Home / News

US Midterms Test Prediction Markets After Brazil Upset

• Polymarket gave Bolsonaro an 80% win chance in Brazil’s runoff, but Lula won 51.8% to 48.2%, exposing major election forecasting flaws. • Prediction market trading volume reached $21 billion in October, up from $90 million in 2024 and $4 million in 2022, intensifying scrutiny. • Studies show only 1% of Polymarket accounts captured three-quarters of profits, suggesting outcomes reflect skilled traders, not broad crowd wisdom.

Prediction markets enter the United States midterm elections as both a prominent forecasting tool and a contested experiment in financialized public opinion. Polymarket and Kalshi, once fringe betting platforms, now supply live odds to major newsrooms and have attracted billions of dollars in trading volume. Their growth rests on a claim: that money-backed markets aggregate dispersed information more efficiently than polls. Recent election misses, however, have sharpened doubts about whether that claim holds when political stakes, retail speculation and insider incentives collide. The most striking failure came in Brazil, where Polymarket priced Flávio Bolsonaro’s victory at about 80% for weeks and touched 84% before the runoff. Luiz Inácio Lula da Silva won 51.8% to 48.2%, despite polls showing a statistical tie. Similar errors followed in Serbia, where markets heavily favored Aleksandar Vučić’s United Serbia before the opposition Student List won 52.6% of the vote. In Wisconsin, Polymarket and Kalshi gave Francesca Hong roughly a 95% chance in a Democratic primary that David Crowley narrowly won. These cases suggest that market prices can become distorted by precedent, hype and uneven information. Scale has made the problem more consequential. Data from DefiLlama show combined trading volume on Polymarket and Kalshi reached about $21 billion in October, compared with more than $90 million around the 2024 presidential election and just over $4 million before the 2022 midterms. The CFTC, under a leadership structure shaped by Trump-era appointments, has helped normalize event contracts as forecasting tools. But researchers at Vanderbilt University examined 2,500 political markets during the final five weeks of the 2024 presidential race and found accuracy inconsistent with efficient information aggregation, with Polymarket at 67% and Kalshi at 78%. The deeper issue is who moves prices. A recent study found that just 1% of Polymarket accounts captured three-quarters of profits since 2022, while two-thirds of accounts lost money. Yale research estimates that only about 3% of accounts belong to skilled traders whose models, data and speed can correct mispricing. In quiet markets, those “sharps” can dominate; in noisy elections, recreational bettors flood in, often treating contracts as political expression rather than probability estimates. When sharp capital is already committed, crowd money can push prices toward a false consensus. Whether the US midterms vindicate prediction markets may depend less on aggregate wisdom than on the quality of marginal traders and the integrity of information. The platforms remain useful indicators of sentiment, liquidity and risk appetite, especially where polling is sparse or expensive. But their failures in Brazil, Serbia and Wisconsin show that prices can mistake momentum for certainty. If the midterms produce another confident miss, the industry’s central promise—that crowds with money at stake see the future more clearly than experts—will require a narrower, and less flattering, definition.