Texas 2026 ForecastSenate · Governor · Early voting

Sources and methodology

Where every number on this site comes from, and how we turn it into a forecast. This is an independent statistical project, not affiliated with any campaign, party or news organization.

Sources

Data	Source	Used for
Past election results by county, 2014–2024	Texas Legislative Council	County patterns, backtests, past-result maps
Voter registration, daily early-vote turnout, election-night results, 2026 primary results	Texas Secretary of State	Turnout, early-vote comparisons, the live needle on Nov. 3
2026 polls	Public poll releases, as compiled on Wikipedia's Senate and governor race pages	The polling average at the center of the forecast
Betting-market odds	Kalshi and Polymarket public prices	Shown for comparison only; not used in the model
2022 early voting, every county, every day	Texas Secretary of State data, as compiled by The Texas Tribune's 2022 tracker	"Same day in 2022" comparisons
2018 early voting, 30 largest counties, every day	Texas Secretary of State 2018 daily reports	"Same day in 2018" comparisons
Same-day (live) in-person turnout	Travis and Montgomery county live trackers (ElectionLink, including their 2022 and 2024 archives); Dallas County's vote-center map; Bexar County's early-vote dashboard	Live today; unofficial until the state posts its numbers
Harris County early voting	Harris County Clerk's daily record of early-voting check-ins, updated each evening	Used for Harris (in person and mail) when the state's file is behind
Denton, Collin and Comal early voting	Each county's own daily reports: Denton's totals by location and returned mail ballots, Collin's totals by site, and Comal's daily lists of in-person voters (counted; only the count is kept)	Used for those counties when the state's file is behind
Tarrant County early voting	Tarrant County's daily in-person totals, posted each morning	Used for Tarrant when the state's file is behind
Who voted early, and in which 2026 primary	Texas Secretary of State daily lists of early voters (2026 general and March 2026 primaries)	"Early voters by 2026 primary." Only county totals are kept; no names or voter IDs are saved or published.
Mail ballot history	Harris County Clerk official canvass reports, 2018–2024	How mail ballots lean when they are counted first on election night
Past poll averages	538, RealClearPolitics, 270toWin, Decision Desk HQ; our own average of October polls for 2018, when none was published	"How accurate have Texas polls been?"
County shapes	U.S. Census Bureau, via us-atlas	All maps

How the forecast works

1. Start from the polls

We average every head-to-head poll from the last 60 days. Recent polls count more (a poll loses half its weight every 14 days), larger polls count a little more, and polls paid for by a party or campaign count less. If one pollster releases several polls, those polls share their weight so a single firm cannot dominate.

2. Measure how wrong polls can be

Texas polls have picked the winner in every recent race, but the final margin has been off by 2 to 5 points, and not always in the same direction. We let the true result differ from the polls by about that much, with a small chance of a bigger miss.

3. Spread the statewide number across 254 counties

Each county's lean relative to the state comes from past races for the same office: the 2018 and 2024 Senate races for Senate, and the 2018 and 2022 governor races for governor. Counties are grouped into six regions (big cities, big suburbs, outer suburbs, South and West Texas, mid-size cities, rural) that tend to move together, so a region can swing more or less than the state. 2026 primary turnout nudges each region slightly.

4. Estimate turnout

Each county's expected turnout is its 2018 turnout rate (the last high-energy midterm) applied to 2026 registration, with random variation by county, region and state.

5. Run 50,000 elections

Each simulated election draws a polling miss, regional swings, county noise and turnout, then adds up every county's votes. A candidate's chance of winning is the share of the 50,000 simulations they win. "74%" means the candidate wins about 3 times in 4, not that they get 74% of the vote.

6. Check it against the past

Run as if it were 2018 and 2024, using only data available then, the model favored the eventual winner both times, the actual margin fell inside its 80% range both times, and it picked the right winner in 251 of 254 counties in 2018 and 243 of 254 in 2024. Two backtests are not proof, but they show the model is not badly off.

Election night

Before votes are counted, the needle shows the forecast. On Nov. 3 it updates every 10 minutes from the state's county-by-county results. Each simulated election is scored against the votes counted so far; simulations that match the count count more, and the needle moves to the share of matching simulations each candidate wins.

Mail ballots and early votes are usually counted first and lean Democratic in Texas cities. We expect that and correct for it, so an early lead in the first results does not move the needle more than it should.

Early voting

Turnout, not votes. Counties and the state report how many people voted, never whom they voted for. Candidate totals stay sealed until polls close on Nov. 3.
Live today uses each county's own website during the day. A county appears only once it has posted numbers for today. These counts are unofficial; the state's numbers replace them once posted.
Comparisons are raw vote counts against the same day of early voting in past years: "142% of 2022 day 1" means 42% more in-person votes so far today than in that county's whole first day in 2022.
Each county's source is listed under Live today ("Where each county's numbers come from"), with the latest day it covers. Counties are compared with 2022 through that same day. If a key county falls behind, or a county's own source stops working, an alert goes out.
Incomplete numbers are left out rather than shown as a false drop: a county whose in-person count is far below 2022 at the same point, or far below the typical county's pace, is marked "not posted yet." If only mail ballots are missing, the county is compared in person with in person. Totals never go down: a county is never shown below what the site already published.
The map filter shows today's in-person votes, mail ballots, or both together, each against the same point in 2022, using the same numbers as Live today.
Which party's voters are turning out faster? Texas has no party registration, so we use the primaries: of the people who voted early or by mail in the March 2026 Democratic primary, what share have voted early this fall, and the same for the Republican primary. Only counties whose lists of this fall's early voters are in are counted. Primary-day voters are not in the state's free lists. Primary voters are a clue to party, not votes.
Key counties are grouped by the role they play: Democratic vote banks, battlegrounds, and Republican vote banks.

Assumptions and limits

Mail ballots in Harris County ran about 30 points bluer than the county in 2024. Outside Harris we assume a smaller lean (about 12 points on the margin); we have no county records to check it.
There are no published 2018 poll averages for these races, so the 2018 rows in the poll-accuracy table are our own averages of October polls.
For 2024, we have day-by-day county early-vote numbers only for counties with archived live trackers.
Third-party candidates are left out; all margins are between the two major candidates.
Two backtests are a small sample. Every election forecast faces this, because elections are rare.

Updates and checks

The site refreshes automatically: every hour from 7 a.m. to 11 p.m. Central during early voting, three times a day otherwise, and every 10 minutes on election night. Each refresh runs a set of automated tests and sanity checks (for example, a poll average that jumps unrealistically or a county count that goes down). If any check fails, nothing is published and the site keeps its last good version.