Skip to content
TexPolls

Methodology

TexPolls collects public election data, checks the sources, and explains the limits. We summarize polls and reported participation; we don’t make election forecasts.

Polling Data

The Headline Average

  • We take a simple average of likely-voter polls in our records dated on or after May 26, 2026, testing the current Democratic and Republican nominees.
  • We leave out polls sponsored by campaigns or political operatives and any poll that our review has excluded because of its method or ballot design.
  • When a poll publishes a result that includes undecided voters who say they lean toward a candidate (a “leaned” result), we use it. Otherwise, we use the first-choice result.
  • Each poll counts equally: we don’t weight by sample size or pollster rating. We chose that design to keep the average simple and easy to check. We average each nominee’s reported percentage and round to one decimal place, without redistributing undecided or other-candidate support. The margin is the difference between those two rounded figures.
  • There is no time window or extra weight for newer polls. We do not adjust for a firm’s typical difference from other polls or fit a trend model. We show each poll as reported.
  • One qualifying poll is labeled as one poll. With none, we show no polling headline. The average summarizes polls; it is not a forecast.
Details: Poll Selection and Display
  • Polls with fewer than 500 respondents statewide or 400 in a district are shown but not counted in the average. Our other review targets are a public source URL and published methodology, sample size, and margin of error. Live-caller, online-panel, and mixed-mode surveys are accepted. IVR-only surveys require an established firm with a track record.
  • Historical polls stay available. We note what each release does not disclose. Registered-voter polls can appear in tables, but never enter the likely-voter average. Appearing in a table does not guarantee eligibility for the headline average. Excluded polls never feed the matchup number on the homepage.
  • We name campaign, political-operative, think-tank, and advocacy sponsors when known. Sponsorship excludes a poll from the average when a campaign or political operative commissioned it; a newsroom, nonpartisan research, or think-tank sponsor does not by itself disqualify a poll.
  • Two-way and additional-candidate ballots can qualify when they test the same major-party nominees. We disclose the ballot because its design can affect results. Different major-party pairings are excluded.
  • The two July Lieutenant Governor polls from Texas Southern University and Texas Pulse / ReconMR remain excluded while we separately review whether their ballot designs can be compared. Adding another poll does not reverse that decision.
  • The October 13 GS Strategy Group / New River Strategies surveys with Cook Political Report are excluded because the release does not publish a margin of error, mode, sample source, or weighting.
  • ActiVote’s statewide samples fall below our 500-respondent minimum. YouGov’s September 28 to October 5 in-house poll and both Trafalgar polls have not published survey-specific methods, and SoCal Strategies has not published a margin of error. Those polls remain excluded from the average. YouGov’s October 26 to 29 poll discloses its sample frame, mode, and weighting, and counts. The ActiVote, YouGov, and Trafalgar polls appear in the tables with their limitations noted.
  • We adopted the leaner rule on October 21, 2026, following the approach 538 used and Silver Bulletin uses. Before that date, we used first-choice results and left out polls that published only a leaned result. When we adopted the rule, we re-entered earlier polls that publish a leaned result, so the whole average uses the same rule.
  • The June Texas Pulse / ReconMR wave was left out after an unannounced revision to its published results. The July wave passed a separate source review; the two July Lieutenant Governor records remain excluded for the ballot-design review above.
  • The general-election tables show undecided support only when recorded. A dash means it is not recorded; we never infer it by subtracting major-party support from 100. Other-candidate support stays separate.
  • General-election matchups show a margin of error when available. Primary horse-race bars omitted it to reduce clutter.
  • During the primary, Runoff Watch appeared when no candidate polled above 50%. Texas requires 50%+1 to win outright. The March primary and May 26 runoffs are decided, so that badge is no longer shown.

Two September SoCal Strategies polls are not shown while we review their source, which is no longer online.

Included Pollsters and Source Reports (43)

Any firm meeting the review criteria can be included. Linked names lead to survey releases in our records; unavailable original releases are noted. Each race’s polling tables and notes identify the recorded dates, sample, method, sponsor, and review limits for each wave. Missing disclosures stay noted.

Early Voting Data

  • The daily chart and county tables compare in-person votes only, matched to the same numbered early-voting day of the 2022 midterm: day 1 with day 1. The headline compares all early ballots, in person and by mail, with all of 2022’s early voting (5,497,027 ballots through the last day, November 4, 2022), rounded down. Missing dates do not shift later comparisons; optional 2024 presidential context stays separate.
  • Daily cumulative reports from the Texas Secretary of State count in-person early votes and mail ballots received through the stated date. These provisional participation counts do not identify candidate choice.
  • When a checked, same-date report from a county election office whose count we collect exceeds the state’s in-person figure by more than 5% of the county count, we add the difference to the displayed total. We name and link the county, show the state’s count, and keep mail unchanged. This is our provisional county-updated estimate.
  • Download the county-by-day CSV, observed revisions, and column guide. Free to use and share. Just credit TexPolls. Licensed under CC BY 4.0.
Details: Incomplete Reports and County Updates
  • Before a state report passes completeness checks, the statewide number is an unofficial count from the state’s newest file: in-person votes plus the state’s own mail count from that same file. A file without mail shows in-person votes only. It starts with the state’s county report. When a county’s figure is missing, out of date, or fails our checks, we can use the county office’s own posted count and name the source on the page. County offices self-label those counts as unofficial.
  • If one of the 10 largest counties by registered voters, or more than a small share of registered voters, lacks a usable count from either source, we show no statewide number for that date. We retain the most recent usable date and label it.
  • Once a state report passes, it becomes the dated base for county updates. A qualifying newer count can lead the page under the same county-coverage rules, with its own date. It includes the state’s mail count when that file has one.
  • We check each county report’s totals before using it. County updates do not replace the state report or hold it back; the state report’s completeness checks stay unchanged.
  • A checked county gap above 5% in the opposite direction is flagged rather than subtracted. Corrections based on the state’s own voter records are identified separately.
  • If a new state date fails checks, we retain the most recent passing report and county updates for that same date. A passing report can still include older counts from a few small counties.
  • Map colors and county-page rank badges use the state report. Downloads keep state and county-office figures in separate columns.
  • We list a county as behind in the state file when its latest daily gain there is under a tenth of its usual daily gain (under 2% on a Sunday, when the law requires only 6 hours), or when its count is under 1 per 1,000 registered voters. We name the largest of these under the statewide total and list them all on our turnout page. The list never changes a figure. Every county must offer voting on the last Saturday and Sunday of early voting, and many counties’ weekend votes first appear in Monday’s file, so a weekend total is a floor until then.
Details: Comparisons, Missing Data, and Downloads
  • Day 1 is compared with day 1, rather than the same calendar date or the number of available reports. Each year’s participation share uses its own registered-voter denominator.
  • Headline and chart comparisons use the same dated county updates. County changes use unrounded shares and are expressed in percentage points.
  • Missing dates remain gaps; one report is a snapshot, not a trend. An unavailable county comparison is distinct from zero. Historical 2022 data may be revised; current 2026 counts remain provisional.
  • The CSV has one running-total row per county on each accepted state-report date. Some small counties may still carry an earlier count. When corrected, the original state figure stays in its own column.
  • The revisions file records changes we observed to earlier figures after the first saved version. The column guide explains the fields.
  • Archived primary party totals describe the ballot selected, not party registration. Their daily comparison uses 2022 primary early voting; the county comparison uses full 2022 Governor-primary votes, labeled separately from November comparisons.

County Maps

  • All 254 Texas counties use boundaries from US Atlas.
  • The early-voting map shows the share of registered voters who voted early. The higher end of the legend means higher participation, not support for a party.
  • The 2024 presidential map uses the Texas Legislative Council’s official returns. It is historical context, not a forecast or a partisan breakdown of current early voters.
Details: Map Scales and Historical Comparisons
  • The participation maps use fixed bands. The turnout-page map, and the homepage map once the state completes a day, color all early ballots at 20%, 25%, 30%, 35%, and 40%. While the homepage shows a newer, unofficial count, its map counts in-person votes only and uses 10%, 20%, 25%, 30%, and 35%. Hatching means no usable rate; it does not establish a reporting delay. Zero votes with a valid registration denominator remains zero.
  • On Turnout, the optional comparison with the statewide rate can put a county in a lower relative category even as its own participation rises, if the state rises faster. The county table separately compares the matched 2022 midterm day.
  • The archived map’s 2022 baseline uses full Governor-primary candidate votes, both parties combined, from the Texas Capitol Data Portal. Those totals are distinct from 2026 primary early voting.

Election Results

Official returns come from the Texas Secretary of State. We distinguish three statuses:

  • Partial: counting continues. Candidates are labeled leading, not winning, following AP style.
  • Called: We mark a race called when the Associated Press or another major news outlet calls it, and we name the source. TexPolls doesn’t call races. The winner gets a checkmark.
  • Final: final reported totals. Certification requires a separate canvass record; a completed reporting percentage does not establish it.
Details: Primary Runoffs and Historical Turnout
  • Texas primaries require 50%+1 to win outright. Otherwise, the top two advance to a runoff. Candidate labels distinguish Advanced to Runoff and Eliminated.
  • The historical runoff dataset covers 2012 onward, with total ballots, registered voters, turnout percentage, and the marquee race. Its main source is the state’s historical turnout tables.
Details: How the Vote Shifted
  • The “How the Vote Shifted” charts on the results page use the two-party margin: Democratic votes minus Republican votes, divided by the votes for both, in points. Third-party and write-in votes are left out, so these margins differ slightly from margins of all votes cast.
  • Races from 2018 to 2024 use the official canvass from the Texas Secretary of State results archive. 2026 uses the state’s unofficial returns until its canvass. County sums match the statewide totals for every race.
  • The charts compare different offices, candidates, and electorates (midterm and presidential years). A county’s change in margin describes the county as a whole; it carries no information about which voters changed or why.

Race Ratings

We show a rating only when Cook Political Report, Sabato’s Crystal Ball, or Inside Elections rates the race. Otherwise, we mark it Not Rated. We don’t make our own ratings. When the forecasters disagree, the badge shows the rating given by most of the forecasters that rate the race, and the source line names those forecasters. If there’s no majority, we choose by hand, and the source line names the forecaster we show.

Campaign Finance Sources

  • Covered federal candidates’ receipts, disbursements, and cash on hand come from the Federal Election Commission.
  • State filings from the Texas Ethics Commission cover Texans for Greg Abbott (committee), Gina Hinojosa (candidate account), Mayes Middleton (filed as David M. Middleton II, candidate account), Nathan Johnson (filed as Nathan M. Johnson, candidate account), and historical Texans for Joan Huffman committee filings from January through June 2025. We show report-period contributions, spending, and cash on hand. Cash on hand is what TEC calls “total political contributions maintained,” which includes bank balances and investments that can be readily converted to cash. Each entry identifies the period and account scope.

Data Processing

  • We collect public records, check them against their sources, and review each update before we publish it. Our data-source inventory identifies the providers.
  • During November early voting, scheduled checks ran throughout the day. A state report became the base only after completeness checks; county updates and newer in-person-only counts followed the rules above.
  • Polls are added after source review, which may follow their release date. Automated return updates do not call races; election-night counts and race calls are separate.
Details: Publication and Freshness

We check records against official sources where they exist, store them as structured JSON, review them, and commit them to our repository. We publish an update once we approve it on the production branch.

Report dates and coverage periods show each series’ age. A site-wide update timestamp does not mean every poll, finance filing, or turnout report was refreshed.

Editorial Independence

We’re independent of political parties, campaigns, government agencies, and news organizations. We publish source data with its limitations, without candidate endorsements or our own election forecasts.

Methodology Questions

How does TexPolls select which polls to include?

TexPolls reviews each poll’s source, sample, method, and sponsor. The headline average uses likely-voter polls of the current major-party nominees dated on or after May 26, 2026, excluding campaign or political-operative polls and polls left out after review. See Poll Selection and Display for the criteria and exceptions.

How is Texas early voting data collected?

November early-voting totals come from Texas Secretary of State cumulative reports of in-person votes and mail ballots received. The daily chart and county tables compare in-person votes only, on the same numbered day of the 2022 midterm general early-voting period (day 1 with day 1), not the same calendar date, and use each election’s own registration denominator. The headline compares all early ballots, in person and by mail, with all of 2022’s early voting. Missing dates do not shift later comparisons. Optional 2024 presidential context is separate. Missing reports remain gaps; November participation counts do not identify party choice. Archived primary comparisons retain their separately labeled periods.

How are Texas election results reported on TexPolls?

Official results come from the Texas Secretary of State. Results are classified as Partial (leaders labeled 'leading' per AP style), Called (winner marked), or Final (final reported totals; certification requires separate canvass evidence). Texas primaries require 50%+1 to win; otherwise the top two advance to a runoff.

Does TexPolls average or weight polls?

Yes. We use a simple, equally weighted average of qualifying likely-voter polls; one poll is labeled as one, and no qualifying poll means no polling headline. See The Headline Average for the calculation, exclusions, and limitations.

AboutDashboard