Prediction-Market Information and Voter Behavior

This study has two interventions and one control condition. The Odds condition provides participants with current prediction market odds for the relevant 2026 U.S. Senate election and a brief explanation of these odds. The Polls condition provides the candidates’ polling averages from the Average row of the Aggregate polls table on the relevant Wikipedia election page, with a source link. Participants receive a brief explanation. The Control condition provides neither prediction-market odds nor polling information.

2026-10-09

2026-11-20

The primary outcomes are (i) candidate preference; (ii) turnout intention; and (iii) intended donation to the preferred candidate’s campaign.

The secondary outcomes are (i) the perceived win probability of candidates; (ii) the expected margin of victory; and (iii) self-reported news sources used.

This study uses a multiple-wave, between-subject, three-arm randomized online survey design. Participants are recruited through online research platforms and first complete a brief screening survey. Eligibility requires participants to be at least 18 years old, reside in Michigan, Ohio, Georgia, North Carolina, or Texas, intend to vote in the relevant 2026 U.S. Senate election, and report that they are still undecided about who to vote.
Eligible participants are randomly assigned within each state to Control, Odds, or. Participants remain in the same condition throughout all survey waves.
In Wave 1, all participants first report candidate preference, winning probabilities, expected victory margins, turnout intention, intended donations, and news sources. The Odds group then sees Polymarket odds, while the Polls group sees polling averages. Both information groups then repeat the belief, preference, and intended-behaviour measures. Control participants complete only the baseline measures in Wave 1.
In Waves 2 through 4, participants remain in their assigned condition. The Odds and Polls groups receive updated odds and polling averages, respectively. All participants complete the belief, preference, and intended-behaviour measures and report their news sources. The displayed information and its source are recorded for each wave. Wave 4 ends with an optional question asking participants to explain any change in candidate preference over the past month.
Participants receive base compensation for each completed survey. Those who complete all four waves are eligible for a study bonus. One bonus-eligible belief response is randomly selected for each eligible participant. A 50/50 random draw selects either the $5 belief-based bonus based on Binarized Scoring Rule method or a drawing with a 0.5% chance of winning $500. Payment follows the rules presented with the selected question.

Not available

Computer-based

Subjects within each US state (strata) are randomly assigned to the three conditions.

No

Not applicable. Treatment is assigned at the individual-participant level and is not clustered.

The study targets at least 1000 participants completing all waves of the study.

The following sample sizes are planned. However, the final sample sizes may differ depending on attrition and logistical/technical constraints that may arise.

Control: at least 300 participants.
Odds: at least 500 participants.
Polls: at least 200 participants.