Prediction markets: information aggregation, contract architectures, and regulatory frameworks

Introduction to prediction markets and their historical development

Prediction markets, also referred to in academic and economic literature as information markets, idea futures, or event futures contracts, are marketplaces where participants trade financial contracts whose payoff depends on uncertain future events. Unlike traditional financial markets, whose primary purpose is capital allocation, risk transfer, or corporate financing, the scientific and economic value of prediction markets lies in their ability to aggregate dispersed information and convert that aggregation into highly efficient forecasts.

In the formulation associated with Justin Wolfers and Eric Zitzewitz, prediction-market prices react to what investors collectively believe about a future outcome, transforming subjective beliefs into a numerical market signal that is often read as a probability-like forecast. That signal is not identical to objective truth, but in many cases it becomes one of the cleanest available real-time summaries of collective expectations.

The historical lineage of prediction markets is long. The use of market prices, especially betting odds, to forecast events stretches back centuries. Political betting markets flourished in the United States in the late nineteenth and early twentieth centuries, and newspapers such as The New York Times treated betting prices as meaningful indicators of election expectations. Interest later faded as scientific polling gained institutional legitimacy, but returned in the late twentieth century alongside the rise of experimental economics and the efficient market hypothesis.

In the present era, blockchain infrastructure and decentralized finance have pushed prediction markets back into global attention. Today they are used, discussed, or contested in domains ranging from elections and macroeconomic indicators to business decision-making, weather, sports-adjacent events, and public-policy expectations. As a result, prediction markets now sit simultaneously inside economics, computer science, legal theory, gambling policy, and platform governance.

Theoretical framework and the dynamics of information aggregation

The strongest theoretical argument for prediction markets is rooted in the efficient market hypothesis. In its ideal form, this view suggests that market prices rapidly incorporate all available information and therefore represent the best available forecast of future events. In prediction markets, that means price becomes a collective estimate rather than merely a quote.

The academic literature has, however, treated this as a conditional claim rather than a naive certainty. The market can aggregate information only if incentives, contract structure, participation, and liquidity allow informed traders to act on their information. In other words, prediction markets can be highly effective without ever becoming perfectly efficient.

Mathematical and economic conditions of information aggregation

The literature has modeled these conditions precisely. Grossman argued that prices can aggregate dispersed private information under strong assumptions about trader preferences and information structure. Yet Grossman and Stiglitz later made the more famous counterpoint: if markets were perfectly informative, no one would have an incentive to acquire costly information in the first place. A functioning information market therefore needs some degree of pricing imperfection in order to reward informed participation.

This paradox matters because it explains why prediction markets are not magic forecasting machines. Their usefulness depends on the continued presence of informed traders who are motivated to push prices away from error and toward better estimates. Without those incentives, the market can become thin, noisy, or overly narrative-driven.

The literature has also debated what exactly a market price represents. Charles Manski showed that the price of an all-or-nothing event contract cannot always be interpreted straightforwardly as the mean probability belief of market participants. Under some assumptions, price may instead reflect a different feature of the belief distribution. Wolfers and Zitzewitz responded by showing that under a log-utility framework, and assuming wealth and beliefs are independent, price can be interpreted as the average belief about the probability of the event. If wealth and beliefs are correlated, however, price becomes closer to a wealth-weighted average belief.

A second classic challenge is the Milgrom-Stokey no-trade theorem. If market participants share common priors and are fully rational, then one participant's willingness to trade reveals information to the other, which in theory should suppress trade. In practice, trade persists because real markets include hedgers, entertainment-driven participants, noise traders, manipulators, and participants with heterogeneous motives. This mixed ecology is not a flaw at the margins; it is one of the reasons prediction markets can function at all.

This is also why attempted manipulation does not always damage price discovery in the way critics first expect. If a trader pushes price away from a defensible estimate, the distorted quote can attract better-informed counterparties who now have stronger incentives to trade against the error. In that sense, some manipulation attempts can end up strengthening the market's information signal rather than permanently corrupting it.

Wisdom of crowds as a forecasting mechanism

Prediction markets are often treated as one of the strongest institutional forms of the “wisdom of crowds” phenomenon. The basic idea is simple: when many individuals with partly independent information contribute to a shared aggregate, the resulting estimate can be more accurate than the opinion of any single expert.

This idea has deep roots in political theory, probability, and decision theory, and is often linked to Condorcet's jury theorem and Francis Galton's classic observations about aggregate estimates. In the context of prediction markets, the aggregation mechanism is not a poll average but a price formed through incentives. That distinction matters because market participants are rewarded or punished financially, which tends to discipline expression more strongly than ordinary opinion surveys.

Applied research has shown that internal corporate prediction markets can perform surprisingly well in forecasting demand, sales, deadlines, and project outcomes. They have been used in manufacturing, retail, and innovation settings as tools for turning dispersed tacit knowledge into operational forecasts. The more general lesson is that prediction markets are not only political or betting-adjacent curiosities; they can also function as organizational intelligence systems.

Contract architecture and price formation mechanisms

The ability of prediction markets to reveal meaningful expectations depends heavily on contract architecture. Even simple markets require well-designed payoff structures that eliminate trivial arbitrage and force prices to map coherently onto expectations. In this respect, prediction markets resemble derivatives markets more than they resemble casual betting slips.

Standard contract types and their information value

More advanced designs can reveal richer information. Families of winner-take-all contracts can be used to approximate an entire probability distribution over outcomes. Nonlinear index contracts can reveal higher moments such as variance. Conditional markets can be used to estimate beliefs about correlations, such as how one event might affect another. Even so, interpretation always requires care: a contract can reveal correlation without revealing causation.

Trading mechanisms and the management of liquidity

Early prediction markets often relied on continuous double-auction models similar to traditional exchanges. These mechanisms use an order book in which buyers and sellers post bids and offers. While familiar and powerful, they require organic liquidity to function well. If participation is too thin, spreads widen and prices stop behaving like reliable summaries of information.

To solve this problem, later platforms adopted algorithmic market makers. Robin Hanson's logarithmic market scoring rule became especially influential because it allows a market to provide continuous liquidity while bounding the market maker's loss. In decentralized systems, constant-product market makers and related automated market maker designs brought similar ideas into blockchain environments.

Another important design decision is whether the market uses real capital or play money. Studies have shown that play-money markets can sometimes produce surprisingly accurate forecasts, especially when reputation and status substitute for financial incentives. But if the purpose is not only forecasting but hedging or genuine risk transfer, real capital becomes far more important.

Technological evolution: decentralized prediction markets

Prediction markets have undergone a major technological shift from centralized web platforms to decentralized prediction markets built on public blockchain infrastructure. Earlier systems such as the Iowa Electronic Markets and later commercial platforms like Intrade depended on centralized operators, custody, and rule enforcement. They also suffered from obvious regulatory bottlenecks: if a jurisdiction objected, the platform could simply be forced to shut down or exit that market.

Decentralized prediction markets such as Augur, Polymarket, and Omen changed the architecture by moving market logic, custody, and settlement into smart contracts. In these systems, no single operator needs to hold user funds in the conventional sense, and no central administrator can unilaterally rewrite settlements after the fact. This makes the markets more transparent and more censorship-resistant, but it does not make them free of governance or integrity problems.

Augur was the best-known early pioneer of the decentralized model. It demonstrated that anyone could create binary, scalar, or multiple-choice markets in a permissionless environment, but also exposed practical weaknesses: high gas costs, slow user experience, and interfaces that were too complex for mainstream users.

Polymarket represents a later generation. By operating on lower-cost infrastructure and using stablecoins rather than volatile native crypto assets as the main settlement layer, it reduced transaction friction and made the user experience feel closer to a mainstream trading platform. Its rise during the 2024 U.S. election cycle showed how prediction markets can evolve into globally visible information products rather than remaining crypto-native experiments.

The oracle problem and dispute resolution mechanisms

The single biggest technical challenge in decentralized prediction markets is the oracle problem. Blockchains are closed systems: they can verify their own internal state, but they cannot directly observe who won an election, what the inflation number was, or whether a hurricane crossed a threshold. Every prediction market therefore needs a mechanism for importing real-world truth into on-chain settlement.

This creates a deep trust problem. If a market settles large sums of money, then the party that controls the oracle or dispute-resolution mechanism becomes the effective judge of reality. That is why oracle design is not a small engineering detail but the central governance problem of decentralized prediction markets.

UMA's Optimistic Oracle

One of the most influential solutions is UMA's Optimistic Oracle. The model is called “optimistic” because it assumes proposed answers are correct unless they are actively disputed. In practice, this means a participant proposes an outcome and posts economic collateral. If nobody challenges the proposal during the challenge window, the result is accepted and settlement proceeds quickly.

If the result is disputed, the system escalates into a heavier verification process in which economically motivated participants vote on the truthful answer. This design depends on a Schelling-point logic: because participants do not know exactly how everyone else will vote, the rational strategy is to coordinate around the most publicly defensible truth rather than around an obviously false answer. The stronger the slashing and staking incentives, the more expensive it becomes to corrupt the result.

Tokenomics, attack vectors, and alternative designs

Token-based oracle systems are never risk-free. If the value secured by a market becomes large relative to the security value of the oracle token itself, the incentive to manipulate the system can rise sharply. This is one reason why researchers and protocol designers have explored alternatives such as broader collateral bases, ETH-backed security models, longer lock-up periods, randomized juror selection, and nonlinear vote-weighting models.

Quadratic-weighting proposals are especially notable because they are designed to make raw capital concentration less dominant in dispute resolution. Under such proposals, voting power grows more slowly than stake size, making it more expensive for a single wealthy actor to dominate a vote purely through capital scale. Whether such systems outperform simpler stake-weighted voting remains an open design question rather than a universally solved problem.

More broadly, the oracle problem shows why decentralized prediction markets are never “fully solved” by smart contracts alone. A blockchain can secure balances, transfers, and settlement logic, but it still needs a credible bridge to contested reality. That bridge is where economics, governance, and legal risk all re-enter the system.

Moral hazard and ethical controversy

The greatest social objection to prediction markets is not technical but ethical. If a market allows participants to profit from war, assassination, terrorism, or human tragedy, then the market may create perverse incentives. Even if most users are passive speculators, the existence of a direct payout tied to a harmful event raises the question of whether someone could be incentivized to help bring that event about.

This concern is often illustrated through the thought experiment of assassination markets. In practice, early decentralized systems faced versions of exactly this problem when users created markets on individual deaths or violent public outcomes. Because decentralized protocols are hard to censor at the contract layer, the burden often shifted to front-end moderation and interface-level restrictions.

The controversy is not only hypothetical. The U.S. government's Policy Analysis Market project in the early 2000s collapsed after public outrage that it might allow speculation on geopolitical violence. Supporters argued that such markets could reveal useful distributed information. Critics argued that the moral cost and incentive structure were unacceptable.

Defenders of prediction markets often respond that ordinary financial markets also move in response to tragedy. Airline stocks can be shorted ahead of bad news; defense stocks may rise in war. In that sense, profit from catastrophe is not unique to prediction markets. The counterargument is that prediction markets make the link between outcome and payout unusually direct, transparent, and psychologically salient.

Regulatory classification and legal frameworks

Regulation remains the single biggest structural constraint on prediction markets. The central legal question is not merely whether the product involves risk, but how it should be classified: as a derivative, as a gambling product, as a hybrid information market, or as some new category that existing law handles poorly.

United States: CFTC event contracts and gambling boundaries

In the United States, the main legal struggle has played out between derivatives regulation and gambling-style objections. The CFTC currently treats event contracts as a derivatives category, and its official contracts-and-products material explicitly defines an event contract as a derivative whose payoff is based on a specified event, occurrence, or value. The same material also notes that CFTC Regulation 40.11 prohibits certain categories of event contracts, including contracts related to terrorism, assassination, war, gaming, or unlawful activity.

The current framework remains highly contested. In May 2024, the CFTC issued a proposal on event contracts that would have treated several categories, including certain political and sports-related contracts, as contrary to the public interest. In February 2026, however, the Commission withdrew that proposal and later in March 2026 sought public comment on a new prediction-markets rulemaking approach. The category is therefore live, politically sensitive, and still under active legal construction rather than fully stabilized.

European Union: MiFID II, binary options, and gambling law

In the European Union, prediction markets sit uneasily between financial regulation and national gambling law. If a contract is classified as a financial instrument or binary-option-like derivative, it may fall into the MiFID II and investor-protection framework. This matters because ESMA's 2018 intervention measures prohibited the marketing, distribution, and sale of binary options to retail investors in the EU.

If the contract instead looks more like event betting, then the platform faces a different obstacle: gambling law in Europe remains largely national rather than harmonized at EU level. This means a cross-border prediction-market operator cannot simply assume one universal legal route across all EU countries. Licensing, advertising, market access, and consumer-protection expectations vary sharply across jurisdictions.

Malta remains especially important in this landscape because the Malta Gaming Authority has long experience regulating remote gaming and commission-based market structures. The MGA's own materials distinguish licensable game types, including a Type 3 category for certain commission-based, non-house-risk models such as peer-to-peer poker, bingo, betting exchange, and similar structures. That makes Malta one of the most relevant European reference points whenever prediction markets start to resemble exchange-style or peer-to-peer products rather than conventional house-banked casino models.

Finland: Lotteries Act, licensing transition, and financial-law boundary

In Finland, prediction markets sit at the intersection of the Lotteries Act and financial-market regulation. If the practical effect of a product is that a user pays to participate in an uncertain event for monetary gain, the gambling-law frame becomes immediately relevant. The Finnish Lotteries Act is therefore a natural starting point for the analysis of any platform actively targeting Finnish consumers.

At the same time, if a prediction-style contract is structured around economic indicators or otherwise looks sufficiently derivative-like, the financial-law frame and supervisory expectations of the Financial Supervisory Authority become relevant. The result is a legal grey zone rather than a neat classification rule.

The Finnish context is further complicated by the country's transition away from the old monopoly model toward a more open licensing structure. That reform process may eventually create clearer pathways for some betting-adjacent products, but it does not automatically resolve how event contracts that resemble both derivatives and gambling should be classified.

Empirical evidence, market efficiency, and behavioral bias

Empirical work on prediction markets is generally supportive, though not uncritical. Large, liquid markets often outperform expert panels and conventional polling at least in timeliness and calibration. They react quickly to new information and can track changing expectations in near real time. That speed is part of what makes them useful to economists, policymakers, and journalists.

But market prices are still produced by human participants, and human participants bring cognitive bias with them. This is especially clear in political prediction markets. Partisan traders often display position persistence, hold on to favored narratives too long, or react asymmetrically to information that harms their preferred outcome. Even so, financial incentives frequently discipline this behavior over time. Participants may be emotionally biased, but markets still pressure them to adapt when the cost of stubbornness becomes large.

As institutional participation and algorithmic trading grow, these markets may become more efficient in some respects and more fragile in others. Better arbitrage can remove naive pricing errors, but it can also transform event markets into an increasingly professionalized asset class that is difficult for casual participants to understand.

Conclusion

Prediction markets have evolved from betting-like electoral indicators into sophisticated information systems and programmable financial instruments. Their ability to aggregate dispersed information is one of the strongest empirical arguments in their favor, and their technological evolution has made them more scalable, transparent, and globally visible than earlier generations of event markets.

At the same time, the category remains structurally unstable. Oracle design, manipulation resistance, moral hazard, national security concerns, and legal classification all remain open problems. This is why prediction markets should be treated neither as a simple extension of sports betting nor as a neutral machine for discovering truth. They are a hybrid market form whose usefulness, risks, and legitimacy all depend on architecture, incentives, and regulation.

Their long-term future is likely to remain strong as a forecasting and market-design tool, but stable mainstream adoption will require more than technological innovation. It will also require legal clarity, credible dispute-resolution systems, and a better answer to the ethical problem of markets built around harmful outcomes.

Selected sources and further reading

This article is a translated and adapted research-style overview. For citation work, the most important practice is to cite the underlying sources directly rather than only this summary page.

1. Justin Wolfers and Eric Zitzewitz, “Prediction Markets,” NBER Working Paper 10504 (2004).
2. Justin Wolfers and Eric Zitzewitz, “Prediction Markets in Theory and Practice,” NBER Working Paper 12083 (2006).
3. Charles F. Manski, “Interpreting the Predictions of Prediction Markets,” NBER Working Paper 10359 (2004).
4. Erik Snowberg, Justin Wolfers, and Eric Zitzewitz, “Prediction Markets for Economic Forecasting,” NBER Working Paper 18222 (2012).
5. Anthony M. Diercks, Jared Dean Katz, and Jonathan H. Wright, “Kalshi and the Rise of Macro Markets,” NBER Working Paper 34702 (2026).
6. Commodity Futures Trading Commission, “Contracts & Products: Event Contracts”.
7. Commodity Futures Trading Commission, “CFTC Issues Proposal on Event Contracts” (May 10, 2024).
8. Commodity Futures Trading Commission, “CFTC Withdraws Event Contracts Rule Proposal and Staff Sports Event Contracts Advisory” (February 4, 2026).
9. Commodity Futures Trading Commission, “CFTC Seeks Public Comment on Advanced Notice of Proposed Rulemaking Relating to Prediction Markets” (March 12, 2026).
10. Commodity Futures Trading Commission, “CFTC Reaffirms Exclusive Jurisdiction over Prediction Markets in U.S. Circuit Court Filing” (February 17, 2026).
11. Commodity Futures Trading Commission, “CFTC Enforcement Division Issues Prediction Markets Advisory” (February 25, 2026).
12. UMA Protocol, “Contracts Documentation”.
13. European Securities and Markets Authority, “ESMA agrees to prohibit binary options and restrict CFDs to protect retail investors” (March 27, 2018).
14. Malta Gaming Authority, “What are the different types of games that are licensable by the Authority?”.
15. Finlex, “Lotteries Act / 1047/2001”.
16. Iowa Electronic Markets, official market archive.