Mechanics
Why prediction markets work: the wisdom of crowds
The intuition behind prediction markets is older than financial markets themselves. Here's why aggregating many independent guesses outperforms any single expert.
In 1907 the statistician Francis Galton attended a livestock fair where 800 visitors guessed the weight of an ox. The average of all guesses was 1,197 pounds. The actual weight: 1,198 pounds. Individuals were wildly off. The crowd, somehow, was almost exactly right.
That's the wisdom of crowds in one anecdote, and it's the same mechanism that makes prediction markets work — with three important refinements.
The three conditions
Aggregating guesses doesn't always work. James Surowiecki, in his 2004 book on the topic, identified three conditions a crowd needs:
- Diversity. Participants should bring different information and perspectives. A crowd of identical experts is worse than a crowd of varied amateurs.
- Independence. Each person's guess should be made without seeing others' guesses first. Otherwise you get cascades — everyone copying the first confident voice.
- Aggregation mechanism. There needs to be a way to combine all the guesses into one number.
How markets achieve this
Prediction markets satisfy these conditions better than most polling or expert forecasting:
- Diversity is enforced by global, anonymous participation. A Polymarket position can be opened by a hedge fund analyst in New York and a college student in Manila. Their information isn't redundant.
- Independence is partial. Traders see the current price (which reflects others' beliefs), so guesses aren't fully independent. But the financial incentive to bet against the crowd when you have private information mostly recovers this.
- Aggregation is the price itself. Every trade adjusts the price by some amount; the price equilibrium is the weighted-average belief of all participants.
Why this beats single experts
A single expert, no matter how smart, has bounded knowledge. A market integrates the knowledge of every person willing to put money behind their view. Experts forecasting elections rely on polls and history. Markets integrate polls, history, on-the-ground intelligence from campaign workers, insider information, and the read of every attentive observer.
Studies consistently find that prediction markets beat individual experts on forecasting questions. They don't always beat the best expert — but they do consistently beat the average expert.
When the wisdom fails
Markets can be wrong. Conditions for failure:
- When the crowd is too homogeneous. A market dominated by US participants will misprice non-US events.
- When there's not enough liquidity. A market with $500 of total volume isn't a crowd, it's three people.
- When manipulation is rewarded. Tiny markets can be moved by someone with money and motive (e.g. influencing perception ahead of an event).
- When tail risks are involved. Markets are bad at pricing very-rare events because the rational arbitrageurs who would correct mispricings can't profit before resolution.