Prediction Markets Are Not Democratic Forecasting Tools — A New Study Proves It

A study published in April 2026 documents something that regular participants on platforms like Polymarket have long suspected: prediction markets are not democratic forecasting instruments. They are arenas where a small cohort of professional traders — market makers and algorithmic takers — systematically extract value from the broader user base.

The numbers are stark. Roughly 3.5% of informed traders, including market makers and skilled takers, capture more than 30% of profits on prediction platforms. Meanwhile, approximately 67% of users absorb the entirety of losses. The study, drawing on transaction-level data from a major prediction market, found that the distribution of outcomes bore little resemblance to the egalitarian model that advocates frequently invoke when arguing for the epistemic superiority of prediction markets over polls or expert panels.

What the Data Actually Shows

The research traces profit flows across thousands of user accounts over a sustained period, distinguishing between market-making activity — where participants provide liquidity and capture the spread — and directional trading, where participants take positions on event outcomes. Market makers, by design, profit from volatility and the spread regardless of which direction an event resolves. Skilled directional traders profit from superior information processing or modelling. Together, these two cohorts account for a disproportionate share of positive returns.

The remaining users — the broad majority — are net losers. This is not a marginal finding. The study notes that the loss-absorption rate among retail participants is consistent across different event categories, including geopolitical outcomes, financial indices, and weather-related contracts. The结构性 inequality is not limited to a single market segment.

The finding complicates a long-standing claim in prediction-market advocacy: that aggregate prices embed genuine collective wisdom. If a small number of sophisticated participants are driving prices while the majority are losing money, the signal may reflect the views of a professional stratum rather than any meaningful consensus.

Who Are the Informed Minority

The study identifies three distinct groups driving the asymmetry. Market makers — often automated systems running continuous orders — profit from the bid-ask spread across large volumes of contracts. They do not require accurate predictions; they require active trading. Their presence stabilises liquidity but concentrates returns in ways that retail participants rarely achieve.

The second cohort comprises skilled directional traders, typically operating with analytical tools, real-time data feeds, and sometimes inside knowledge of the event domain. These participants are most likely to identify mispriced contracts before the broader market corrects them. The study notes that this group expands and contracts depending on the event type — political outcomes attract more skilled political traders; financial contracts attract more quants.

The third group, smaller and less consistent, comprises casual users who happen to align with outcomes by chance or through genuine insight. Their wins are real but not structurally replicable; the study finds that individual retail winners tend not to sustain returns across consecutive event cycles.

The Wisdom-of-Crowds Claim Collapses

The foundational argument for prediction markets rests on a specific version of the wisdom-of-crowds thesis: that aggregating many independent judgments produces a forecast more reliable than any single expert or poll. Francis Galton observed this effect in livestock guessing contests in 1906; Robin Hanson and others extended the logic to financial markets, arguing that real-money incentives produce sharper signal than voluntary surveys.

But the new data suggests a different model is operating. Rather than aggregating independent judgments, prediction markets appear to be pricing contracts through a contest between a professional class and a retail class, where the professional class has structural advantages — lower transaction costs, superior information, algorithmic speed — that the market mechanism does not equalise.

This matters for the political use case. Governments, media organisations, and research institutions have increasingly cited prediction market odds as evidence of public or expert sentiment on policy questions. If the prices reflect the views of a 3.5% stratum rather than a crowd, those citations may be unwarranted. The market is not aggregating opinions; it is revealing where the professionals place their money.

Implications for Platform Design and Governance

The findings raise questions about whether prediction markets can be reformed to deliver on their democratic promise. Some researchers have proposed royalty mechanisms or liquidity subsidies targeted at retail participants. Others have suggested tiered access — restricting certain contract types to verified sophisticated investors — to prevent retail losses from distorting prices.

Platform operators face a different calculus. Polymarket and comparable platforms depend on both retail volume — which generates controversy, attention, and viral spreads — and professional liquidity — which keeps markets viable. The tension between those two needs may be structural rather than resolvable through product design.

The study does not conclude that prediction markets are epistemically useless. Markets that attract skilled traders and active market makers may still produce prices that outperform naive baselines. But it challenges the stronger claim — that prediction markets are legitimate democratic tools for aggregating public judgment on factual questions. What they appear to be, more accurately, is a mechanism for professional traders to profit from retail enthusiasm for event resolution.

Whether that is a problem depends on what you want prediction markets to be. If the goal is accurate forecasting, the professionalisation of the space may be acceptable — better traders, better prices. If the goal is democratic participation in factual reasoning, the study suggests the market form is structurally ill-suited to the task. The crowd, it turns out, is not wise. A very small crowd is.

This article was published on 2026-04-27. Monexus covered this study primarily through its science and markets desks rather than the crypto desk, reflecting the study's focus on economic behaviour rather than platform-specific token dynamics.