2026 Ohio Statewide Poll — Methodology (Pre-Commitment)
Published October 23, 2026, before the start of data collection.
This document states, in advance of collecting any responses, how MeridianPoll will conduct, analyze, and report its 2026 Ohio statewide poll. It is published before fielding and archived with a timestamp so that our methodological choices cannot be said to have been made after seeing the results. Where a decision was fixed in advance, it is stated here; we will report the poll as described below regardless of what the findings show.
Survey and Field Design
Subject: A statewide survey of Ohio voters covering the 2026 elections for Governor and U.S. Senate, the generic U.S. House vote, and presidential job approval.
Mode: SMS text-to-web. Respondents receive a text message with a personalized link to a web survey.
Instrument: The complete questionnaire, including every question and response option in the order administered, is published separately.
Field window: Messages are sent within an 8:00 a.m. to 9:00 p.m. local-time window, in compliance with applicable rules.
Sampling
Frame: Registered Ohio voters who voted in the 2024 general election, drawn from the L2 voter file.
Contact data: Cell phone numbers at L2 confidence scores 1 and 2.
Non-Affiliated/Other Party voters: Because non-affiliated and other-party voters respond at a substantially lower rate than other voters, they are contacted at a higher rate, so that enough respond to yield a usable number of respondents. This is done to achieve adequate representation, not to over-represent them. They are weighted to their correct share of the electorate in analysis.
Exclusions: Records in states whose texting statutes we exclude on, identified by area code, are not contacted.
Likely-Voter Model
Turnout is modeled using a rule defined before data collection that combines the respondent's stated likelihood of voting with their intended method of voting. Respondents are classified into three nested turnout models:
Low model: the most restrictive screen.
Base model: our headline estimate.
High model: the most inclusive screen.
Response combinations that are logically incoherent are excluded. The complete classification rule and the set of excluded combinations are fixed in advance and held internally until the results are published, at which point the rule and the resulting counts are disclosed in full. This protects the integrity of the screen while the survey is in the field, and the rule is not adjusted after results are seen.
Weighting
Demographic and party weighting: Age, gender, race/ethnicity, education, and party are weighted to targets derived from the L2 voter file. Party weighting uses the party recorded in the voter file, not self-reported party.
Recalled-vote weighting: 2024 recalled presidential vote is weighted to the certified 2024 Ohio presidential result. This is the only variable not weighted to the voter file.
Self-reported party: Collected for descriptive reporting only; it does not affect weighting.
Reporting
Because the response rate is low and the composition of the 2026 electorate is uncertain, horse-race results are reported as a range across the Low, Base, and High turnout models rather than as a single point estimate. The Base model is the headline figure.
Results for a subgroup are reported only where the number of respondents supports it. Subgroups below a minimum are not reported; marginal subgroups are shown as directional, with respondent counts shown.
Impartiality
MeridianPoll is independent and is not conducting this poll on behalf of any campaign, party, or committee. We follow the pre-committed methodology described above and report results regardless of whether they are favorable to any candidate or party.