Election misses put prediction markets’ crowd forecasting claims under scrutiny

Polymarket assigned Aleksandar Vucic a 70 percent probability of winning Serbia’s prime ministerial race, but he lost with 39.2 percent of the vote, according to InfoBets. The outcome adds to the forecasting misses highlighted by the outlet.

Favored candidates fall short

InfoBets also cited a Wisconsin primary in which both Kalshi and Polymarket put Francesca Hong’s chances of winning at 95 percent. She lost, making that contest another instance in which the platforms’ heavily favored outcome did not materialize.

The figures measure different things: Vucic’s 70 percent market probability concerned his likelihood of winning, while 39.2 percent was his share of the vote. The gap between those numbers is not itself a measure of forecast error; the relevant outcome for the market forecast was his defeat.

A broader question about pricing

According to InfoBets, prediction platforms have an uneven forecasting record despite their argument that collective judgments can produce useful forecasts. The two election examples illustrate that high market probabilities do not guarantee results.

InfoBets also cited a study of 2,500 political markets that found prices inconsistent with rational information aggregation. That finding raises a broader question about how effectively market prices reflect available information, beyond the individual election misses.

What it means for prediction markets

• Probability is not certainty: Both the 70 percent assessment for Vucic and the 95 percent assessment for Hong left room for defeat.
• Individual misses offer limited evidence: These outcomes show favored candidates losing, but alone do not establish whether similarly priced contracts win at the expected frequency.
• The study widens the scrutiny: The finding across 2,500 political markets challenges the information-aggregation argument more broadly than either election example.