A new Action Network survey of 2,495 prediction market users, who trade on platforms such as Kalshi or Polymarket, finds that 38% have traded against their own stated political preference for profit. Among 18-24 year-olds, that number climbs to 55%, more than double the 27% reported by traders 55 and older.
Key findings
- 38.3% of prediction market users have traded against their own political preference for profit, rising to 54.8% among 18-24 year-olds and falling to 26.5% among those 55 and older.
- 57.3% of prediction market users plan to trade on this year's midterms. Among users who specifically trade politics markets, that jumps to 83.5%.
- 63.0% of prediction market users think Democrats will flip the Senate.
- Half of prediction market users (50.8%) say prediction markets are more accurate than polls, and 48.6% say the same of betting odds. Only 16.5% say prediction markets beat none of the three.
- California traders are the most willing to trade against their own politics for profit, at 45.3%, against 28.0% in Illinois.
Trading against your own politics for profit
Among the 2,495 Americans who trade prediction markets, 38.3% say they've traded on an outcome they didn't personally want to see, purely because it looked like the more profitable position. That leaves 61.7% who say they haven't.
The age split is the sharpest divide in the data:
| Age | Traded against own politics |
| 18-24 | 54.8% |
| 25-34 | 44.4% |
| 35-44 | 37.5% |
| 45-54 | 34.9% |
| 55+ | 26.5% |
Willingness to trade against your own politics also tracks closely with which markets someone trades. Politics-market traders are the most likely to say they've done it (62.3%), followed by tech traders (61.1%) and weather traders (56.1%). Sports traders, who make up the bulk of the 2,495, sit lowest at 38.7%.
Men are somewhat more likely than women to say they've done this (39.7% versus 35.9%). Living arrangement matters too: traders living with a partner report the highest rate at 42.4%, while divorced traders report the lowest at 27.3%.
If you plan on trading on the midterms this year, check out our Polymarket promo code before you do.
Midterm trading plans
57.3% of prediction market users say they'll be trading on this year's midterms, against 42.7% who say they won't. Interest is highest among 18-24 year-olds (61.1%) and lowest among those 55 and older (52.5%), though the gap between age groups is narrower here than it is on the politics-for-profit question.
The category someone already trades is the bigger predictor. Among users who trade politics markets specifically, 83.5% plan to trade the midterms. That falls to 79.2% for tech traders, 75.4% for weather traders, 72.0% for finance traders, 70.8% for entertainment traders and 58.5% for sports traders.
Who thinks Democrats flip the Senate
63.0% of prediction market users think Democrats will flip the Senate this year, against 37.0% who think they won't.
State views on a Senate flip
This state-level breakdown comes from the wider survey sample, the original group of Americans surveyed (4,398), including both prediction market users and people who don’t use prediction markets, rather than from the 2,495 prediction market users alone. The figures below reflect US states with at least 85 respondents in that wider sample (see methodology for how this breakdown differs from the rest of the piece).
| State | Think Dems flip Senate |
| Maryland | 73.4% |
| Georgia | 70.5% |
| Virginia | 67.6% |
| Michigan | 67.3% |
| Pennsylvania | 67.2% |
| California | 66.3% |
| Wisconsin | 65.6% |
| New York | 65.4% |
| Florida | 62.4% |
| North Carolina | 62.3% |
| Illinois | 61.8% |
| Texas | 60.4% |
| New Jersey | 58.5% |
| Ohio | 58.5% |
| Tennessee | 57.6% |
| Indiana | 56.1% |
| Arizona | 55.5% |
Maryland is the highest of the reliably sampled states at 73.4%, Arizona the lowest at 55.5%, an 18-point spread. No state in this group drops below half, so even the most skeptical state still shows a majority expecting a flip.
Confidence in a flip doesn't track cleanly with who currently holds a state's Senate seats. Maryland, Georgia, Virginia, Michigan, California, New York, Illinois, New Jersey and Arizona are all represented by two Democratic senators, but those nine states span the entire range in the table, from the highest reading to the lowest. Arizona is the clearest outlier: a swing state with two
Democratic senators is the least confident of any state that Democrats flip the chamber nationally. Georgia, a similarly competitive state with the same party holding both seats, sits near the opposite end at 70.5%.
The six states with two Republican senators, Florida, North Carolina, Texas, Ohio, Tennessee and Indiana, cluster more tightly between 56% and 62%. None drops below half, meaning Americans in solidly Republican-held states still lean toward expecting a Democratic Senate flip. Pennsylvania and Wisconsin, the two states in the sample with a split delegation, land higher still, at 67.2% and 65.6%.
The survey asked about control of the Senate nationally, not the outcome of each respondent's own state race, so there's no reason a state's current delegation should predict the answer directly. The Arizona-Georgia gap suggests something more specific than national party control is shaping local sentiment, though the survey doesn't say what.
Do traders trust prediction markets over polls and odds?
Asked whether prediction markets are more accurate than polls, pundits or betting odds (respondents could pick more than one), prediction market users came down clearly in favor of the format:
| Prediction markets are more accurate than… | Share who agree |
| Polls | 50.8% |
| Betting odds | 48.6% |
| Pundits | 20.4% |
| None of these | 16.5% |
Just over half of users say prediction markets beat polls, and roughly half say the same of betting odds. Only 16.5% say prediction markets don't outperform any of the three.
State differences
Sample sizes outside the largest states get thin fast, so this section only covers states with at least 85 respondents among the 2,495 prediction market users. Smaller states are excluded entirely rather than reported on a handful of responses (see methodology).
California stands out on both measures that have state-level detail. It has the highest share of traders willing to trade against their own politics for profit (45.3%) and the highest share planning to trade the midterms (61.4%).
Illinois sits at the other end on both counts: 28.0% say they've traded against their own politics for profit, the lowest of any state with a usable sample, and only 47.0% plan to trade the midterms, also the lowest.
| State | Traded against own politics | Plans to trade midterms |
| California | 45.3% | 61.4% |
| New York | 41.1% | 58.3% |
| Florida | 40.4% | 54.3% |
| Texas | 40.3% | 59.3% |
| Ohio | 39.4% | 57.4% |
| Georgia | 37.1% | 56.2% |
| Pennsylvania | 35.2% | 59.3% |
| North Carolina | 34.5% | 58.6% |
| Michigan | 32.9% | 54.1% |
| Illinois | 28.0% | 47.0% |
Methodology
The research was commissioned by Better Collective and carried out online by Research Without Barriers (RWB). All surveys were conducted between October 2 and October 6, 2026. The wider survey sampled 4,398 Americans in total, of whom 2,495 said they trade on prediction markets.
Which sample each section uses:
- Trading against your own politics for profit and Midterm trading plans, including their state breakdowns in “State differences,” were questions asked only of the 2,495 prediction market users. These figures are a direct result from that group, not a subset of a larger ask.
- Who thinks Democrats flip the Senate and Do bettors trust prediction markets over polls and odds? were questions asked of the full 4,398-person sample. The percentages reported for “prediction market users” in these two sections are the subset of those 4,398 responses that came from the 2,495 prediction market users, filtered out after the fact, not a question fielded to that group alone.
- The state-by-state breakdown in State views on a Senate flip reflects all 4,398 respondents in each state, not filtered to prediction market users, since no state-level cut by trading status exists for that question.
The question on prediction market accuracy (polls, pundits, betting odds) allowed respondents to select more than one option, so the reported figures do not sum to 100%.
Both state-level breakdowns apply the same rule, at least 85 respondents, to their respective sample. Among the 2,495 prediction market users, that yields 10 qualifying states: California, Texas, Florida, New York, Pennsylvania, Illinois, Ohio, Georgia, North Carolina and Michigan. Among the full 4,398-person sample, it yields 17: Maryland, Georgia, Virginia, Michigan, Pennsylvania, California, Wisconsin, New York, Florida, North Carolina, Illinois, Texas, New Jersey, Ohio, Tennessee, Indiana and Arizona. All other states, in both cases, returned samples too small to report reliably and have been excluded from this piece.
All research conducted adheres to the MRS Codes of Conduct (2023) in the UK and ICC/ESOMAR World Research Guidelines (2023). RWB is registered with the Information Commissioner's Office and is fully compliant with the General Data Protection Regulation (GDPR) and the Data Protection Act (2018).
You can find the full dataset here.
