Diversity, Efficiency, and Incentives in Political Prediction Markets

Abstract

Critics of political prediction markets argue that prices are biased because traders are not representative of voters. We test this claim by analyzing data from political betting markets spanning more than 100 years. Our analysis yields two main findings: (1) Acros markets, traders were not representative of the voters. (2) Prices were not biased by political preferences of participants. For the 2024 election, we analyzed anonymized data from Kalshi, a CFTC-regulated exchange, the largest regulated political prediction market in the U.S. Kalshi's traders were 86% male and half were under 30. In its first month of listing election contracts, Kalshi's prices were more accurate than every poll-based forecast in the national race, the seven battleground states, and nineteen Senate races, and no poll-based forecast would have been profitable to trade against them in any group of presidential contests. Its volume came from areas that voted 28% Republican, while 60% of volume went to Republican positions. Traders with a single trade placed 73% of their volume on Harris. The association between a trader's location and the direction of positions declines with trading activity and is zero among active traders, who account for most of the volume. Prices can be accurate with unrepresentative traders when the forecasts of profit-motivated traders do not depend on their demographics. Two other markets corroborate our findings. In New York's 1900 market, bettors were older than the population and drawn from finance. On PredictIt in 2016, which limited each trader to $850 per contract, positions followed registered party more closely, as our theory predicts when market restrictions do not allow profit-motivated traders to set the price. Prediction market efficiency is driven by financial incentives over demographic representativeness.