Tracking the Early Vote in Tarrant County

• Tarrant County’s daily list of returned mail ballots and its daily totals report. The big number at the top is the county’s own total.
• The county voter file and its 2018–2020 history file, plus older county rosters for the 2022 elections and the 2024 and 2026 primaries.
• Statewide 2026 primary rosters, for voters who voted in another county.
• Maps: county voting precincts, city limits, city council and school district maps, and congressional plan C2333.
• The Census Bureau’s 2020–2024 American Community Survey, including its counts of adult citizens.
• The county’s 2024 presidential results by precinct, and final turnout for 2018–2024 from the Texas Legislative Council.
• Polling places: the county’s lists of early voting sites and Election Day vote centers, placed on the map with the Census Bureau’s address lookup, and each site’s early-vote totals from the county’s 2018–2024 daily reports.
• Candidates: the Texas Secretary of State’s list of qualified candidates for the Nov. 3 general election, statewide and Tarrant County offices, as of Oct. 11, 2026. Each race lists its candidates by party in alphabetical order (Democratic, Green, Independent, Libertarian, Republican), then write-ins. That is not the order on the ballot. The special elections for Texas Senate District 22 and Texas House District 93 aren’t in that list; they come from the Governor’s proclamations of Aug. 18 and 20, 2026.

Before every update, each number at the top of the page, in the daily chart, the Mail line and the Voters tables is worked out again from the county’s own files and compared with the page, and the page is checked in a browser for broken or out-of-date text. If anything disagrees, the page is not updated.

Early in-person and Election Day ballots are added as the county reports them. On election night, results come from the county’s cumulative and precinct results reports.

• Each named ballot is matched to the voter file by its state voter ID. If the ID isn’t on file, a ballot is matched only when the name, middle initial and city point to exactly one person.
• Older rosters fill in elections the voter file leaves out, and history the county has blanked for some voters.
• Redacted ballots carry only a precinct and dates. They count in every total and on the map, but never get a party label.
• If the county’s count runs ahead of its list, the extra ballots show only in the top number, since they have no precinct yet.

A primary ballot shows which party a voter picked that year, not how they’ll vote in November. Five March primaries count: 2018, 2020, 2022, 2024 and 2026.

• Strong: that party at least 4 of 5 times.
• Mid: 2 or 3 times.
• Weak: once.
• Mixed: both parties equally. Voters who chose both count as weak for the side they chose more often.
• No party history: general elections only, no Tarrant history, or a redacted ballot.

Younger voters are judged only on the primaries they were old enough for. A voter with no primary on record but a party label in the voter file (from a 2026 primary runoff, or history the county blanked) counts as weak for that party.

This is bookkeeping, not a poll or a forecast. Each voter with a Democratic history counts 1 toward Democrats, each with a Republican history 0, and a mixed record ½. Each ballot with no party history counts as its precinct’s Democratic share of March 2026 primary ballots (the county’s share if the precinct had none).

The range shows how far the unknowns could move it: the low end counts every one of them as Republican, the high end as Democratic. Every precinct and district panel uses the same formula; areas with fewer than 5 ballots show only their counts.

• Voters’ party history: the average score above. Diagonal lines mark areas where more than half of the ballots have no party history.
• March primary result: Democratic minus Republican share of the area’s March 2026 primary ballots.
• Shift from the primary: arrows comparing voters returning ballots with the area’s March primary voters. Left means more Democratic, right more Republican, and a longer arrow is a bigger gap. Areas need at least 5 returning voters with a party history.
• Shift from 2024: the same comparison against each area’s 2024 Harris share of the two-party vote. 2024 precincts are moved to today’s areas by where their voters are registered now.
• Turnout: ballots returned divided by registered voters (active and suspense).

Districts are built from voting precincts, and each ballot is placed by the district codes the county gives it. Precincts split by a city or district line are approximate.

Figures come from Census block groups, split into precincts and districts by how much of each block group’s land falls inside, so small precincts are approximate. Race uses adult citizens, which fits the electorate better than total population. Areas with fewer than 100 adult citizens are never selected.

• Past years use the county’s daily totals for the 2018–2024 November elections. Mail ballots that arrived after Election Day are left off.
• The same-day comparison divides each year’s ballots by that year’s registration.
• The turnout bar marks final turnout in the 2018 and 2022 midterms.
• No mail arrives on Sundays and federal holidays, and the county doesn’t always count on weekends, so lines go flat and then jump.

• The county’s first report, due by 7:15 p.m., has countywide totals for mail and early in-person votes only. Election Day votes are added through the night, and precinct-by-precinct results follow.
• With each report, the top of the page switches to results: ballots counted, the race picked under “Results for” (U.S. Senate first, County Judge and every other countywide race in the menu), and a list of every race with each candidate’s votes. The Mail, Early in person and Election Day tabs show the votes cast that way.
• Until precinct results are posted, district panels (Congress, Texas House and Senate, State Board of Education, Commissioner, Justice of the Peace) show their race’s Tarrant County totals. Once they are posted, the map is shaded by the party lead and every panel shows its races. Cities, council districts and school districts cut through precincts, so their figures are estimates and say so.
• The estimates and daily snapshots from before results stay on the map: in each panel under “Estimate before results,” in the Shading menu and on the snapshot slider.
• No race calls. This website will not update races as called until the results have been certified by the county, which may take multiple days or weeks. Until then, candidates are listed in party order, never by who is ahead, and nothing is marked as won.
• Before results go up, every race total is checked against the county’s report and every district against its own precincts. A report that can’t be read or doesn’t add up is not posted, and the page keeps the last one that did.

• A primary choice is not a November vote. People cross over, and many November voters skip primaries.
• Most mail voters are 65 or older, so early numbers lean older than the full electorate.
• Redacted ballots and voters who skip primaries add no party information, which is why the range is wide.
• When the county blanks a voter’s history, older rosters restore most of it, but a few voters may sit one tier too low.
• Small areas can swing a lot on a few ballots.

• In Texas, voter files and lists of who returned a mail ballot are public records that anyone can request. They show whether someone voted and which primary they chose, never how they voted. Ballots are secret.
• No names or addresses appear on this site. It shows only counts for precincts, districts and the county.
• The page holds totals, not individual ballots: counts for each precinct and day by party history, and separate counts by age, sex and registration period. Any count from 1 to 4 shows as “<5”, and an area with fewer than 5 ballots shows no breakdown, estimate or map color.
• Ballots the county redacts stay unidentified.

Every record comes directly from official government sources (Tarrant County Elections Administration, the Texas Secretary of State, the Texas Legislative Council, the U.S. Census Bureau, the cities of Fort Worth, Arlington, and Grand Prairie, and Fort Worth ISD). The records are analyzed as published, with nothing added, altered, or invented.

While the gathering of records and the design mockups are organic, Claude was used to write the code that sorts and analyzes the data, and to code with the mockups to assist in UI/UX development, using Python (pandas, NumPy, GeoPandas, Shapely, openpyxl), Poppler, matplotlib, HTML, CSS, and JavaScript.

Given the water, energy, and carbon footprint of using AI, donations will be made to organizations working to hold AI and data center developers accountable for their use of precious resources.