The Texas 2036 Voter Poll interviews about 1,000 voters each wave. Yet, Texas has about 18 million registered voters. Is our voter poll representative of statewide voters? Here are some common questions we receive, and why we feel confident about our results.
Is our voter poll sample representative of statewide voters?
Accuracy depends less on the share of the population you interview and more on how you select and weight your sample. A sample of about 1,000 voters can describe an electorate of millions, as long as the people interviewed are chosen and weighted to resemble that electorate. In fact, 1,000 respondents is common for statewide political polls and even some national polls.
We were careful in how we selected and weighted our sample. Our pollster, Baselice & Associates, Inc., ensured representation across Texas by creating interview quotas by media market, and by ZIP code for high population counties. The weighted distribution across the six media markets fell within about one point of each market’s share of the ballots cast in the most recent general election. We also weighted each poll so that the mix of respondents — sex, age, race and ethnicity, party, region, education — matches the Texas electorate. The weighted result roughly lines up by sex, race and ethnicity with Census estimates of who voted in Texas in 2024.
Figure 1. After weighting, the August 2026 poll was within 2 points of Census estimates of 2024 Texas voters by sex, race and ethnicity
Share of each group: August 2026 poll before and after weighting, and estimates of Texas voters in November 2022 and November 2024
Note: We weight the poll on every trait shown, so the weighted shares reflect the weighting targets. The exit poll uses different degree categories. Source: Texas 2036 Texas Voter Poll, conducted by Baselice & Associates; U.S. Census Bureau, Current Population Survey; ABC News exit poll.
How confident are we in this sample of 1,000 or more voters?
In August 2026, we interviewed 1,369 voters, but weighting makes some responses count more than others. That leaves us with about as much information as 1,000 interviews.
Each time you interview a different set of 1,000 Texas voters, you may get a slightly different number. The margin of error is how much that number moves. Our margin of error is 3.1 points, so a reported 62% means the real figure is very likely between 59% and 65%.
Narrowing the margin of error is feasible but it takes many more interviews. In our poll, about 2,100 interviews would bring the margin down from 3.1 to about 2.5 points, and about 3,200 interviews would bring it to 2 points. Cutting the margin in half takes about four times as many interviews. Each additional point of precision costs several times what the last one did, and for most of what we ask, a 3.1-point margin of error is precise enough to be useful.
How close does the poll come to what actually happens?
Survey researchers answer that question by comparing a poll against measures taken outside it: certified election results, official turnout records and large government surveys.
In our August 2023 poll, we asked how voters would vote on three constitutional amendments on that November’s ballot. Majorities favored all three, and all three passed. But leaving out respondents who were unsure, support in the poll ran five to 10 points above the certified result on each one (Figure 2).*
Figure 2. Leaving out unsure respondents, both polls ran ahead of the certified result on each 2023 amendment
Share in favor: range from the share for (left end) to for plus unsure (right end), and the share for among respondents who chose a side, in the Texas Voter Poll (August 2023) and the Hobby School poll (October 2023); certified share of votes cast in favor, Nov. 7, 2023
Source: Texas 2036 Texas Voter Poll, August 2023, conducted by Baselice & Associates; University of Houston Hobby School of Public Affairs, Texas Trends 2023; certified results from the Texas Secretary of State, compiled by the Texas Legislative Reference Library.
Our poll showed which way Texans were leaning, and it showed that correctly all three times. Its support numbers ran higher than the final vote counts.
That gap is typical of polling on ballot measures, and several things likely contribute to it. Our interviews ended 70 days before Election Day, and support for a ballot measure tends to slip as the vote nears. The University of Houston’s Hobby School of Public Affairs polled likely voters on the same amendments six weeks closer to the election and ran 8.0 points high, against our 8.4, so timing is likely only part of the explanation. The voters who decided the amendments were also a small group: 14.4% of registered voters cast a ballot in November 2023. In other words, our numbers describe what Texans support when we ask, not how an election will come out.
Most of what we ask never goes to a vote. Those three ballot measures were unusual for us. We designed this poll to track what Texans think about state issues over time. We ask what Texans want the state to do about water, schools and health costs — questions with no official count to check against. That is why how we build and weight the sample matters as much as it does.
Every poll we have run since 2020, with full methods, is at texas2036.org/pollingdata.
* The questions used the ballot wording in Senate Joint Resolution 75 (Proposition 6), House Joint Resolution 125 (Proposition 8) and Senate Joint Resolution 74 (Proposition 14) of the 88th Legislature. Online respondents saw “unsure” as an answer; phone respondents had to volunteer it.
