Never out of style in U.S. America

The New York Times's "We're Not Nate Silver" column, the Upshot, ran some interesting demographic noodling Thursday, noting that some demographic assumptions about the U.S. electorate may actually be kind of wrong, which could maybe, just maybe, mean that Donald Trump could find a way to become president without winning any significant portion of the black or Hispanic vote. Not that anyone thinks it's a good idea, just that it may be mathematically possible:

One of the biggest reasons Donald Trump is considered to be a long shot to win the presidency is the diversity of the country.

As Joe Scarborough of MSNBC put it, “There are not enough white voters in America for Donald Trump to win while getting routed among minorities.”

But a growing body of evidence suggests that there is still a path, albeit a narrow one, for Mr. Trump to win without gains among nonwhite voters.

Here's the statistical dealio: the Times's Nate Cohn and pals figured out that exit polls in 2012 didn't accurately reflect the actual number of "white, older working-class voters" who might be likely to vote for a Trump, and that there are actually "millions more" of them than you'd know just from looking at exit polls. That's because while exit polls are fairly reliable at getting information on why people voted as they did, they're not an especially good measure of the actual demographic composition of the electorate:

“There are campaigns and journalists who take the exit polls as the word of God about the shape of the electorate and their voting propensities,” said Michael McDonald, a political scientist at the University of Florida who researches voter turnout. “They’re meant to tell us why people voted. They’re not designed to tell us much about the demographic profile of the electorate.”

The exit polls have a series of subtle biases that depict a younger, better-educated and more diverse electorate. Mr. McDonald tentatively reached this conclusion in 2005, and the pattern has been seen in a broader set of data.

Survey Nerds will want to look a lot more closely at why that is, but we promised ourselves we wouldn't get bogged down in talking about sampling methodology. Cohn identifies two sources for better demographic estimates:

The first -- and longest-standing -- source of alternative data is the Current Population Survey, known as the C.P.S. Conducted by the Census Bureau, it is the same monthly survey that yields the unemployment report. After elections, it includes a question about whether people voted.

A second source is the so-called voter file: a compilation of local records on every American who has registered to vote, including address, age and whether the person voted in a given election.

Cohn went with data from a Democratic number-crunching firm, Catalist, whose numbers have been found to be "unbiased and more accurate than public voting records." Looking at these two sources, Cohn lays out the exit polls' misunderestimate of the older, whiter, poorly educated people Donald Trump loves so much:

White People Chart via New York Times

So if you compare the most generous estimate from Catalist with the figures from the exit polls, you come up with the exit polls missing some 10 million more voters who are white, over 45, and without a college degree. This might explain some other polling surprises at this stage in the election, says Cohn. A "larger pool of potential voters than generally believed" could explain why Trump "is competitive in early general election surveys against Hillary Clinton."

But this analysis could still turn out to be good news for Hillary Clinton, since the numbers also tend to refute the conventional wisdom that in 2012,

Barack Obama did very poorly among whites and won only because young and minority voters turned out in record numbers. This story line led Republicans to conclude that they had maximized their support from white voters and needed to reach out to Hispanics to win in 2016.

If the white electorate was bigger than assumed, then Obama was probably "not as weak among white voters as typically believed," which also implies Donald Trump wouldn't necessarily have a lock on the less-educated older whiteys either. So while there may be more possibly Trump-ready voters, Cohn reminds us that demographics aren't destiny:

The lower turnout among Hispanic and young voters implies that it’s possible -- even easy -- to imagine a huge increase in Hispanic and youth turnout in 2016. And Mr. Obama’s strength among Northern white voters raises doubts about whether the Republicans, including Mr. Trump, can assume that white working-class voters are receptive to conservative candidates.

For Trump to actually win, he'd have to get much more of the older-whiter-less educated white vote than Mitt Romney did in 2012, when the Republicans thought they'd already maxed out their share of the white vote. So far, the early (and really unreliable this early) polling seems to show Trump doing that, but it's unclear whether he can keep up his popularity with that demographic in swing states; also, if he loses more support from other groups, like educated white voters, he's likely to be toast. As this week's Republican panic about Trump's open racism suggests, that could very well be the direction he's headed. Ultimately, Cohn concludes, it looks likely that Trump will "lose too much ground among well-educated and nonwhite voters to win. But the diversity of the country in itself does not rule out a victory for Mr. Trump." So it's an interesting bit of mathematical noodling, but all in all, there's good reason to hold off on buying that plane ticket to Canada just yet.

[NYT]