Two reports can give different turnout percentages for the same election because they divide by different populations. One may describe participation among registered voters. Another may use citizens of voting age. A third may use all people of voting age or a separately constructed eligible-population estimate.
The difference is not automatically an error. A percentage is a relationship between a numerator and a denominator, and both need definitions. The data source, date, geography, and meaning of “voted” also matter.
This article explains measurement. It does not determine any person's voting eligibility or describe current procedures for registering or casting a ballot.
One invented election, several valid arithmetic results
Suppose a fictional jurisdiction records 800 participating voters under a clearly defined administrative measure. For the example, it has 1,000 registered voters, 1,250 citizens of voting age, and 1,500 people of voting age in total.
| Denominator used | Calculation | Turnout under that definition |
|---|---|---|
| Registered voters | 800 / 1,000 | 80% |
| Citizen voting-age population | 800 / 1,250 | 64% |
| Total voting-age population | 800 / 1,500 | About 53.33% |
The numerator stays 800 throughout. The percentage rises as the denominator becomes smaller. A headline saying only “turnout was 80%” cannot be compared fairly with another saying “turnout was 64%” until their bases are identified.
These numbers are invented. They do not describe any real election or imply that every member of a listed population is eligible under applicable law.
The Election Assistance Commission's EAVS report explicitly discusses multiple turnout denominators and their limitations. The choice depends on the question and on the coverage of the numerator, not on which percentage looks largest.
Registered voters and voting-age residents answer different questions
A registered-voter denominator asks about participation relative to the registration list or specified registration category. A voting-age-population denominator asks about participation relative to a broader population defined by age.
Citizen voting-age population narrows that population by citizenship. It is still a statistical category, not a complete individual eligibility determination. Other legal conditions and the data source's coverage can matter. A measure called voting-eligible population has its own construction that should be read in the source methodology rather than inferred from its name.
The Census Bureau's voting FAQ explains why removing noncitizens from the population base changes a rate for a given numerator. The arithmetic is a denominator change; it does not mean that additional ballots were found.
Labels such as active registered voters and total registered voters can also refer to different administrative categories. Their definitions, reference dates, and list-maintenance processes matter to a comparison. Do not treat a category name from one dataset as universally identical to a similar name in another.
The numerator needs as much attention as the denominator
“Votes” can refer to different things in ordinary discussion: people who participated, ballots cast, ballots counted under a defined measure, or votes recorded in one particular contest. Those quantities should not be substituted without explanation.
An individual can participate in an election without recording a valid vote in every contest on the ballot. A contest's total therefore need not equal the number of participants. The applicable reporting definitions explain the relationship.
The EAVS report defines its participation measure and notes a change in wording across survey years. That is a reminder to read the actual variable definition when building a time series. A familiar field label can hide a change in exactly what was requested.
For an invented example, 800 people participate but 760 votes are recorded across the options in one contest. Using 760 to calculate overall participation would answer a narrower contest-based question. It would not automatically reveal that 40 people failed to participate in the election.
Survey reports and administrative counts are different evidence
A population survey can ask respondents whether they voted and collect demographic information. Administrative election records document events in an election process. The sources have different purposes, coverage, and possible measurement limitations.
The Census report on the 2022 election labels its survey population and its citizen voting-age turnout basis. Its percentages should be described as survey estimates, not silently rewritten as an official count of ballots.
Survey responses can be affected by nonresponse, recall, reporting behavior, and sampling. Administrative data have their own definitions and completeness issues. A difference between the two is a reason to examine the methodologies, not sufficient evidence by itself that either source is intentionally deceptive.
Our survey margin-of-error guide explains one part of survey uncertainty. Sampling error is not the only possible source of discrepancy, and it should not be used as a catch-all explanation for every mismatch.
Geography and dates must line up
A numerator for one boundary should not be divided by a population estimate for another. Changes in jurisdictional boundaries, coverage of special populations, and the availability of subarea estimates can all affect the match.
The population denominator may also come from an estimate produced before or after the election. That does not automatically make it unusable, but the source should explain the reference period and why it was chosen.
Our one-year and five-year ACS guide explains why a population estimate labeled with an ending year may represent a longer collection period. An election is an event at a particular time; a multi-year population estimate is a statistical summary across a period.
Combining them requires a clear account of what the resulting ratio represents. It should not be presented as though every input were a complete census taken on election day.
Registration growth can change one turnout rate more than another
Return to the fictional jurisdiction with 800 participants and 1,000 registered voters. Its registered-voter turnout is 80%. Suppose registration grows to 1,100 in a later comparable election, and participation grows to 850.
Participation increased by 50 people, yet registered-voter turnout falls to about 77.27%. The denominator grew faster than the numerator. A statement that “turnout fell” needs to identify whether it means the rate or the number of participants.
If the population base also changed, population-based turnout would need its own calculation. One ratio's direction cannot be assumed from another's.
Our percentage-points explanation also matters here. The change from 80% to about 77.27% is a decrease of about 2.73 percentage points, while the relative percentage decrease in the rate uses 80% as its starting base.
Combined rates need combined denominators
Suppose two fictional districts have turnout of 90 out of 100 and 500 out of 1,000 registered voters. Their rates are 90% and 50%. The simple average is 70%, but their combined registered-voter turnout is 590 divided by 1,100, about 53.64%.
The smaller district should not receive the same weight as the larger one when the question concerns all registered voters combined. Add compatible numerators and compatible denominators before dividing.
Missing data introduce another complication. If only jurisdictions with both required inputs are included, the reported combined rate describes that coverage. Read footnotes before assuming a national label necessarily means every jurisdiction contributed to every calculation.
What a rate cannot establish by itself
A turnout percentage describes participation under a definition. It does not by itself identify why people participated or did not participate, whether a particular process was convenient, or what every nonparticipant believed.
Those questions require additional evidence. A rate comparison can motivate an investigation, but a causal explanation should not be inserted merely because two places have different percentages.
An accurate citation of turnout can be brief: name the election, geography, participation measure, denominator, and source. If the result is a survey estimate, say so. Those details make the number comparable and useful without asking a single percentage to describe the entire election or the circumstances of every voter.
Sources
- U.S. Election Assistance Commission: 2024 EAVS Report
Turnout measures can use different denominators; the EAVS defines its counted-ballot numerator and discusses coverage and limitations of population and registration bases.
- Census Bureau: Voting and Registration FAQ
Survey-reported voting is distinct from administrative election counts, and voting-age and citizen voting-age bases produce different rates.
- Census Bureau: Voting and Registration in November 2022
The report defines voting-age and citizen voting-age populations and labels its survey-based turnout denominator; its estimates have the survey’s population coverage and limitations.