Election-night coverage can change quickly, but the terms used in headlines do not all mean the same thing. This guide explains how to read election polls, forecasts, projections, exit polls, unofficial tallies, and certified results so you can follow news updates carefully, compare claims fairly, and know which information is still subject to change.
Overview
The most important distinction is between information that measures opinion and information that records votes. A poll asks a sample of people how they may vote or how they currently view a contest. A forecast or model estimates likely outcomes using polling and other inputs. A projection is a news organization’s or analyst’s judgment that the available vote information is sufficient to identify a likely winner. An unofficial result is a count reported during or after vote counting, while a certified result is the legally completed outcome confirmed by the responsible election authority.
These categories answer different questions. A poll describes sentiment at a particular time; it does not count votes cast on Election Day. A forecast describes probability under stated assumptions; it is not a guarantee. A projection interprets partial results and may be made before every ballot is counted. A certified result follows required counting, reconciliation, canvassing, and any applicable review procedures.
Terminology and procedures vary by country, state, province, territory, county, or municipality. Before relying on an election-night label, check how the newsroom defines it and compare the information with the official election authority for that contest.
| Term | What it measures | How to interpret it | Can it change? |
|---|---|---|---|
| Poll | Responses from a selected group of people | A snapshot of opinion, not a vote count | Yes; new polls can show different results |
| Margin of error | Sampling uncertainty expressed around a survey estimate | Use it with the sample, method, and question wording | It does not account for every possible error |
| Forecast | A model’s estimated chances or expected outcome | Read the assumptions and probability carefully | Yes; models update as inputs change |
| Projection | An assessment that a candidate is likely to win | It reflects available vote data and analytical judgment | It may be revised if information was incomplete or misstated |
| Exit poll | Survey responses from people after voting | Useful for analysis, but not an official count | Yes; early samples and weighting can be adjusted |
| Unofficial result | Votes reported before the formal process is complete | A running tally, often incomplete | Yes; later reporting, corrections, and outstanding ballots matter |
| Certified result | The legally finalized outcome | The authoritative result for the jurisdiction | Only through procedures allowed by election law |
What to track
When reading a poll, do not focus only on the topline percentage. Record the details that determine what the number can reasonably tell you:
- Sample size: Note how many people were surveyed and whether the figure refers to all adults, registered voters, likely voters, or another group. A larger sample can reduce random sampling uncertainty, but it does not automatically make a survey representative.
- Margin of error: A stated margin of error describes a range associated with sampling variation under particular assumptions. It is not a guarantee that the true result falls inside the range, and it generally does not capture every source of survey error. Avoid treating two close poll numbers as meaningfully different without considering the uncertainty.
- Methodology: Check whether respondents were contacted online, by telephone, in person, or through a mixed approach. Look for information about weighting, response rates when available, question order, field dates, and whether the pollster identifies a likely-voter screen.
- Turnout assumptions: Election outcomes depend on who votes, not simply on the preferences of all respondents. A poll that uses one likely-voter model may produce a different result from one using another model.
- Question wording: Small wording differences, candidate order, response options, and whether undecided voters are prompted can affect answers. Read the full question when it is available.
- Reporting time: Polls describe a period when interviews took place. They may not reflect later events, new information, late campaign activity, or changing voter decisions.
For forecasts and projections, track the inputs and the confidence language. “Leads,” “favored,” “likely,” and “projected” are not interchangeable. A model showing a candidate ahead is different from a statement that the contest has been called. A projection may also depend on which precincts have reported, the expected composition of outstanding votes, and whether the remaining ballots come from areas with known voting patterns.
For results, watch the total number of votes counted, the share of precincts or reporting units included, the areas still outstanding, and whether the reported figures are unofficial. Percent of precincts reporting is not always the same as percent of votes counted, because reporting units can differ greatly in size.
Cadence and checkpoints
Use a different reading routine before, during, and after an election.
Before voting
Save the official election authority’s website for the relevant jurisdiction. Review the election date, offices on the ballot, ballot-counting explanations, and any published rules about canvassing, recounts, or certification. For polling, compare multiple surveys rather than relying on a single release. Record field dates, sample definitions, methodology, and margin of error so that older and newer polls are not compared as if they were identical.
During election-night updates
Start with the official results page when available, then use reputable news organizations for context and projections. Check the timestamp and identify whether the numbers are updated continuously or in batches. Ask what has been counted, what has not, and whether the displayed margin is based on all reported votes or only the current unofficial total.
After polls close
Treat exit polls and early tallies as provisional information. Do not confuse a news projection with certification. As additional ballots are counted, compare the new total with earlier updates and look for correction notices. If a result is close, pay particular attention to the jurisdiction’s official explanation of outstanding ballots and next procedural steps.
Step-by-step verification workflow
- Identify the exact contest, jurisdiction, and election date.
- Open the official election authority’s results or election-information page rather than relying only on a screenshot or social media post.
- Check the page’s timestamp, status label, vote totals, reporting units, and any notes about unofficial figures.
- Compare the authority’s figures with at least one established news report, looking for discrepancies in timing or presentation.
- Read official notices about canvassing, provisional or absentee ballots, recounts, corrections, and certification if they apply.
- Share the result using precise language: “unofficial tally,” “projected winner,” or “certified result,” rather than using “won” when the status is unclear.
How to interpret changes
A change in a poll is not necessarily evidence that voters moved. It may reflect a different sample, turnout screen, weighting method, question wording, or field period. Compare trends only when the polls are sufficiently similar and consider the average pattern rather than treating one release as decisive.
A change in a forecast may result from new polling, revised assumptions, additional information about turnout, or a model update. Read the model’s explanation before describing the change as a shift in public opinion.
An unofficial result can change for ordinary administrative reasons, including the reporting of additional vote batches, corrections, reconciliation, or the counting of ballots that take longer to process. A narrowing margin does not by itself prove a problem, and a widening margin does not by itself establish that all ballots have been counted.
Be especially careful with percentages presented without denominators. “Candidate A has 52%” could refer to a poll, a partial count, or a model estimate. Ask: 52% of what group, measured when, using which method, and with what status? This simple check prevents many misleading comparisons in breaking news and live election coverage.
When to revisit
Return to this guide before each primary, general, local, or special election, and whenever a major change affects how voting or counting is administered in the relevant jurisdiction. Polling standards, result dashboards, reporting conventions, and official links can change, so editors should review the terminology and verification links before every major election cycle and after meaningful administrative updates.
Readers can use a shorter checklist whenever a new headline appears:
- Is this a poll, forecast, projection, unofficial tally, or certified result?
- What is the date or time of the information?
- How large and representative is the sample, and what is the stated margin of error?
- What turnout assumptions and methods were used?
- How many votes or reporting units remain?
- Does the official election authority confirm the status?
Bookmark the official election page, record the status of important updates, and wait for formal confirmation before treating an election as finalized. For broader verification habits during fast-moving coverage, see How to Verify Breaking News Before You Share It and How to Find Reliable Local News and Community Updates Near You. Careful distinctions take only a few minutes, but they make election-night news easier to understand and less likely to mislead.
