.

— Purview From the Bridge (@Yastreblyansky) August 4, 2019

Detroit definitely saw turnout drop dramatically. It’s an outlier in that respect. Milwaukee’s vote was significantly suppressed. Clinton got the same net votes out of Philly/Pitt as Obama had and did better in the suburbs. She still turned a comfortable win into a loss.

— Martin Longman (@BooMan23) August 4, 2019

Or Facebook propaganda did, in Detroit and Ann Arbor and Milwaukee and Madison, discouraging voters you would have expected to be energized.