- Exit polls are conducted by Edison Research for the 6-member NEP consortium (ABC, CBS, NBC, CNN, Fox, AP); results are released only after polls close in each state to prevent influencing voters still in line
- 2022 exit polls overstated Democratic performance due to differential non-response bias — Democratic-leaning voters are more likely to participate in exit polls than Republicans, creating a systematic skew that has appeared across multiple election cycles
- The dramatic growth of early voting since 2020 has fundamentally changed exit poll dynamics: fewer voters cast in-person ballots on Election Day, meaning traditional polling-place exit polls now capture a smaller and less representative share of the full electorate
- For 2026, Edison Research has expanded its telephone exit poll component to reach early voters — the combined phone + in-person methodology is designed to reduce the demographic gaps that caused the 2022 overestimates
The National Election Pool: How It Works
The National Election Pool (NEP) is a consortium of six media organizations — ABC News, CBS News, NBC News, CNN, Fox News, and the Associated Press — that jointly fund a shared exit polling and vote-count reporting operation. Rather than each network running its own separate exit poll (expensive and logistically redundant), they pool resources to hire Edison Research, the primary contractor, which conducts surveys at selected precincts nationwide.
Edison deploys trained interviewers to a stratified random sample of precincts in each state. Sample sizes vary by state competitiveness: a swing state Senate majority math might have 50-75 precinct locations with 50-100 interviews each, yielding 3,000-5,000 total responses for that state. Interviewers approach every third to fifth voter exiting the polling place and ask them to fill out a paper questionnaire — anonymously — covering candidate choice, key issues, and demographic data.
Why 2022 Exit Polls Were Off — and What Changed
The 2022 midterm exit polls significantly overestimated Democratic performance in several demographic categories, particularly among Hispanic voters and younger men. The primary cause is differential non-response bias: Democratic-leaning voters are more likely to agree to participate in exit poll interviews than Republican-leaning voters. This creates a systematic tilt toward Democratic respondents in the raw data that weighting procedures do not fully correct.
A second issue is the dramatic growth of early and mail voting. In 2022, approximately 40-45% of votes were cast before Election Day in many competitive states. Traditional polling-place exit polls only capture Election Day in-person voters — a shrinking and increasingly skewed sample. Edison's response for 2026 is to expand the phone component of the survey, calling registered voters who self-reported voting early, and combining phone and in-person data with different weighting procedures.
How to Read Exit Poll Data on Election Night
The most important rule: exit poll topline numbers released within the first two hours after polls close are based on small, partially weighted samples and should not be treated as outcome predictions. Networks have learned this lesson repeatedly. Early exit data is useful for demographic breakdowns — which candidate is winning among seniors, which issues voters cited as most important — but unreliable for margin estimates.
The adjusted and weighted final exit polls, released 24-72 hours after election night, are the authoritative versions. These are recalibrated to match actual vote results, making them reliable for demographic analysis. Political scientists use these final versions for academic research on voter behavior. The "exit poll" numbers cited in post-election journalism are almost always the final adjusted versions, not the election-night preliminary data.
Demographic breakdowns (gender, age, education, race). Key issue identification. Candidate approval among specific groups. Cross-tabs for political analysis.
Exact vote margins. Early-night outcome predictions. Capturing early/mail voters accurately. Detecting late-shifting Hispanic and working-class voter trends.
Expanded phone component for early voters. Revised weighting for known non-response bias patterns. Better Hispanic voters sampling procedures after 2020 and 2022 misses.
Frequently Asked Questions
How do exit polls work and who conducts them?
Edison Research conducts exit polls for the National Election Pool (ABC, CBS, NBC, CNN, Fox, AP). Trained interviewers survey voters at randomly selected precincts. In 2026, a phone component reaches early and mail voters. Results are weighted, embargoed until polls close, and released progressively through election night.
Why did exit polls fail in 2022?
Two main causes: differential non-response bias (Democrats more likely to participate in exit surveys, creating systematic tilt) and the growth of early voting (polls only captured Election Day in-person voters, missing a large portion of the electorate). Edison has expanded phone interviewing for 2026 to address both issues.
How should viewers interpret exit poll data on election night?
Early election-night exit data should not be treated as outcome predictions — it's based on small partial samples. Demographic breakdowns are more reliable than topline margins. The final adjusted exit polls released 1-3 days later are the authoritative versions used by researchers.
