Looking at the data behind prediction markets
Dan Schwarz audits Polymarket and Kalshi against five promised benefits of prediction markets, finding volume concentrated in gambling rather than useful societal foresight.
What Matters
- 90% of Kalshi’s trading volume is sports betting; 80%+ of Polymarket volume is sports, crypto, or elections.
- Risk monitoring is the one category with supply-demand balance: 2,821 markets, $3.8B volume, 199 daily/weekly conflict trackers.
- Markets detect risks only after stories are already large—useful for monitoring, not discovery.
- Health markets failed: Kalshi’s weekly COVID vaccine markets averaged $8,000 each and missed Omicron at 3% odds one week before resolution.
- 85% of “interpreting news” volume ($1.25B total) is US federal interest rate markets—terrain CME futures and Bloomberg already cover.
- Accountability markets: two-thirds of $173M volume is Epstein file speculation, diagnosed by Rohanifar et al. (2026) as “prediction laundering.”
- AI markets ($25M volume) use Arena voting scores as judge—not credible by expert standards, serving bettors not decision-makers.
