Exit polls on election night work by interviewing voters immediately after they cast their ballots at physical polling places, combined with extensive phone and online surveys designed to capture early and mail-in voters. Conducted by major research organizations on behalf of media consortiums, these polls provide the earliest look into who voted, what issues drove their decisions, and how different demographic groups behaved.

While viewers on election night see polished graphics on their screens, the underlying machinery of an exit poll is an incredibly complex, fast-moving logistical feat. Understanding how this data is gathered, weighted, and used by network decision desks is key to interpreting the numbers without being misled by early, incomplete results.

The Two Main Players: NEP and AP VoteCast #

For decades, exit polling in the United States was dominated by a single entity. Today, the landscape is split into two major systems, each using a fundamentally different methodology to capture the electorate.

The National Election Pool (NEP) and Edison Research #

The National Election Pool is a consortium of major news networks—including ABC, CBS, NBC, and CNN. Since 2003, the NEP has contracted Edison Research to conduct its exit polls.

Edison’s approach relies on the traditional, classic exit poll model. On Election Day, Edison sends hundreds of field interviewers to randomly selected voting precincts across the country. These interviewers stand outside the polling places and ask voters to fill out a brief, anonymous paper or digital questionnaire as they exit. Because this method cannot capture those who vote before Election Day, Edison supplements these in-person interviews with extensive pre-election telephone and online surveys of mail-in and early voters.

AP VoteCast and NORC #

In 2018, the Associated Press and Fox News departed from the NEP to create a new system called AP VoteCast, conducted by NORC at the University of Chicago. Other networks, such as PBS NewsHour, also use this data.

Rather than sending interviewers to stand outside physical polling places, AP VoteCast is not a traditional “exit” poll. Instead, it is a massive, multi-mode survey of registered voters conducted in the days leading up to Election Day, continuing until the polls close. It utilizes a combination of a random sample drawn from state voter files and a self-selected panel of online respondents. This methodology was designed specifically to address the decline in Election Day turnout and the massive rise in mail-in and early voting.

Comparing how these different methodologies capture voter behavior can be fascinating, especially when tracking how pre-election sentiment aligns with final outcomes. To see how these real-time shifts compare to broader public opinion, you can use interactive election tracking tools to view polling averages and sentiment trends leading right up to the close of voting.

Step-by-Step: How Traditional Exit Polls Are Conducted #

To understand how Edison Research pulls off a traditional exit poll on behalf of the NEP, it helps to look at the process as a sequence of highly coordinated steps.

[Precinct Selection] ➔ [Random Voter Interviewing] ➔ [Data Transmission] ➔ [Weighting & Calibration]

1. Selecting the Precincts #

Edison cannot send interviewers to all of the tens of thousands of voting precincts in the United States. Instead, they use a statistical method called stratified probability sampling.

First, they divide a state into geographic and demographic regions. Within those regions, they select specific precincts that represent a microcosm of the state’s political makeup. Precincts with higher populations or higher historic voter turnouts have a higher probability of being selected.

2. Interviewing Voters at a Random Interval #

To prevent interviewers from introducing bias by only talking to people who look friendly, young, or of a certain demographic, Edison enforces a strict systematic sampling interval.

For example, an interviewer might be instructed to approach every third or fifth person who exits the building. If that person refuses, the interviewer cannot simply pick the next person nearby; they must wait for the next person in the designated interval.

The survey itself is brief. It fits on a single sheet of paper or a small tablet screen, asking about:

• Demographic information (age, race, gender, education, income, religion).
• Candidate choice.
• Key issues (the economy, foreign policy, healthcare, social issues).
• Opinions on the candidates’ character traits.

3. Gathering Early and Mail-In Data #

Because a massive percentage of the electorate votes before Election Day, traditional exit polls would be highly inaccurate if they only polled on-site voters. To correct for this, Edison conducts landline, cellular, and online surveys in the weeks leading up to the election. This ensures that the early-voting cohort is represented proportionally in the exit poll model.

4. Transmitting and Weighting the Data #

Throughout Election Day, interviewers call in their raw numbers or upload their digital data to a central database at three or four designated times (e.g., mid-morning, early afternoon, and late afternoon).

As the raw data comes in, statisticians apply weighting. If a precinct’s actual turnout demographics are known to be 52% female, but the exit poll sample is 58% female, the responses are mathematically adjusted so that men and women reflect their actual proportions of the voting population.

The “Quarantine” Room: Preventing Leaks and Market Panic #

Exit poll data is highly sensitive. If early, unweighted exit poll numbers were leaked to the public at 11:00 AM on Election Day, it could cause chaos. It could suppress voter turnout in western states where polls are still open, or cause wild fluctuations in the stock and prediction markets.

To prevent this, the NEP enforces a strict protocol known as the Quarantine.

On the morning of Election Day, representatives from the participating news networks (including their chief political directors and statisticians) enter a secure, isolated room in a secret location. Upon entering, they must surrender all mobile phones, smartwatches, tablets, and any other communication devices.

In this room, the representatives have early access to the raw, unweighted exit poll data. They spend the day analyzing the demographics and issue priorities, preparing their graphics and talking points for the evening broadcast. No one in the Quarantine room is allowed to communicate with the outside world until 5:00 PM Eastern Time, at which point the networks are permitted to publish general, non-outcome-related data (such as what issues voters prioritized). No projections about who won a state can be made until the official voting polls in that state have closed.

While exit polls capture the final decisions of voters on election night, checking live presidential approval ratings and pre-election polling trends beforehand offers vital context on how those voter decisions were shaped over months of campaigning.

Why Exit Polls Can Be Misleading (and How They Fail) #

Exit polls are incredibly useful for understanding why people voted the way they did, but they are not designed to be a perfect, real-time vote count. Historically, there have been famous instances where exit polls pointed to the wrong winner.

The most famous modern exit poll failure occurred in the 2000 Presidential Election. Early exit polls in Florida led networks to project Al Gore as the winner of the state before voting was fully completed in the western Panhandle (which is in a different time zone). Later that night, as actual votes were counted, the networks had to retract their projections and place the state back in the “undecided” column, eventually projecting George W. Bush, before retracting that projection as well.

Because of these vulnerabilities, network “decision desks” do not rely on exit polls alone to project a winner. Instead, they treat exit polls as just one input, blending them with actual, incoming precinct-level vote counts and historical voting patterns as the night progresses.

How to Read Exit Polls Like an Expert #

If you are watching the coverage on election night, here are a few rules of thumb to help you interpret the exit polls without falling into common traps.

• Ignore the horse-race numbers in early leaks. If you see a leaked screenshot on social media at 2:00 PM showing one candidate leading by 10 points in a swing state, ignore it. Early exit poll data is highly unweighted and frequently biased toward whichever party’s voters are more enthusiastic about talking to interviewers early in the day.
• Focus on demographics, not margins. Exit polls are at their best when explaining who voted. Look at the breakdowns: How did college-educated women vote? Did suburban voters shift compared to four years ago? These cross-tabulations are highly reliable even if the overall statewide margin is too close to call.
• Look at the “Why.” Exit polls are the only tool we have to understand voter motivation on election night. Pay attention to the “most important issue” questions. If 40% of voters in a state say the economy was their top priority, and that group broke heavily for one candidate, you have found the narrative of the election.
• Watch how the numbers change. As actual votes are counted, exit polls are adjusted and “calibrated” to match the real results. The exit poll numbers you see at midnight are far more accurate than the ones you saw at 7:00 PM because they have been mathematically anchored to the actual vote tally.

If you want to follow these shifting dynamics on election night and beyond, downloading a comprehensive election companion app makes it easy to monitor everything from approval trends to prediction markets in one clean interface.

Frequently Asked Questions #

Do exit polls affect voter turnout? #

There is concern that releasing exit poll results early in the day can depress turnout, particularly in western states where polls are still open. To prevent this, US media networks have a strict agreement not to project the winner of any state until all the polls within that specific state have officially closed.

Why do exit polls sometimes change throughout election night? #

Exit polls are not static. When they are first broadcast at 7:00 PM, they are based entirely on survey responses. As the night goes on and official county-by-county vote counts are reported, statisticians “weight” the exit poll data to match the actual, incoming vote totals. By the end of the night, the exit poll is a hybrid of survey data and real election results.

Are exit polls used to detect election fraud? #

In some countries with developing democratic institutions, exit polls are used by international observers as an independent check against official vote tallies to detect potential election fraud. However, in the US, exit polls are designed for media analysis and are not precise enough to serve as an official audit of the voting systems.

What is the difference between an exit poll and a pre-election poll? #

Pre-election polls ask registered or likely voters whom they intend to vote for, often weeks before the election. Exit polls ask actual voters who they just voted for moments after they left the booth or after they completed their mail-in ballot. This eliminates the uncertainty of whether a respondent will actually show up to vote.