Narratives

By James Kwak

[Updated to add another headline leading with “white voters.”]

Two days later, some of the world’s leading newspapers—or their headline-writers, at least—are saying it was all or largely about race:

The respective roles of race and class in this year’s election are a highly contentious issue. I’d like to add to that contentiousness as little as possible while pointing out that this race-based framing isn’t really supported by exit poll data. I want to get ahead of the vitriol by stipulating that the exit polls don’t provide conclusive evidence for either side.

OK, here’s the data:

 

Those are vote shares in the presidential election by racial or ethnic group. The numbers at the right show you the shift from the previous election.* In this case, the Democratic-Republican gap among white voters shifted by 8 points toward the Republican. That’s evidence that the election was about white voters, right?

Except those are the 2012 exit polls. The 8-point shift is relative to the 2008 exit polls.

Here’s the equivalent chart for 2016:

As you can see from the right-hand column, Trump did better than Romney among every racial or ethnic group. In fact, if you subtract off how he did among all voters (2 points better than Romney), his performance among whites relative to his overall performance was 1 point worse than Romney’s.

What about income? This is 2016:

There are two factual statements you can make about this picture. One is that Trump lost the “working class” (under $50,000) vote. You will hear a lot of people make that statement. The other is that he did much, much better among the working class than Romney: about 11 points better (the

Looking at these pictures alone, at first glance, the story seems to be more about class than race. In politics, change happens at the margin. Trump is still not the candidate of the working class—Clinton is—but he was able to appeal to them much more successfully than Romney or McCain. As for whites, they have voted for the Republican in every election since at least 1968, and Trump didn’t expand that advantage significantly over where it stood in 2012.

But as I said earlier, I don’t think you can necessarily infer from the exit polls that class was the dominant factor and race was less significant. The problem is that we know a larger proportion of working class people voted for Trump than for Romney, but we don’t know why they voted for Trump, at least not from the data we have. We can make some guesses, but again the exit polls provide support for both stories:

(How each of these two questions provides support for a different story is left as an exercise for the reader.)

At the end of the day, we know that the “white working class” supported Trump much more strongly than it supported Romney, but we can’t tell from polling data if that was because of their judgments about Trump’s policies, their feelings about race, or their feelings about their economic status. In practice, different people in the same demographic group make political choices based on different combinations of those (and other) factors.

I think it’s important to try to understand the relative importance and the interactions of these different motivations, and how those have shifted over time. But if there’s one thing I want you take away, it’s that you can’t answer these questions by looking at aggregate polling data—even though many people will try to do exactly that in the next few days.

* Thanks to the Times for this presentation, and for the ability to switch easily from election to election. When anyone cites a poll number, your first question should be: “Relative to what?” In this case, I think the previous election is the most obvious baseline for interpretive purposes, although it certainly isn’t perfect.

** In 2012, Romney won the >$200K group by 10 points, while Trump won it by 1–2 points; you don’t see that shift in the picture because the 2012 data don’t have a break at $250K.