Every reporting unit's votes are sorted into the method categories its state or county actually publishes. Categories are never inferred: if a source doesn't separate two methods, they stay merged here.
| Category | What it covers |
|---|---|
| Election day | Ballots cast in person at a polling place or vote center on election day. |
| Early in-person | Ballots cast in person before election day at early-voting sites or vote centers. |
| In-person | In-person ballots where the source doesn't distinguish election day from early (e.g., Hawaii's voter service centers). |
| Mail / absentee | Ballots returned by mail or drop box and received by the state's deadline. |
| Late mail | Mail arriving after election day under a postmark grace period, where reported separately (e.g., Maryland, Virginia). |
| Pre-election day | All ballots cast before election day where early in-person and mail can't be separated (e.g., Arizona's "early ballots," Iowa's absentee). |
| Late pre-election day | Pre-election-day ballots processed after election day: Arizona's "late earlies." |
| Provisional | Ballots held for eligibility review and counted after verification. |
| Affidavit | New York's name for provisional ballots. |
| Overseas | Military and overseas (UOCAVA) ballots reported statewide rather than by locality, shown as a pill at the bottom of the map (Alaska, Maine, Rhode Island). |
| Unallocated | Votes a source reported without assigning a method. Kept as their own row, never redistributed. |
| Total | Used when a unit publishes only a combined total, with no method split at all. |
Every county, town, district, and ward carries a grade describing how rich its published method breakdown is. The grade measures the source's reporting, not the accuracy of the votes themselves.
| Grade | Meaning |
|---|---|
| C6 – C1 | The number of distinct vote methods the source breaks out. A C5 county reports five separate methods (say, election day, early, mail, late mail, provisional); a C1 county reports just one. |
| U | Totals only: the unit publishes a single combined number with no method split. Entire states fall here (e.g., Colorado, Wisconsin, Massachusetts) because their official results carry no breakdown. |
| F | Not yet sourced. Reserved for units with no usable data; none currently on the map. |
A state's headline grade is the most common grade among its units. The coverage strip in each state's sidebar shows the full distribution, and the Coverage map view colors every unit by its grade.
About 92% of units carry certified official results. The rest were captured from county or state reporting pages ahead of certification and are flagged unofficial in each unit's source note. Every unit records its deviation from the certified total where one exists, and a firm rule applies throughout: numbers are never scaled, corrected, or redistributed to force agreement with certified totals; deviations are recorded and disclosed instead.
Some counties publish method breakdowns that don't cover every ballot; the remainder is kept as an explicit unallocated row (83 units, mostly in New Mexico, Florida, and New York). Sources are held to these standards before a breakdown is used:
| Data | Source |
|---|---|
| Election returns | MIT Election Data + Science Lab (MEDSL) precinct-level returns; state Secretary of State certified results (web and PDF); county Statements of Vote; OpenElections; Ballotpedia. Each unit's source is named in its panel. |
| Boundaries | U.S. Census Bureau cartographic boundary files (1:500k, 2023) for counties and New England towns; Maine GeoLibrary (METWP24) for Maine townships; Vermont Center for Geographic Information; Census TIGERweb for New Hampshire's grants and purchases. |
| City labels | GeoNames gazetteer. |
Found a bug? Have better data?
Reach out on X: @ferghugg, bug reports, corrections, new or richer sources for any state, and requests for access to the raw and processed data are all welcome. DMs are open.
VoteMode is an independent project. Results are presented as reported by their sources; grades describe reporting richness, not vote accuracy. Last data build: July 2026.