Exit polls survey people who have just voted, and today they also survey the large share who voted early or by mail, so the “exit” part is only one piece. The results are weighted during the night to match the actual vote as it’s counted, and news organizations use them mainly to explain who voted and why. Race calls rely far more on real vote counts than on exit polls.
The system changed in 2025. After the research firm SSRS acquired Edison Research, the TV networks that had run one exit poll joined the Associated Press and Fox News, which had run a rival survey, so most major outlets now work from the same voter survey for 2026.
Who runs exit polls in 2026? #
From 2018 through 2024, there were two competing systems:
- The National Election Pool (NEP): ABC, CBS, CNN and NBC, with Edison Research conducting the survey since 2003. It interviewed voters at a sample of polling places and supplemented that with phone and online surveys of early and mail voters.
- AP VoteCast: Created in 2018 by the Associated Press and Fox News with NORC at the University of Chicago. It skipped polling-place interviews and instead surveyed tens of thousands of registered voters in the days before and on Election Day, online and by phone.
In 2025, SSRS bought Edison Research, and the networks joined the AP and Fox to survey voters together. The resulting survey, called The Voter Poll by SSRS, was used for the 2025 elections and is planned for November 3, 2026. It relies mainly on online and phone interviews, with a smaller in-person component at polling places than the old exit polls had.
Why aren’t exit polls just in-person anymore? #
Because a large share of Americans now vote before Election Day. In 2020, most voters cast ballots early or by mail, and even in the 2022 midterm the share was high. An old-style exit poll that only interviewed people leaving polling places on Election Day would miss those voters, and early and Election Day voters have leaned differently in recent elections.
So modern voter surveys start before Election Day, reaching people by phone, text and online, and ask whether they’ve already voted or how they plan to.
How does a traditional in-person exit poll work? #
The in-person piece still follows a design pollsters have used for decades:
- Pick precincts. Pollsters choose a random sample of polling places within each state, weighted so larger precincts are more likely to be picked, and grouped so the sample reflects the state’s regions and past voting patterns.
- Interview at an interval. An interviewer approaches every third or fifth voter leaving, for example, rather than picking people who look approachable. If someone declines, the interviewer notes their approximate age, sex and race so refusals can be accounted for.
- Ask a short questionnaire. Vote choice, demographics, the issues that mattered most, views of the candidates and the direction of the country.
- Report during the day. Interviewers send results to a central team several times on Election Day.
How are exit polls weighted? #
Weighting happens in stages through the night:
- Before polls close, the survey is weighted using demographic targets and turnout estimates.
- As votes are counted, analysts adjust the survey so its vote-choice numbers match the actual results in each state. By the end of the night, the exit poll’s topline matches the count, and the demographic breakdowns shift along with it.
That’s why the numbers you see at 7 p.m. and the ones you see the next morning can differ. The later numbers are more reliable because they’re anchored to real votes. For more on how weighting works in general, see how pollsters weight survey data.
What is the exit poll “quarantine room”? #
Early exit poll data is sensitive, and leaks in past elections caused confusion. So the organizations involved keep a small team in a secure room in New York on Election Day, where participants hand over their phones and other devices. They can study the early data but can’t share it until an agreed release time, traditionally 5 p.m. Eastern, and even then only for non-horse-race findings such as which issues voters cared about. By long-standing practice, the networks don’t project a winner in any state until its polls have closed.
Why have exit polls been wrong before? #
| Problem | What happens | How it’s handled |
|---|---|---|
| Differential response | One side’s voters are more willing to talk to interviewers | Weighting to demographics and, later, to actual results |
| Early and mail voters | Missed by polling-place interviews | Pre-election phone, text and online surveys |
| Unrepresentative precincts | A few sampled precincts behave oddly | Adjusting to real precinct and county results |
| Early leaks | Unweighted numbers spread online | Quarantine and embargo rules |
The most famous miss came in 2000. Networks projected Al Gore the winner of Florida around 7:50 p.m. Eastern, before polls had closed in the state’s western Panhandle, using early exit poll data and incomplete returns. They retracted the call, later called the state for George W. Bush, then retracted that too. In 2004, early exit poll numbers leaked online showed John Kerry doing better than he did. Both episodes are why exit polls now play a smaller role in race calls.
How should you read exit polls on election night? #
- Ignore early leaked toplines. Afternoon numbers circulating on social media are unweighted and often misleading.
- Focus on who and why. The best use of exit polls is explaining how groups voted and what issues mattered. Our guide to reading crosstabs helps here.
- Mind small groups. A breakdown for a small demographic can carry a wide margin of error.
- Expect revisions. Exit poll numbers keep changing until the count is mostly in. Months later, studies that match surveys to voter records, such as Pew Research’s validated voter studies, give a more precise picture of who actually voted.
For following the count itself, see our guide to getting live election results on your phone.
What should you know before November 3? #
Exit polls tell you what happened; pre-election polls tell you what to expect. Knowing where the averages stood makes the night easier to read, because you can see whether a state’s early count is running ahead of or behind expectations. Election Tracker groups Senate and governor polls by race with a 30-day average, averages the generic congressional ballot, and shows presidential approval. It doesn’t show exit polls or results, but it gives you the baseline going in. It’s free on iPhone.
Frequently asked questions #
Are exit polls accurate? #
Their final, fully weighted numbers are matched to the actual vote, so the topline is accurate by design. The early numbers, before weighting to real results, have been off in several elections. Their demographic breakdowns are useful but carry real uncertainty, especially for small groups.
Why do exit poll numbers change during the night? #
Because they’re reweighted as actual votes are counted. Analysts adjust the survey so it matches the real results in each state, which also shifts the breakdowns by age, race, education and other groups.
Do exit polls affect turnout? #
That’s the concern behind the rules. Networks agreed long ago not to project a state’s winner until its polls close, and early data is kept under embargo, to avoid discouraging people who haven’t voted yet.
What’s the difference between an exit poll and a pre-election poll? #
A pre-election poll asks people how they plan to vote, often weeks ahead. An exit poll asks people who have already voted, on Election Day or earlier by mail or in person, how they voted and why.
Who conducts exit polls in 2026? #
SSRS, which acquired Edison Research in 2025, conducts The Voter Poll for a group that now includes ABC, CBS, CNN, NBC, the Associated Press and Fox News.