What are prediction markets? Prediction markets are exchanges where people trade contracts that pay out based on future events, so the market price aggregates dispersed private beliefs into a single probability estimate.
A prediction market is an exchange where people buy and sell contracts that pay out based on the outcome of a future event — so the trading price becomes a collective, real-time probability estimate. The theoretical foundation is Friedrich Hayek's 1945 argument that prices aggregate knowledge no single mind holds: each trader brings private information, and the price summarizes all of it. The Iowa Electronic Markets, running since 1988, provided the classic evidence — Berg, Nelson, and Rietz (2008) found the market was closer to the eventual outcome than 74% of 964 national polls, and outperformed polls most clearly at horizons more than 100 days out. In the 2024 US elections, markets such as Polymarket and Kalshi signaled the presidential winner with higher confidence than polling aggregates in the final week and absorbed late news faster, though accuracy varied and thin down-ballot markets showed larger errors; research suggests combining market prices with polls beats either alone. In the decision-making lifecycle — attend, frame, generate, investigate, aggregate, deliberate, allocate, choose, implement, monitor — prediction markets are an aggregation mechanism: they compress many people's beliefs into one number. But a price aggregates <em>beliefs</em> while discarding the <em>reasoning</em> behind them, and voting aggregates preferences rather than beliefs. Argumentree aggregates judgment differently — through multi-dimensional rating that produces a consensus score while preserving the full argument trail, so the decision keeps its reasons.
What Are Prediction Markets?
A prediction market turns a crowd's scattered beliefs into a single price — a live probability estimate of whether some future event will happen. Here is how they aggregate information, how accurate they really are, and where they break down.
Last updated: 2026-07-18
TL;DR
A prediction market lets people trade contracts that pay out on a future event, so the price becomes a collective probability estimate. The idea rests on Friedrich Hayek's insight that prices aggregate dispersed knowledge. The evidence is real but bounded: the Iowa Electronic Markets beat most polls over 1988–2004, and 2024's markets read the presidential race confidently — yet thin markets, manipulation, and legal limits all constrain them. Crucially, a market aggregates beliefs and throws away the reasoning. Argumentree aggregates judgment through rating instead, and keeps the argument trail intact.
A Brief History of Prediction Markets
The idea that a price can forecast the future runs from mid-century economics through decades of empirical testing to today's high-volume online markets.
Friedrich Hayek publishes The Use of Knowledge in Society, arguing that the price system aggregates information dispersed across countless individuals that no central planner could ever gather. It is the theoretical root of prediction markets.
The University of Iowa launches the Iowa Electronic Markets, a real-money research exchange for election forecasting, operating under regulatory no-action relief. It becomes the most-studied prediction market in the world.
Robin Hanson introduces the logarithmic market scoring rule (LMSR), an automated market maker that keeps thin markets liquid — a mechanism widely used by later platforms.
Philip Tetlock publishes a roughly two-decade study of 284 experts and about 28,000 probability judgments, finding the average expert barely beat simple extrapolation — sharpening the question of who, or what, forecasts well.
Berg, Nelson, and Rietz show the Iowa Electronic Markets beat 74% of 964 national polls over five presidential elections, and outperformed most clearly at long horizons.
In IARPA's forecasting tournaments, Tetlock's Good Judgment Project identifies "superforecasters" — ordinary people whose calibrated, frequently updated forecasts reportedly outperformed intelligence analysts with access to classified data.
Tetlock and Dan Gardner's Superforecasting popularizes the traits behind accurate prediction: probabilistic thinking, frequent belief updating, and actively open-minded reasoning.
High-volume markets such as Polymarket and Kalshi trade the US elections at scale, and US courts open the door to regulated election contracts. The markets read the presidential race confidently in the final week.
Post-2024 analyses find markets incorporated late information faster than polls but showed larger errors in thin down-ballot markets — and that blending market prices with polls tends to beat either source alone.
How a Price Becomes a Forecast
The mechanism is simple to state and subtle in effect. A contract's price is the crowd's probability, updated continuously as money changes hands.
Contracts that pay on outcomes
A typical contract pays a fixed amount (say $1) if an event happens and nothing if it doesn't. If it trades at 63 cents, the market is implicitly saying the event has roughly a 63% chance.
Traders reveal private information
Anyone who believes the price is wrong can profit by trading against it — so people are paid to reveal what they privately know. Hayek's dispersed knowledge gets pulled into a single number.
Continuous updating
As news breaks, traders move the price immediately, which is why markets often absorb late-breaking information faster than periodically published polls or models.
Market makers keep it liquid
Automated mechanisms like Hanson's logarithmic market scoring rule (2003) or a continuous double auction ensure there is always a price, even when few people are trading.
How Accurate Are Prediction Markets?
The track record is genuinely good — but the honest version comes with conditions attached.
The Iowa Electronic Markets record
Berg, Nelson, and Rietz (2008) found the market was closer to the eventual result than 74% of 964 national polls across the 1988–2004 presidential elections, and beat the polls most decisively at horizons more than 100 days before the vote.
2024: confident but uneven
In 2024, Polymarket and Kalshi signaled the presidential winner with higher confidence than polling aggregates in the final week and incorporated late news faster — but accuracy varied by platform and race, with larger errors in low-liquidity down-ballot markets.
Markets plus polls beat either alone
Recent research suggests the strongest forecasts come from combining market prices with polling aggregates rather than relying on markets by themselves — a reminder that markets are one signal, not an oracle.
Superforecasting: The Human Complement
Markets aggregate a crowd; forecasting research asks who inside that crowd is worth listening to. The two are complementary.
Most experts forecast poorly
Tetlock's Expert Political Judgment (2005) found the average expert, across roughly 28,000 judgments, barely beat simple rules — and confident, famous experts often did worse.
But some people forecast well
In IARPA's tournaments, the Good Judgment Project's "superforecasters" outperformed the base rate and even some analysts with classified access — by thinking in probabilities, updating often, and staying actively open-minded.
A field of its own
How individuals achieve calibration — and how to measure it — is a distinct topic in its own right, part of the forecasting and calibration literature that sits alongside, but apart from, market-based aggregation.
Where Prediction Markets Fail
A market is only as good as its liquidity, its rules, and its participants. The known failure modes are well documented.
Thin markets
When few people trade, a handful of orders can swing the price, and the "crowd" is too small to be wise. Down-ballot and niche markets are especially vulnerable.
Manipulation
A trader with an agenda can try to move a price for signaling value; the research generally finds manipulation attempts are corrected quickly in liquid markets, but thin ones are more exposed.
Favorite–longshot bias
Markets tend to slightly overprice unlikely "longshot" outcomes and underprice heavy favorites — a systematic distortion long observed in betting markets.
Legal and access limits
Real-money prediction markets face regulatory restrictions that limit who can participate and which questions can be traded — thinning the very crowds that make them accurate.
Beliefs without reasons
A market outputs a probability, not an explanation. It can tell you the crowd thinks something is 63% likely, but not why — so it cannot be interrogated, taught, or audited the way a documented argument can.
Markets, Votes, and Deliberation: Three Ways to Aggregate
Prediction markets are one of several mechanisms for turning many minds into one answer. They differ in what they aggregate and what they preserve.
Markets aggregate beliefs
A price expresses how probable the crowd thinks an event is. It is fast and self-correcting, but it compresses everything into a single number and discards the reasoning.
Voting aggregates preferences
Voting systems combine what people want rather than what they believe is true — and how you aggregate preferences (majority, ranked-choice, and other rules) can change the outcome, a problem studied in social choice theory.
Deliberation aggregates reasons
Structured deliberation combines arguments and evidence, producing not just a decision but the reasoning behind it — the part markets and votes throw away.
How Argumentree Aggregates Judgment
Argumentree is a deliberation tool, not a market. It aggregates judgment while keeping the reasoning a price would discard:
Rating instead of pricing
Participants rate arguments on helpfulness, clarity, accuracy, and completeness; those ratings aggregate up the pro/con tree into consensus scores — a collective judgment, like a market price, but attached to specific reasons.
The reasoning is preserved
Where a market gives you a number, Argumentree gives you the number and the argument behind it — every claim, objection, and rating kept in a full audit trail.
Independent input first
Contributions are made independently and asynchronously, protecting the diversity and independence that make any aggregation mechanism — market, vote, or rating — trustworthy.
A decision you can interrogate
Because the reasons are on the record, a stakeholder can ask why the group concluded what it did — something a bare probability estimate can never answer.
Frequently Asked Questions
What is a prediction market?
A prediction market is an exchange where people trade contracts that pay out based on the outcome of a future event, so the market price aggregates dispersed beliefs into a single probability estimate. The idea rests on Friedrich Hayek's 1945 argument that prices aggregate knowledge no single mind holds.
How accurate are prediction markets?
Berg, Nelson, and Rietz (2008) found the Iowa Electronic Markets were closer to the outcome than 74% of 964 national polls across the 1988–2004 presidential elections, especially at long horizons. In 2024, markets like Polymarket and Kalshi read the presidential race confidently, but accuracy varied and thin down-ballot markets showed larger errors — and research suggests combining markets with polls beats either alone.
How do prediction markets aggregate information?
Each trader who thinks the price is wrong can profit by trading against it, so people are paid to reveal what they privately know. Their trades move the price, and the price continuously summarizes the crowd's collective probability estimate — Hayek's dispersed knowledge pulled into one number.
What is the difference between prediction markets and polls?
Polls ask people what they intend or believe at a moment in time; prediction markets have people stake money on outcomes, so prices update continuously and reward accuracy. Markets often absorb late news faster, but they can be thin or manipulated, and combining both signals tends to forecast better than either.
What are superforecasters?
Superforecasters are individuals identified in Philip Tetlock's Good Judgment Project (during IARPA's 2011–2015 tournaments) whose forecasts consistently outperformed the base rate and even some analysts with classified access. They think in probabilities, update their beliefs frequently, and reason in an actively open-minded way.
What are the limitations of prediction markets?
Prediction markets can be inaccurate in thin, low-liquidity markets, are subject to manipulation and favorite–longshot bias, and face legal restrictions that limit participation. They also output a probability without the reasoning behind it, so the result cannot be interrogated or audited the way a documented argument can.
References & Further Reading
Hayek, F. A. (1945). The Use of Knowledge in Society. American Economic Review, 35(4), 519–530.
The theoretical foundation: prices aggregate dispersed knowledge.
Berg, J., Nelson, F., & Rietz, T. (2008). Prediction Market Accuracy in the Long Run. International Journal of Forecasting, 24(2), 285–300.
The Iowa Electronic Markets vs. 964 polls — the 74% result.
Hanson, R. (2003). Combinatorial Information Market Design. Information Systems Frontiers, 5(1), 107–119.
The logarithmic market scoring rule (LMSR).
Tetlock, P. E. (2005). Expert Political Judgment: How Good Is It? How Can We Know? Princeton University Press.
The 284-expert study on the limits of expert forecasting.
Tetlock, P. E. & Gardner, D. (2015). Superforecasting: The Art and Science of Prediction. Crown.
The Good Judgment Project and the traits of accurate forecasters.
Wolfers, J. & Zitzewitz, E. (2004). Prediction Markets. Journal of Economic Perspectives, 18(2), 107–126.
A widely cited survey of how prediction markets work and perform.
View source →Related Concepts
Aggregate judgment — and keep the reasons
A market gives you a number; Argumentree gives you a consensus score and the argument behind it. Aggregate your team's judgment through structured rating, with the full reasoning on the record.
Start Free Trial