“prediction markets failed to beat basic statistical models for disease forecasting”
Tests whether Polymarket forecasts for influenza hospitalizations and measles cases can beat established benchmarks. The markets fail to outperform the CDC FluSight ensemble for flu or simple ARIMA baselines for measles. Two specific failure modes identified: traders placed probability mass on impossible cumulative outcomes, and low trading volume undermined price discovery.
Some technical background helpful
Platforms mentioned: Polymarket