Questions: Prediction Markets and Information Aggregation

Question 1

A prediction market shows a 30% probability that a new drug will receive FDA approval. You've done extensive research and believe the true probability is closer to 65%. What should you do, and what happens to the market price as a result?

• A. Nothing — the market price reflects all available information so your estimate is probably wrong
• B. Buy contracts; your purchases push the price toward 65%, correcting the mispricing
• C. Sell contracts; higher probability means the market will overcorrect upward
• D. Report your estimate to a prediction market administrator who will adjust the price

Answer: (B)

This is the core mechanism of prediction markets: anyone who believes the market is mispriced has an incentive to trade and profit from the discrepancy. Buying pushes the price up; your purchases move it toward the true probability as you profit from the mispricing. Option A confuses efficient markets with omniscient ones — the market is only as accurate as current participation allows, not infallible. Option C is backwards: if you think the probability is higher than the price, buying (not selling) is the profit-seeking move.

Question 2

A political pundit says a prediction market showing 60% for a candidate is 'just averaging what people think.' Why is this characterization wrong?

• A. The market aggregates only the opinions of credentialed experts, not average participants
• B. The price is a financial equilibrium where anyone who disagrees has an incentive to trade — it aggregates incentivized information, not stated opinions
• C. The market is more accurate simply because it includes more data points than any individual poll
• D. Market participants are vetted for epistemic calibration before being allowed to trade

Answer: (B)

Prediction markets don't average opinions — they create a financial equilibrium where disagreement is expressed through trades. If the price is 60% and you believe 80% is closer to the truth, you profit by buying, and your trade moves the price toward 80%. People with accurate information are rewarded; poorly calibrated participants lose money over time. This self-correcting mechanism is fundamentally different from a poll, where confident but wrong answers carry no cost. The price reflects revealed financial confidence at equilibrium, not a head count of beliefs.

Question 3

A prediction market price of 70% for an event means that 70% of participants believe the event will occur.

• T. True
• F. False

Answer: False

The price does not tell you what fraction of participants hold a given belief. It is the market-clearing equilibrium — the point at which the marginal buyer and seller agree to trade. A single well-funded, well-informed participant could move the price substantially. The 70% price is better interpreted as the collective probability estimate implied by all trading activity at equilibrium, where anyone who disagrees has a financial incentive to act. Counting heads has nothing to do with it.

Question 4

In domains with many active, financially-incentivized participants, prediction markets tend to be better calibrated than expert panels.

• T. True
• F. False

Answer: True

Research by Arrow, Hanson, and others has consistently found that well-populated prediction markets outperform expert surveys on calibration — events assigned 70% probability actually occur roughly 70% of the time. Financial incentives filter out overconfidence: there is a cost for being confidently wrong. Thin markets with few participants are a genuine concern, but when participation is sufficient, the incentive mechanism produces better-calibrated estimates than expert opinion, which carries no financial consequence for error.

Question 5

Why do financial incentives make prediction markets better information aggregators than surveys or expert panels?

Model answer: Financial stakes punish overconfidence and reward accuracy. Anyone who believes the market is mispriced can profit by trading against it — and if correct, they gain; if wrong, they lose. This means well-calibrated participants grow in influence while poorly calibrated ones are weeded out. Crucially, anyone with private information that differs from the market price is motivated to act on it, encoding that information into the price. Polls and expert panels carry no such consequence for error, so they fail to aggregate dispersed information as effectively.

The key insight is not that markets are smarter than any individual expert — it is that markets *incentivize accuracy* in a way polls cannot. A stated belief in a survey costs nothing. A financial bet on a wrong belief costs real money. This asymmetry drives traders to research carefully, update on evidence, and avoid confident errors. The result is a price that reflects dispersed private information from across all participants, including information no single expert or panel possesses.