Prediction markets for policy

A prediction market lets participants trade contracts whose payoff depends on a future event: a contract might pay one unit if a policy target is met by a certain date and nothing otherwise. The market price can be read as a crowd estimate of the probability of the event. Prediction markets for policy use such prices to inform public decisions: for example, by forecasting the effects of a law, the risk of a crisis or the outcome of an election.

A more ambitious version, the decision market or conditional market, asks traders to predict an outcome under each of several possible decisions ("GDP growth if the policy is adopted" versus "if it is rejected"), so that the comparison of prices indicates which decision is expected to work better. Robin Hanson's futarchy proposal would make such markets binding: voters would choose goals, and markets would choose the policies expected to achieve them.

The case for prediction markets rests on the idea that prices aggregate dispersed information, following Friedrich Hayek's argument about the market as a discovery process. Traders with better information or analysis can profit by moving prices towards accurate values, so the market rewards accuracy and penalises error. Compared with expert panels, markets are argued to be:

• Incentive-compatible: participants have money at stake, which discourages cheap talk.
• Continuous: prices update in real time as information arrives.
• Open: anyone with relevant information can participate, including people outside official hierarchies.

Wolfers and Zitzewitz describe the conditions under which market prices approximate mean beliefs about probabilities, and they note pitfalls: thin trading, manipulation and the difference between correlation and causation in conditional markets. In 2008 a group of prominent economists, including Kenneth Arrow, argued in Science that regulators should create a safe harbour for small-stakes research and policy markets.

Forecasting mechanisms compared

• Iowa Electronic Markets (since 1988). The University of Iowa runs small-stakes markets on elections under a regulatory no-action letter. Reviews of its record report that it has outperformed large polling organisations: over four presidential elections its prices in the final week erred by about 1.5 percentage points on average, against 2.1 points for the final Gallup poll.
• The Policy Analysis Market (2003). The US Defense Advanced Research Projects Agency planned a market in contracts on geopolitical risk, such as economic growth, civil stability and military developments. After press reports in July 2003, critics attacked it as "terrorism futures" and the agency dropped the proposal before launch. Hanson, who guided the research, later wrote that the episode showed general acceptance of such markets was still a long way off.
• Forecasting tournaments. From 2011 to 2015 the US intelligence research agency IARPA ran a geopolitical forecasting tournament. The Good Judgment Project won by aggregating forecasts from trained volunteers; Atanasov and colleagues found that well-aggregated prediction polls were competitive with, and sometimes more accurate than, prediction markets.
• Corporate markets. Firms including Google, Hewlett-Packard and Ford have run internal markets to forecast sales and project deadlines.
• Regulated and crypto markets. Following a 2024 US federal court ruling against the Commodity Futures Trading Commission's attempt to block election contracts, regulated exchanges such as Kalshi expanded, and the blockchain-based Polymarket attracted large volumes during the 2024 US election. State regulators have since challenged sports-event contracts as unlicensed gambling.
• Futarchy experiments. Decentralised organisations, notably MetaDAO on Solana, use conditional markets to decide on proposals, the closest existing approximation of Hanson's design.

• Manipulation. Wealthy actors can move prices in thin markets, either to profit or to influence perceptions. Studies of historical and experimental markets suggest that manipulation is often short-lived, but high-stakes policy markets would attract stronger efforts.
• Thin markets and long horizons. Many policy questions resolve years later and attract few traders, so prices may reflect a handful of participants.
• Correlation versus causation. Conditional prices reflect what traders expect to observe if a decision is taken, not necessarily what the decision causes. If a policy is adopted only in bad times, its conditional market may look pessimistic for reasons unrelated to the policy.
• Moral objections. Betting on deaths, attacks or disasters can be seen as repugnant and could create incentives to cause the event. The Policy Analysis Market controversy shows how politically potent this objection is.
• Inequality of influence. Market weight follows money, not citizenship, which sits uneasily with democratic equality.
• Regulatory constraints. In many jurisdictions event contracts are regulated as derivatives or prohibited as gambling.

Prediction markets have a solid record as forecasting tools in elections, corporate planning and some economic indicators, though well-designed forecasting tournaments achieve similar accuracy without money at stake. Their use in public policy has been held back by moral, legal and political objections rather than by evidence of poor accuracy. Binding decision markets remain experimental and confined to small online organisations. The most realistic near-term role is advisory: aggregated probability estimates that sit alongside expert advice and public deliberation, rather than replacing democratic choice.

• Wolfers and Zitzewitz (2004), "Prediction Markets", Journal of Economic Perspectives
• Arrow et al. (2008), "The Promise of Prediction Markets", Science
• Atanasov et al. (2017), "Distilling the Wisdom of Crowds: Prediction Markets vs. Prediction Polls", Management Science
• Robin Hanson, "The Policy Analysis Market", Innovations 2(3) (2007)
• Robin Hanson, "Shall We Vote on Values, But Bet on Beliefs?", Journal of Political Philosophy 21(2) (2013)