Prediction markets often outperform experts on bounded questions like elections — but the record on complex geopolitics, slow-moving structural events, and rare disasters is uneven.
The Iowa Electronic Markets beat 75% of pollsters on US presidential elections from 1988 to 2008. Polymarket called Trump 2024 weeks before mainstream pollsters. But prediction markets failed badly on COVID-19 timing, on the Russia-Ukraine war's duration, and on Brexit's actual economic effects. The Hayek-style information-aggregation argument works best when the question is well-defined, outcomes resolve in months, and trader sample is large.
