clock menu more-arrow no yes mobile

Filed under:

How exit polls work: when they're released, which states they cover, and what they mean

How to interpret an exit poll.


The American public will find out whether Hillary Clinton or Donald Trump has won the 2016 presidential election long before the last vote is counted.

Or rather, we’ll think we know. Because as polls are closing across America, media outlets are releasing exit poll results — predicting who will win a state, and providing more information about who really turned out to vote and why.

The exit polls will shape the story of the election — they’ll provide the record that people will refer to in the future when they talk about what issues mattered and how our 45th president built a successful coalition.

But to understand how accurate the exit polls really are — and whether you should trust their predictions about who’s going to win — you have to understand how the exit polls are conducted, and why. If you’re a critical consumer of exit poll data, you’re less likely to be duped by bad information on election night, and more likely to understand whether the exit polls are really telling the story of the 2016 election.

How do exit polls work?

Every November election, exit polls are conducted by a group of media outlets called the National Election Pool: NBC, CBS, ABC, Fox, CNN, and the Associated Press. They hire a pollster to conduct the exit poll, but they're the ones that own the information — and that get to be the first to report the results.

hillary election party 2008 sad
Exit polls can really ruin the mood at an election night party. (This one was in 2008.)
Joe Raedle/Getty

The actual polling happens in two parts.

The most visible part of the poll happens in person on Election Day. An army of thousands of interviewers are sent to hundreds of polling places around the country. Interviewers approach a certain number of voters who are leaving the polling place — the exact fraction surveyed is secret — and ask them to fill out the written exit poll survey. Pollsters estimate they’ll interview about 85,000 people on Election Day.

But part of the exit poll has already happened before Election Day. As early voting has become more popular, it's gotten harder to predict vote totals just by talking to people who vote on Election Day. So for the past several elections, exit pollsters have started calling people and asking if they voted early or absentee — then conducting exit poll interviews by phone. (In 2016, pollsters estimated they’d contact about 16,000 voters this way.)

What can you learn from exit polls?

The primary purpose of the exit poll is to allow TV networks and the AP to project who's won races as soon after the polls close as possible.

That means that in 2016, only 28 states are going to have state exit poll results published. That includes obvious presidential battlegrounds (Colorado, Virginia, Wisconsin), states with key Senate races (Indiana, Missouri), and states that are just really big (New York, California).

If a state is clearly a safe state for either Democrats or Republicans — Massachusetts or Tennessee, for example — exit pollsters still send people to do interviews there, for the purpose of the national poll. But they don’t collect enough interviews to publish reliable poll results.

The exit poll pool cut back its efforts since 2012, when 31 states were surveyed in depth. That could lead to some surprises. Alaska, for example, isn’t being surveyed this year but has been surprisingly tight in presidential polling. But it’s hard to imagine either Hillary Clinton or Donald Trump paving a path to 270 electoral votes without the exit poll identifying the winner.

But the exit poll isn’t just about whom people voted for — that’s why there are interviewers even in safe states. Voters are asked to provide basic demographic information like gender, age, and ethnicity. Furthermore, they're asked some questions about their personal viewpoints and behaviors — like their religion and churchgoing habits — and questions about major issues facing the country.

That means the exit poll data is actually more detailed, in some ways, than the official US Census vote tallies that come out several weeks after the election. It can offer the first hints — and often the most important ones — to what voters thought this election was about. That's very important to pundits as they try to interpret what it means.

Candy Crowley debate
CNN’s Candy Crowley is one of the people who’ll have to do some high-speed interpretation of exit poll results.
John Moore/Getty

In 2004, for example, post-election chatter focused on "values voters." Voters who attended religious services regularly had overwhelmingly voted for George W. Bush. That narrative came out of the exit poll data.

Of course, what voters say is important to them is partly what campaigns have told voters is important — there's political science research suggesting that when a campaign hammers particular issues, those are the issues that the candidate's supporters say are most important to them. But the exit poll is still the best opportunity the national media has, in some ways, to figure out who voted, why, and how.

This is the 21st century. Why aren’t we seeing exit poll results in real time?

The media outlets running the exit poll want to be able to describe who's voting, and whom they're voting for, to the public as early as possible. But they don't want to have any influence on who ends up voting — they don't want anyone deciding not to vote because they've already seen what the exit polls say and they don't think their vote will matter. (There's some evidence that this happened back in 1980, when some outlets projected that Ronald Reagan would win the presidential election before polls closed on the West Coast.)

In some countries, like the United Kingdom, it's actually illegal for any media outlet to report exit poll results before the polls close. In the US, it's not illegal, but there's a binding agreement among the media outlets that run the exit poll that none of them are allowed to leak any results before the polls have closed.

Sometimes, networks slip up. In 2014, for example, Fox News showed early exit poll results in New Hampshire before polls had closed there in a tight Senate race. The network argued that it hadn’t broken the exit poll rules because it technically didn’t show how many respondents had voted for Jeanne Shaheen or Scott Brown — it showed how many people said they would vote for each candidate if the race came down to a runoff. (Other outlets disagreed, but Fox wasn’t kicked out of the consortium that uses the exit poll.)

In 2016, some media outlets are trying to find their way around that agreement: Slate, for example, is using its own election data tool to share turnout projections in real time. But it’s not using the official exit poll results.

So when do exit poll results come out?

Reporters are allowed to see some of the exit poll results as they're being compiled throughout the day, but they're under super-strict security — we're talking no-phones-allowed-in-the-room-where-the-results-are-kept levels of security. And there's a strictly regimented schedule for when exit poll results can get released.

Around 5 pm ET, media outlets are allowed to start reporting what the exit poll says about who turned out to vote — the racial, age, or party breakdown of voters. But these are preliminary results, and they're going to be skewed toward people who voted early in the day. So groups who tend to vote later in the day — like young voters — might be underrepresented in the stats that first get announced.

As soon as polls close in a particular state, media outlets are allowed to project who's going to win elections there, based on the exit poll results. In deep blue or deep red states, where the outcome of the election isn't really in doubt, media outlets don't waste any time projecting winners. So the minute 7 pm hits on the East Coast, for example, you can expect to see CNN and the AP make a bunch of projections at once. In states with closer races, media outlets will often wait to get the final exit poll results (including people who voted right before the polls closed), or wait to see how actual vote tallies stack up when precincts start reporting official vote totals.

Don't exit poll results ever get leaked?

There has never been an actual leaked exit poll in the US. But there have been plenty of hoaxes.

If you see anyone on Facebook sharing "LEAKED EXIT POLL RESULTS" while the polls are still open, be very, very skeptical.

This will be a little tricky in 2016, because Slate’s real-time Votecastr project might confuse people into thinking that exit poll results are being leaked or released throughout the day. As long as you understand what those results really are — estimates of who will win based on turnout projections — you’ll be fine, but don’t mistake them for “leaked” exit poll results.

And definitely don't decide not to vote just because you saw something in a leaked exit poll.

Are exit polls always right?

No. In fact, there are some particular challenges that exit polls have faced for the past several elections that they still haven't found a way to work out:

Early voters. The phone poll for early voters is a relatively new addition to the exit poll— and it’s still a relatively minor one, compared with in-person polling. Early voting itself, meanwhile, has gotten very popular very quickly. In key states like Nevada and Florida, it’s estimated that fewer people will show up to vote on Election Day than showed up during early voting.

The exit poll understands the huge role early voters will play — pollsters estimated to Pew that 35 to 40 percent of all voting will happen early this year — but it’s not clear that their polling can accurately capture who those people are. It runs into the problems any phone poll has — namely, that it's difficult to poll people who only have mobile phones. And because this year saw such a huge surge in early voting, it’s hard to use past years to predict how representative a sample is.

Wendy Davis, after voting early on October 20, 2014.
No exit pollster here!

It’s unlikely that any state is going to get called the wrong way because the exit polls didn’t include enough early voters there. Networks are aware of the early voting data, and if, say, the exit polls suggest Donald Trump will eke out a narrow victory in Nevada, networks will probably wait until some of the votes are counted to see whether Trump was really able to surmount Hillary Clinton’s early voting lead there. But the demographic and other data the exit poll provides might be skewed in favor of people who voted in person — who might not be the voters who decided this election.

Small groups. Like any poll, the smaller a sample size is, the less likely it is to be representative. So the exit poll is pretty reliable when it comes to large demographics (men, women, Democrats, Republicans) but less reliable when it gets to small demographics (young voters, Jewish voters).

Voters of color. In addition to the general problems with smaller voting demographics, analysts believe the exit poll has a tendency to oversample a particular kind of voter of color — the kind who lives in majority-white areas.

Here's the logic. Even though the public doesn't know exactly how the exit poll chooses where to go, it's possible to make some educated guesses. The exit poll is trying to predict the margin of victory for one candidate over another across the state. So when it decides which polling places to put interviewers outside of, it's reasonable to assume that it's choosing lots of swing precincts — precincts that are harder to predict and likely to affect the outcome. Those are going to be largely white precincts.

Alternatively, says Matt Barreto of Latino Decisions, exit pollsters might choose a precinct as a benchmark based on the last cycle. For example, if a precinct voted for the Democratic senator 70 percent to 30 percent in 2008, the pollster might choose to put an exit poll interviewer at that precinct to see if the Democrat is getting less than 70 percent of the vote this time around. But pollsters are not necessarily paying attention to the racial makeup of those precincts.

Here's why this is a problem: The voters of color pollsters run into in majority-white precincts might not be representative of the voters of color across the state. In particular, according to Latino Decisions, voters of color living among whites are "more assimilated, better educated, higher income, and more conservative than other minority voters."

Check out the difference in the percentage of nonwhite voters who had a college degree in 2010, according to the US Census versus the exit poll:

Education level nonwhite voters exit polls Latino Decisions

And the problem is even worse for Latino voters, because exit polls are almost never offered in Spanish — even though more than a quarter of Latino voters prefer Spanish to English. So the exit polls oversample English-speaking Latinos.

All these issues together mean that the exit polls sometimes think Latino voters are much more favorable to Republicans than they actually are. In 2010, for example, Harry Reid won reelection to the Senate by turning out Latinos to vote against his Republican challenger, Sharron Angle, who was running as a hardcore immigration hawk. But according to the exit polls, 30 percent of Nevada Latinos voted for Angle — many more than voted for John McCain for president in 2008. (When the official vote tallies came out, it became clear that more than 90 percent of Latinos had voted for Reid.)

sad Sharron Angle supporters
It turns out their candidate didn’t get 30 percent of the Latino vote after all...

So are exit polls usually biased toward Republicans?

Nope! As a matter of fact, even with the issues listed above, exit polls have historically been biased toward Democrats more often than they've been biased toward Republicans.

In 2004, for example, the exit polls overestimated John Kerry's share of the vote (by "more than one standard error") in 26 states; it overestimated George W. Bush's share in only four states. The reason for the error? Bush voters were more likely than Kerry voters to refuse to answer a pollster's questions after they left the voting booth.

Are the exit polls going to accurately predict the 2016 election?

Traditionally, the exit polls don’t get the outcome wrong that often. The errors in exit polls aren't enough to get the outcome of the race wrong — they just might misstate how much the victor won by, or who supported him/her the most.

Admittedly, this is not a typical election, and it’s possible the electorate won’t be typical either. The data from early voting suggests that pollsters might have underestimated the Latino vote and (perhaps) misjudged how many of them would vote for Hillary Clinton. That’s the sort of error the exit polls would also be liable to make. On the other side, Republican champions of Donald Trump suggest he’ll be able to turn out unexpected numbers of white voters without college degrees — who are liable to live in deep red areas where the exit pollsters won’t be either.

But these are reasons to be cautious of the demographics that the exit polls present in their detailed data. Exit polls aren’t the only reason a state gets called. By the time exit polls can be released, networks have a day’s worth of information about how an election has gone in a particular state — and if they think that information shows the exit poll might be wrong, they’ll wait to call the state for Trump or Clinton.

If the polls close on Election Day and CNN immediately projects that your preferred presidential candidate will lose your state, don't hold your breath for an eventual victory. But if the exit polls project your candidate will win and he or she ends up losing, it’s the fault of the exit poll, not proof of a rigged election.

What to watch for in the early states

Sign up for the newsletter Sign up for Vox Recommends

Get curated picks of the best Vox journalism to read, watch, and listen to every week, from our editors.