2016 exit polls: how to read early results like an expert

The first wave of exit poll results has been released — with coverage by NBC News, ABC News, and Politico, among others — and an information-starved electorate is picking over it, trying to find significant information.

But it’s very easy to go astray when looking at exit polls — especially early ones.

The exit polls are designed to allow networks to call races in particular states as quickly as possible. So they’re going to be at their most reliable when they’re projecting whether Hillary Clinton or Donald Trump will win, say, Missouri (and who will win the Senate race there). But those results won’t be released until polls close in each state.

What we’re left with in the early exit poll, instead, are preliminary results purporting to show who voted across the US and why. That’s not the exit polls’ strongest suit.

If you’re using the national exit polls — especially early on — to divine what the electorate looked like in 2016, whom individual demographics voted for, or what factors were most important as voters selected candidates, you might want to slow your roll. But when it comes to poll results that hint at voters pre-election polls might have missed — or show divides between supporters of each candidate — the early exits might help you divine the national mood.

When the early exits contradict the pre-election polling, it might not always be the exit poll that got it wrong. Pre-election polling might underestimate how many low-propensity voters will end up making it to the polls — and given that preliminary reports indicate that turnout might be up in several key states like Florida, that could mean that the pre-election polling had a particular blind spot. (Conversely, if a demographic doesn’t show up to vote in expected numbers, it can also skew a pre-election poll.)

In 2016, for example, pre-election polling found that voters saw Trump as slightly more “honest and trustworthy” than Clinton. But early exit polls suggest that while neither is seen as all that honest, Clinton beats Trump: 38 percent of voters see her as honest and trustworthy; only 32 percent see him the same way.

That could be a problem with the early exits — or it could mean that pre-election polling missed a lot of Clinton supporters who saw her as trustworthy and came out to the polls.

Seventy-two percent of Clinton supporters are scared of what Trump would do in office; 60 percent of Trump supporters are scared of what Clinton would do in office.

Sixty-eight percent of Clinton supporters are “very confident” in the accuracy of the vote count; only 28 percent of Trump supporters are.

Answers like this — that show a huge divide between how supporters of each candidate see the election and the country — aren’t things that vary systematically by demographic groups that might be better or less well-represented in early exit polling. So while these questions tell us very little about who’s going to win, they also aren’t likely to change as the night goes on — or to be wildly misrepresentative of the electorate as a whole. (If anything, early exit polls can understate the divisions between candidates’ supporters because they’re looking at moderate, “bellwether” areas.)

The early exit polls show that Clinton and Trump are tied among white voters with a college degree, according to MSNBC. Latino voters, meanwhile, appear to be supporting Clinton in lower numbers than they supported Barack Obama in 2012 (71 percent in 2012 to 65 percent now) — while the same number of Latinos are supporting Donald Trump, 27 percent, as supported Mitt Romney in 2012.

Both of these numbers are slightly out of line with where each demographic group has been polling: Clinton has been slightly ahead among college-educated white voters in most polls, while Trump is polling behind Romney.

The exit poll findings might shift as more results come in: In 2012, for example, initial exit polls reported that Obama was leading Romney 69 percent to 30 percent among Latinos, a margin 5 points smaller than his ultimate lead.

But even after the final numbers come in, the exit polls might not do a great job of assessing these votes. That’s because exit polls concentrate on bellwether polling places, which are liable to swing from one candidate to the other depending on who’s winning — not safe red or blue precincts where the question is how much a winning candidate will run up the vote.

As a result, the kinds of, say, Latino voters that the exit polls will interview might not be representative of the bulk of Latino voters. And the exit polls might be more likely to interview suburban white voters with college degrees in swing districts than well-educated white urbanites — who might plausibly be more likely to vote for Clinton.

One big reason that divides between Clinton supporters and Trump supporters are likely to persist as the night goes on: Each side is picking up on the message its campaign is sending.

Unsurprisingly, early exit polls say that neither Clinton nor Trump is particularly well-liked: 54 percent of respondents have an unfavorable view of Clinton, while 61 percent of respondents see Trump unfavorably.

But ask voters what they actually like to see in a candidate, and you’re just as likely to hear them parrot things the campaigns have been telling them are important.

Donald Trump is running as a change candidate; unsurprisingly, his supporters overwhelmingly say they want a president who will change things. Hillary Clinton is running as a competent and experienced leader; unsurprisingly, that’s what her supporters say they want. Values such as “cares about people like me” — an attribute that’s been important in past presidential elections — didn’t really play a role in this one, and accordingly, only 15 percent of early exit poll respondents said it was the most important factor in deciding their vote.

We have no idea how many people are still waiting to vote — either waiting in line in states where polls are closing soon or simply planning to vote later today on the West Coast. Early exit polls are skewed in favor of people who vote earlier in the day. And those tend to be older voters: 19 percent of voters in early exit polling in 2012 were 65 or older, but by the time the final results came in, that group only accounted for 16 percent of voters.

The effect this has on other issues might be negligible. But when thinking about who’s going to win the race, pundits often look at demographic groups with which one candidate is particularly strong — like Trump with white voters without a college degree, or Clinton with black voters — and see how large a percentage of the electorate that group makes up. That is going to shift throughout the night.