Prediction markets and polls answer different questions: a poll measures what people say they will do today, while a prediction market prices the chance of an outcome, folding polls, news and everything else traders know into one number that moves in real time. Studies of US elections have generally found market prices at least as accurate as poll-based forecasts at the same distance from the vote, with notable misses for both. On Nimb, election markets show that number, its move and the money behind it, as on the politics hub.
What is the difference between a poll and a prediction market?
| Poll | Prediction market | Forecasting model | |
|---|---|---|---|
| Measures | Stated vote share in a sample | Chance of an outcome | Chance, from polls and other data |
| Updates | When fieldwork ends | Every second | Daily or so |
| Inputs | Answers from a sample | Anything traders know, with money at stake | A stated method |
| Known weaknesses | Sampling error, non-response, turnout | Thin money, biased trader base, long-shot bias | Model choices, polling errors |
| Transparent method | Usually published | The price is public; the reasons are not | Usually published |
A 52% vote share in a poll and a 52% chance in a market are not the same claim. A candidate who leads by 4 points in polls may have a 75% chance to win; a market price compares with a forecast model, not with a poll's share.
Which is more accurate?
- In an early study, Berg, Nelson and Rietz compared the Iowa Electronic Markets with 964 polls in US presidential elections from 1988 to 2004 and found the market closer to the result 74% of the time (International Journal of Forecasting, 2008).
- Later work found the edge shrinks when polls are turned into probabilities properly rather than compared raw (Erikson and Wlezien, 2008), and both markets and models missed in some years.
- In 2024, Polymarket's prices favoured the eventual winner of the US presidential election by more than most poll-based forecasts did in the final weeks. One election is one data point, not proof.
Across many markets, what matters is calibration: do 70% prices come true about 70% of the time? See Polymarket calibration and are prediction markets accurate?
Prediction markets vs forecasters and superforecasters
Superforecasters (people selected for accuracy in forecasting tournaments, such as those run for the US intelligence community's IARPA programme) and crowd platforms compete with markets on the same questions. Research from those tournaments found that teams of skilled forecasters, aggregated well, matched or beat prediction markets on many geopolitical questions (Atanasov et al., Management Science, 2017). Markets have the edge on speed and on questions with a lot of money behind them; skilled forecasters on questions too thin or too new to trade well.
How should I read a market next to the polls?
- Compare the market with a forecast model's probability, not with a poll's share.
- Look at the move after a new poll: a market that does not react may already have priced it, or may be thin. Reading a price chart.
- Check the money: a 60% price on $20k traded is weaker evidence than on $20M. Volume.
- Check the rules: "wins the election" and "is inaugurated" are different contracts.
Frequently asked questions
Are prediction markets better than polls?
They answer a different question. As forecasts of who wins, markets have done at least as well as poll-based forecasts in studied US elections, with misses for both.
Why do markets and polls disagree?
Polls give a share; markets give a chance and include other information. Disagreement can also come from a thin market or a trader base that leans one way.
Were prediction markets right about the 2024 election?
Polymarket's prices favoured the winner in the final weeks. One election cannot show a market is reliably right; calibration over many markets can.
Last reviewed 8 November 2026 · Data from Polymarket · Not financial advice · Nimb is not affiliated with Polymarket.