Methods and disclosure

How Free State Research conducts its public polls, and everything you need to check our work. We follow the disclosure elements of the American Association for Public Opinion Research's Transparency Initiative.

Latest release

Kansas Voting Research Poll · fielded October 5–10, 2026 · released October 13, 2026 · revised October 18, 2026 (see Revision below)

The toplines give every question in the exact wording and order respondents saw it.

Who paid for it and who did it

Conducted and paid for by Free State Research LLC, an independent Kansas research firm. No campaign, party or outside group sponsored or paid for this survey. If a future poll has a sponsor, we will name it on the release.

Who we surveyed

Population: Kansas registered voters.
Sample frame: the Kansas voter registration file. Voters were drawn in batches balanced by region, party, age and turnout history. Each phone number was checked as a mobile line before any text was sent.
Likely voters: registered voters who say they will definitely or probably vote in November. In the October poll, 10 of 574 respondents said 50-50 or less and are excluded from the likely-voter results.

How we reached them

Mode: a text message with a unique, single-use link to an online questionnaire (text-to-web). One link per invited voter; a link can't be answered twice.
Dates: October 5–10, 2026, with one reminder text to people who had not yet answered.
Completion: 25,564 invitations sent, 24,470 delivered, 574 completed (2.3% of delivered invitations).
Anyone who replied STOP was removed and is never texted again.

Precision

The margin of sampling error is ±4.1 percentage points at the 95% confidence level for both the registered-voter and likely-voter samples, and larger for subgroups and for questions asked of only part of the sample. Sampling error is not the only source of error in any poll; see Limitations below.

Weighting

Every respondent is matched to the voter file. Results are weighted by party registration, age, region, gender, education, race, 2024 presidential vote and 2022/2024 general-election voting history, so the sample matches our central estimate of the November 2026 electorate:

2018 + primary boost: Republicans 52.2%, Democrats 29.5%, unaffiliated and other 18.3%; 2024 Trump voters 54.5%, Harris voters 38.3%. It is the November 2018 electorate by party, from the Kansas voter file, with one point moved from Republicans to Democrats for this year's August primary. Ages follow the 2022 midterm electorate in the voter file.
Weights are trimmed so no respondent counts for less than 0.2 or more than 5 times the average.
Undecided voters are allocated by the way they lean, then by their choices in the other races, then by party, and otherwise in proportion to decided voters. We report the vote both before and after allocation.

We also tested weighting on whether each respondent voted in the August 2026 primary, where our respondents differ most from the electorate. It did not change the results meaningfully, and because it nearly halves the effective sample, it is not used in the published figures.

Turnout scenarios

Who votes is the biggest unknown in any election poll, so we don't hide our assumption. Every release shows each race under eight possible electorates, from a repeat of November 2022 to a strong Democratic year, and the toplines show the real Kansas electorates of recent elections for comparison. We commit to our turnout model before the election (we revised it once, on October 18; see Revision below) and will publish how each scenario compared with the actual result afterward.

Revision, October 18, 2026

We revised the October release on October 18. The October 13 version was weighted to the FSR Turnout Model (Republicans 51.6%, Democrats 31.6%, unaffiliated 16.8%), which assumed a more Democratic electorate than any Kansas general election on record. The revised release centers on the 2018 + primary boost electorate above, the best recent midterm for Democrats plus a small allowance for this year's primary. The FSR Turnout Model remains one of the eight published scenarios. The interviews are unchanged; only the weighting target changed.

October 13 (FSR Turnout Model): Hamilton 49%, Marshall 42%; Mann 48%, Kobach 45%; Masterson 46%, Holscher 46%.
October 18 (2018 + primary boost): Hamilton 47%, Marshall 43%; Mann 47%, Kobach 46%; Masterson 48%, Holscher 44%.
The original files: release, likely-voter toplines, registered-voter toplines.

A check against a known result

Reweighted to the November 2022 electorate, the October poll's likely voters give Kris Kobach 50.6% and Chris Mann 49.4% for Attorney General. The actual 2022 result was Kobach 50.8%, Mann 49.2% (Kansas Secretary of State). That's a check on the method, not a forecast: 2026 voters and 2026 candidates are different.

Data quality

Every completed response was matched to its voter-file record before it was counted.
Unique invitation links prevent duplicate or outside responses.
Numbers are line-checked before texting, and landlines are skipped.

Limitations

The survey reaches only voters with a mobile number on the voter file.
People who answer polls by text differ from those who don't. Our respondents are more likely than the electorate to vote in every election. We weight on 2022 and 2024 voting history to correct for this, along with party, age, region and other measured traits, but weighting can't correct for everything.
Results depend on who turns out. That's why we publish every scenario, not just our own.
A poll is a snapshot of the dates it was in the field, not a prediction.

Questions

Reporters and researchers can reach us at info@freestateresearch.com.