Texas 2026: the polls

The average is deliberately simple, so it can be checked by hand. For each race, on each day, it takes every poll whose fieldwork ended in the previous 30 days and keeps only each pollster’s most recent one, so a pollster that releases often does not outvote one that releases rarely.

Each poll’s weight is the square root of its sample size (samples above 2,000 count as 2,000), halved for every 14 days since its fieldwork ended. The average share is the weighted mean of the published shares, and the margin is the Democrat’s average minus the Republican’s. Nothing is adjusted for a pollster’s record or lean.

Polls of all adults, and polls sponsored by a party, a campaign or a partisan group, are listed but not averaged. Where a poll reports both likely and registered voters, the likely-voter figures are used. When no independent poll ended within 30 days, the page shows no average rather than a stale one.

Polls are entered by hand from each pollster’s own release, which every row links to. A poll is a snapshot with a margin of error of several points for each candidate, and an average narrows that error without removing it; a lead of a few points in the average is not a forecast of the result.