Trump Losing College-Educated Whites? He Never Won Them in the First Place

After Donald J. Trump’s upset victory in the 2016 presidential election, one data point from the network exit polls jumped out at analysts: his two-point win among college-educated white voters.

Many pre-election polls had suggested they would favor Hillary Clinton. And now, more than a year later, polls again show Mr. Trump with striking weakness among well-educated white voters, implying he has alienated many who backed him in 2016.

But it is increasingly clear that there’s another explanation: The exit polls were wrong. Mr. Trump never won college-educated white voters, and therefore he hasn’t lost nearly as much ground among the group as it seems.

The exit polls’ overestimate of Mr. Trump’s support among well-educated white voters is symptomatic of the survey’s deeper problem: a tendency to severely underrepresent older, white voters without a college degree. Thanks to an odd methodological quirk in exit polling, this underlying problem ultimately biases the rest of the survey, along with the analysis that follows.

At this point, there is no question that the exit polls underestimate the number of white working-class voters, especially older ones, by a considerable amount. This is clear in census and voter file data, which are widely preferred as measures of the composition of the electorate by pollsters, academics and analysts.

The newest evidence comes from a Pew Research study on voter registration files. To my knowledge, it is the first nationally representative probability sample matched to vote history data from the 2016 election. According to the Pew data, 45 percent of 2016 voters were whites without a college degree — 11 percentage points higher than the 34 percent vote share calculated by the exit poll, which is conducted by Edison Research for a consortium of media organizations.

How could the exit poll be off by so much? In fact, most polls are represented by too many well-educated respondents; most high-quality surveys weight their samples to adequately represent less educated voters. But the exit polls aren’t weighted by education, and so they start and end with far too many well-educated voters. (The exit polls are weighted only by demographic characteristics that the exit poll interviewer can visually identify, like gender, whether someone is African-American, and a rough guess of age; they can’t guess education.)

This winds up biasing the rest of the survey because the exit polls are weighted to match the actual result of a far less educated country. In general, the exit polls underestimate Republican support, probably in no small part because they overrepresent young, nonwhite and well-educated voters. But this process leaves the underlying educational bias of the sample intact, and the result is that Republican-leaning voters are given more weight to compensate for an electorate that represents Democratic-leaning voting groups.

In the end, the exit poll overestimates Republican support among most demographic groups, including well-educated white voters, and it overestimates the number of voters from Democratic-leaning voting blocs, like young, nonwhite and well-educated voters.

Those two facts go a long way toward explaining why there were so many Democratic-friendly analyses of voting demographics and polls over the last decade. The exit polling made it seem as if the Republicans, in a diversifying country, could not win with big gains among older, white working-class voters.

And if the exit polls overestimate Republican vote share among various demographic groups, then pre-election national polls will seem to show Democrats doing well, since they’re weighted by education and aren’t adjusted to match a result. The exit poll tendency to overestimate Republican vote share will also tend to obscure Republican strength in pre-election polls, like Mr. Trump’s big gains among white voters without a college degree.

Over the last year or so, national media organizations have made a concerted effort to cover those white working-class voters. Some have argued that those efforts have gone too far. But so far there has been little or no movement to better represent white working-class voters in public political data and analysis. Only a handful of state pollsters have begun to weight by education, and the exit polls have not announced any methodological changes intended to remedy these issues either.

The consequences can be significant. For better or worse, lawmakers and campaigns make policy decisions based on political calculations, and the exit polls play a huge role in shaping how people judge the landscape of American electoral politics. Even Fox News’s Sean Hannity briefly supported comprehensive immigration reform in November 2012, after the exit polls made it seem as though Hispanic voters, not white working-class Midwesterners, decided the election in President Obama’s favor.

One way to avoid the issue: retire the use of the exit polls in post-election analysis, and make apples-to-apples comparisons between current and older pre-election polls. The most important trends of the 2016 election were more apparent using this approach, including Mr. Trump’s strength among white working-class voters. And today, it would make President Trump seem more resilient among well-educated white voters.

Powered by WPeMatico