Prediction Markets

Any firm wishing to frame 2026 as part of a planning process may be tempted to glance at prediction (or information) markets such as Polymarket (or Kalshi). 

Built on a decentralized architecture, Polymarket hosts prediction markets where users trade contracts (using cryptocurrencies) linked to real-world outcomes. Each contract settles at $1 if the event occurs and $0 if it does not. Thus, a contract trading at 63 cents implies a 63% likelihood of the event occurring.

Collective beliefs are continuously translated into probabilities with brutal simplicity. According to Polymarket, for example, there is a 23% chance of a Fed rate cut by March; a 51% chance that Rick Rieder is selected as Fed Chairman; a 19% chance of an AI bubble bursting by the end of the year; and a 32% chance that the Supreme Court rules in favor of the administration’s tariffs. The platform also covers M&A and IPOs.

How reliable are these predictions? According to the ‘wisdom of the crowd’ idea, independent guesses made by a diverse population can outperform experts. The original story is that of Sir Francis Galton’s 1906 fair: hundreds of fair participants were invited to guess the weight of an ox. It turned out that the diverse crowd’s average estimate was astonishingly accurate, within a fraction of a percent of the true weight. Under the right conditions (diversity and independence), dispersed ignorance appears to aggregate into something resembling intelligence.

In ‘Prediction Markets: a Systematic Review and Meta-Analysis’ (2020), the authors show that prediction markets are, on average, about 80% more accurate than alternative forecasting methods such as polls or expert judgment. Specifically, according to ‘Exploring Decentralized Prediction Markets’ (2025), prices on Polymarket are, on average, directionally well-calibrated: higher-priced outcomes do occur more often.

Prediction markets can be so powerful that some large (tech) firms have deployed internal models to tap into their employees' wisdom.

Yet, these markets have important limitations: (i) participation is self-selecting, skewed toward a narrow crypto-adept demographic in the US and Europe; (ii) liquidity varies sharply by topic, inviting manipulation through fake trades; and (iii) unlike Galton’s fair, where everyone had one vote, predictions are weighted by capital.

Perhaps most critically, they face the reflexivity problem: participants react to price movements rather than forming independent assessments, an issue amplified by the echo chambers of social media.

To preempt a thought, equity markets are different. Stock market traders speculate on what they think everyone else thinks the ox weighs, rather than the ox's actual weight. In other words, it is a crowd betting on its own moves, hence the persistent risk of herd behavior.

In conclusion, prediction markets compress uncertainty into a single, seductive, and demonstrably informative number. Yet, they are best treated not as oracles but as a challenge to one’s worldviews.

Besides, only questions with a binary outcome can form the basis of prediction markets, while the world runs on nuances. Or does it still?