What Are Prediction Markets and How Do They Work?

Prediction markets turn uncertain future events into tradable contracts. A typical market asks a yes-or-no question, pays a fixed amount if the event occurs, and lets buyers and sellers trade before resolution. The price is often read as the crowd's implied probability—but that interpretation has assumptions and limits. This guide explains the mechanics, what makes a market informative, how settlement works, and where prediction-market signals fit beside stocks, polls, forecasts, and social data.

What a prediction market is

A prediction market is an exchange or platform where participants trade contracts linked to the outcome of an event. Questions can concern economic releases, elections, weather, policy, company milestones, technology, sports, or other objectively resolvable events. In regulated U.S. derivatives markets these products are commonly called event contracts.

The U.S. Commodity Futures Trading Commission explains that event contracts are frequently structured around a binary outcome with a fixed payout and an expiration. Participants can use them to speculate, to express a forecast, or in some cases to hedge exposure to a real-world event. The underlying is not a company share. It is the occurrence or measured result described in the contract rules.

That distinction separates prediction markets from surveys and traditional betting. A survey asks respondents what they believe. A market asks participants to trade at a price, creating a financial incentive to reveal information and challenge prices they think are wrong. A sportsbook typically sets odds and manages a book; an exchange-style prediction market matches participants through an order book or market mechanism.

How yes-and-no contracts work

Consider the market: “Will the central bank cut its policy rate at the September meeting?” The contract rules define what counts as a cut, the meeting, the official resolution source, the trading deadline, and the payout. Suppose a yes contract trades at $0.64 and pays $1 if the event occurs and $0 if it does not.

• A buyer paying $0.64 can receive $1 if the market resolves yes, for a gross gain of $0.36 before fees and taxes.
• If the market resolves no, the yes position expires at $0 and the buyer loses the amount paid.
• The holder may be able to sell before settlement at the current market price rather than waiting for the event.
• A no contract expresses the opposite side and pays according to the contract's defined structure.

Real markets add bid-ask spreads, order types, fees, position limits, partial fills, and liquidity constraints. The displayed probability may be a last trade, midpoint, or platform convention rather than the exact price available for your desired size. Always distinguish a headline number from the executable order book.

Why price is interpreted as probability

For a binary contract paying $1 on yes and $0 on no, a price of p dollars is commonly read as an implied probability of approximately p. Kalshi's current educational materials use this direct mapping: a yes price of 70 cents indicates a market-assigned probability around 70%. The CFTC likewise describes event-contract prices as reflecting traders' perceived likelihood.

“Approximately” does important work. Trading costs create a wedge. Thin order books can be moved by one participant. Limits may prevent informed traders from correcting a price. Participants may have hedging motives or different risk tolerances. Contract wording can make the event subtly different from the headline question. A market price is a forecast produced by a mechanism, not a direct measurement of objective truth.

Probabilities are also calibrated, not judged one event at a time. If events priced around 70% occur roughly seven times out of ten across a meaningful sample, that probability band is well calibrated. When a 70% event fails, the forecast was not necessarily “wrong”; the remaining 30% outcome had a real chance.

How markets aggregate information

Participants bring different data, models, domain knowledge, and interpretations. When they trade, their willingness to buy or sell moves the price. People with strong evidence can take larger or more aggressive positions; those without conviction can abstain. The resulting price compresses many views into one continuously updated number.

The classic review by economists Justin Wolfers and Eric Zitzewitz found that well-designed prediction markets can aggregate dispersed information and often produce accurate forecasts relative to moderately sophisticated benchmarks. Research using the University of Iowa's Electronic Markets has examined which market characteristics are associated with accuracy. These findings support prediction markets as a useful forecasting mechanism, not as an infallible one.

Information aggregation works best when the question is clear, the outcome is verifiable, informed participants can access the market, liquidity is sufficient, and traders have reasons to correct mispricing. It weakens when rules are ambiguous, the market is obscure, costs are high, or the participant pool shares the same blind spot.

Resolution and settlement: the part that matters most

Traders often focus on probability and overlook resolution. The contract rules—not a general sense of what happened—determine the payout. A strong rule names the measured variable, time window, official source, publication or determination time, edge cases, and what happens if data is delayed, revised, unavailable, or ambiguous.

Kalshi's help centre notes that markets may wait for an official source agency before settling, even when participants believe the event has effectively concluded. Its rules summaries specify the value being measured, timeline, and verification source. Other platforms use their own resolution systems, which may involve designated sources, an oracle, disputes, or platform review. Researchers should compare mechanisms rather than assuming every “yes” contract settles the same way.

Read the full rules before interpreting a price. “Will inflation be below 3%?” is incomplete until you know which index, release, month, seasonal adjustment, revision treatment, and rounding convention applies. A trader can forecast the broad economic story correctly and still lose because the contracted measurement differs.

Are prediction markets accurate?

They can be highly informative, but “accurate” needs a benchmark and a horizon. Compare a market with polls, expert forecasts, statistical models, or a simple base rate at the same timestamp. Measure calibration and scoring across many resolved questions. Do not cherry-pick a famous win or miss.

Markets may react faster than periodic surveys because prices update when participants trade. They also combine public and privately held interpretations. But speed can amplify rumours, and apparent precision can exceed the quality of the underlying information. A liquid political market and a thin technology milestone market should not receive the same confidence merely because both display two decimal places.

Watch for base-rate neglect, correlated participants, favourite-longshot effects, low limits, manipulation attempts, and markets whose participants are selected by geography or platform access. A well-functioning market can absorb attempts to move price if other traders have capital and information to take the opposite side; a small market may not.

How prediction markets relate to stock research

Prediction markets and stocks encode different objects. A stock price reflects discounted expectations about a company's future cash flows, risk, capital structure, and many possible outcomes. An event contract isolates one defined question. That isolation can make it a useful input when a company or sector is sensitive to a specific event.

A market on a rate decision can provide context for banks, growth stocks, and rate-sensitive sectors. A market on regulatory approval can help frame the event probability for an exposed company, although the stock's price reaction still depends on what was expected, the economic value of the outcome, and second-order effects. An election market can inform scenario analysis, but it does not estimate the magnitude of a policy's impact on each business.

Combine the tools carefully. Use the prediction market for the event probability, primary research for the payoff under each scenario, options or price data for market-implied risk, and stock sentiment for how the public conversation is reacting. If all signals move together, the event may be entering consensus. If they diverge, investigate the timing, contract definition, and source groups before assuming one is smarter.

Risks, market integrity, and regulation

Prediction-market contracts can lose their full purchase price. Other risks include thin liquidity, wide spreads, inability to exit at the displayed price, ambiguous rules, settlement disputes, operational risk, platform restrictions, fees, taxes, and misuse of non-public information. A probability that looks attractive can still offer poor expected value after costs.

In the United States, the regulatory position depends on the product and venue. The CFTC oversees federally regulated derivatives markets and publishes a current consumer guide to prediction markets and event contracts. In 2026 it issued staff and enforcement advisories and proposed a framework concerning certain enumerated event-contract activities. A proposal is not a final rule, and access or legality can differ by jurisdiction. Check the current regulator, platform, and contract terms rather than relying on an old article.

The CFTC advises customers to understand costs, rules, settlement sources, rights, and platform authenticity. That advice also improves non-trading research: a market is only as interpretable as the mechanism producing its price.

A practical prediction-market research checklist

1. Read the exact question and full rules. Record the resolution source and deadline.
2. Inspect the order book. Note spread, depth, volume, and whether the displayed probability is executable.
3. Translate price into payoff. Include fees, taxes, and the possibility of exiting early.
4. Choose a benchmark. Compare with a base rate, poll, expert forecast, or model at the same time.
5. Identify new information. Explain why the price moved rather than treating movement itself as evidence.
6. Check calibration over a sample. Do not judge probabilistic forecasts from one resolved market.
7. Separate event probability from asset impact. For stock research, model both the outcome and its economic payoff.

Primary sources and further reading

• CFTC: understanding prediction markets and event contracts
• CFTC: March 2026 prediction-markets advisory
• CFTC: June 2026 event-contract proposal
• Kalshi: what prediction markets are
• Kalshi: how contract prices are determined
• American Economic Association: Prediction Markets
• University of Iowa: what makes markets predict well