Overview of How we know when someone wins an election
This NPR Politics Podcast episode explains how the Associated Press determines when to “call” an election, why those calls matter, and why election night usually reveals only part of the story. AP elections vice president David Scott walks through the organization’s process for gathering vote totals, combining them with survey data, and making high-confidence decisions about when a candidate has truly won.
How the AP makes a race call
What a “call” means
- AP does not use the word projection; it says a candidate has won only when it is certain the trailing candidate can no longer catch up.
- The standard is certainty, not speed.
- AP’s calls are based on the idea that election results are news, and the public deserves a clear, evidence-based answer.
How AP gathers information
- AP relies first on the vote count from roughly 5,000 reporting units nationwide.
- It uses multiple sources:
- county and local election offices
- online result feeds
- emails, photos of tally sheets, handwritten notes, or whiteboards
- direct interviews with election officials
- AP then performs quality control to check for inconsistencies or possible reporting errors before publishing results.
Scale of the operation
- AP deploys about 5,000 people on election night.
- It covers roughly 7,000 elections in a general election cycle, including runoff races.
The voter poll: how AP understands the electorate
What it is
- AP’s voter poll is an evolution of its old VoteCast system and the networks’ traditional exit polls.
- It combines methods to capture:
- people who vote early
- people who vote on Election Day
- a fuller picture of voter behavior and motivation
Why it matters
- The voter poll helps explain who voted, how they voted, and why.
- It provides the demographic and issue-based data used in election analysis, such as voting patterns by race, age, or party.
Accuracy, mistakes, and lessons learned
AP’s error standard
- AP says its standard is 100% accuracy.
- Even though mistakes are rare, the organization studies every error closely to improve future calls.
Examples of incorrect calls this year
- AP said it had made about 12,000 calls in primaries and special elections so far this cycle, with four incorrect.
- Causes included:
- bad information from election officials
- an ineligible candidate being counted
- a technical data entry error
- underestimating remaining votes in a race
Arizona 2020 and other lessons
- AP revisited how it analyzes advance ballots after the close Arizona presidential call in 2020.
- The episode also references the Florida recount in 2000 as a turning point that pushed election coverage toward more sophisticated data and decision-making.
Why race calls matter
- Race calls are part of the news, not just a convenience for impatient viewers.
- AP sees its role as a mirror of the voters and a safeguard against misinformation.
- Calling a race when it is clearly decided helps prevent confusion and rumor from filling the information gap.
What to watch on election night
Be patient with timing
- Different states close polls at different times, and some cross time zones.
- Results may begin arriving before all polls are closed statewide, but a true AP call won’t happen until the data support certainty.
Understand the structure of U.S. elections
- The U.S. system is highly decentralized, with states and counties controlling much of the process.
- Because election administration varies widely, results take time to fully unfold.
- That complexity makes it harder for misinformation to take hold, but it also means election night often lasts well beyond the first wave of returns.
Key takeaway
The main message of the episode is that election night is only the beginning. AP’s race calls are built on a careful, evidence-based process meant to balance speed with certainty, and voters should expect a gradual rollout of results rather than instant answers for every race.