Prediction Markets & Information

Prediction markets have always carried a strange allure. On the surface, they look like gambling, but underneath they are something much richer: a collective attempt to peer into the future by putting money where one’s mouth is. Their story stretches from Renaissance Europe, through Wall Street experiments, into today’s blockchain-fueled revival. Along the way, they have promised wisdom, provoked scandal, and forced regulators into uncomfortable debates.

The First Flickers of Prediction Trading

When we think of prediction markets today, we imagine sleek web platforms and crypto tokens. But the first true prediction markets emerged in far less digital settings. In Renaissance Rome, bets were placed on papal conclaves. Bankers, clerks, and ordinary townsfolk traded odds on which candidate might emerge as Pope. The practice became so widespread that Pope Gregory XIV issued an official decree in 1591, threatening excommunication for those who wagered on papal succession. These wagers were thought to carry genuine political weight, fueled by information leaking from insiders who attended the conclaves, and hence his decision.

Later, London coffee houses became informal trading floors for political outcomes. Pamphlets and gossip set the odds, and wagers reflected not just luck but also deep networks of information exchange. In both Rome and London, markets formed organically wherever information was scarce and uncertainty was high. These proto-markets carried two traits that remain central today: liquidity flowed through close social networks, and the sharpest edge belonged to those with inside knowledge.

By the late 19th and early 20th centuries, the United States saw its own flourishing of election betting markets. These “political stock markets” generated odds that were often as reliable, if not more so, than contemporary polls. A century later, academics revived the idea with the Iowa Electronic Markets, which allowed small-scale wagers on presidential outcomes. For several election cycles, these tiny markets routinely outperformed traditional polling averages, showing just how powerful the mechanism of incentive-driven aggregation could be.

Of course, not every attempt ended well. The U.S. Defense Department’s 2003 “Policy Analysis Market,” designed to forecast geopolitical risks, collapsed after a public backlash over what critics called “terrorism futures.” The lesson from this would be that even when markets could generate insight, public perception and ethics could shut them down overnight.

Prediction Markets Rediscovered on the Blockchain

For years prediction markets lingered in the academic and corporate fringes. Then blockchain technology arrived, promising a new way to decentralize, automate, and globalize their use. Ethereum, with its smart contracts, opened the door to protocols that could operate without a central bookmaker, using code and economic incentives to handle everything from order matching to payouts.

Augur was one of the pioneers. Launched in 2018, it allowed users to create markets on almost anything. Its rollout was rocky, but it proved a point: decentralized infrastructure made it possible to run a prediction market beyond the reach of any one regulator or company.

Soon after, Polymarket emerged as a more user-friendly alternative. While it quickly gained traction, it also drew the attention of U.S. regulators. In 2022, the Commodity Futures Trading Commission fined the platform and forced it to shut down certain markets. Yet Polymarket survived, regrouped, and expanded globally. By 2024 it was processing billions in volume, and by 2025 it had reportedly secured a path to operate in the United States more openly, showing that prediction markets could eventually coexist with regulatory frameworks. Most recently, the Intercontinental Exchange (ICE) agreed to invest up to $2 billion in Polymarket.

Blockchain not only provided new venues but also solved some old market design problems. Automated market makers kept even thinly traded questions liquid, and token incentives attracted a broader range of participants. The downside was just as clear: decentralization made it harder to prevent markets from unsavory or manipulable events, and pseudonymity raised the risk of coordinated manipulation. As with earlier eras, the technology could not resolve the core tension between harnessing information and avoiding moral hazard.

Insider information: the paradox at the core

This tension is most obvious when we talk about insider information. In traditional financial markets, insider trading is banned because it undermines fairness and investor confidence. Russian and Ukrainian hackers have been accused and prosecuted for breaking into the SEC’s EDGAR and other filing systems to steal non-public corporate filings and then trade on that market-moving information. The schemes prompted SEC enforcement actions and criminal prosecutions including indictments, settlements, and convictions  after authorities found traders had profited from the stolen filings. This just goes to show the value of information from the inside.

But in prediction markets, the very purpose is to encourage people with unique information to contribute. If a campaign staffer knows internal polling data or a hospital worker has early signals of an outbreak, their trades make the market price more accurate. Without them, the market might just be a dressed-up guessing game.

Economist Robin Hanson has even argued that manipulators and insiders can sometimes help markets by creating profit opportunities for informed counter-traders. Push a price too far from reality, and others will step in to correct it, leaving the final market price even sharper. Yet this is only true in well-designed systems with enough liquidity and diversity. In small or thin markets, one insider or manipulator can overwhelm everyone else.

The law sits awkwardly here. In the U.S., insider trading rules target securities, not event contracts. That leaves the CFTC as the main regulator for prediction platforms, particularly those tied to political or financial events. The line between legitimate informed trading and problematic outcome control is still fuzzy. What if a politician bets on their own reelection? Or a corporate executive bets on a regulatory decision that their lobbying could influence? Such scenarios show why regulators remain uneasy, even if academics see informational value.

Do Prediction Markets Really Work?

The evidence is mixed. When they are liquid, diverse, and carefully designed, prediction markets have a track record of beating polls and expert forecasts. The Iowa Electronic Markets often produced more accurate election probabilities than Gallup or other polling giants. Internal corporate markets at firms like Google have also shown that they can surface useful information from employees.

But when participation is thin or when markets are poorly framed, they can fail spectacularly. Markets can stampede into the wrong outcome if too many like-minded traders reinforce each other’s biases. They can also be vulnerable to “informational cascades,” where individuals ignore their own knowledge and simply follow the crowd.

The blockchain era has sharpened both sides of this reality. On one hand, open access and algorithmic market makers have drawn in new voices and kept markets liquid. On the other, pseudonymous trading makes it easier for coordinated groups to manipulate prices, and decentralized platforms cannot always stop markets that cross ethical lines.

Thermodynamics and Prediction Markets

To deepen the metaphor between physical entropy and informational entropy, we can turn to visualizations drawn from thermodynamics. The rotating surface plot P for a van der Waals gas provides more than a physical curiosity: it offers a geometric way to see how uncertainty responds to changes in underlying conditions.

In thermodynamics, entropy is not a static quantity, t shifts as temperature and pressure change, and the derivative surface captures the sensitivity of the system to those shifts. Regions of the surface where gradients steepen indicate zones of instability or rapid change, much like markets under stress. In calmer areas, the surface flattens, representing periods where uncertainty responds sluggishly to inputs.

Prediction markets can be seen through the same lens. The “temperature” of a market might correspond to volatility or the rapidity of information flow, while “pressure” could analogize the constraints of liquidity, regulation, or insider participation. Just as the entropy derivative plot shows that the same physical system behaves differently under varying pressures, prediction markets display different sensitivities depending on their structure. A thin, illiquid market may overreact to small inputs of information, while a deep, liquid market absorbs them smoothly.

The rotating view of the entropy surface underscores this dynamic nature. Information uncertainty is not a flat line or a simple curve; it is a landscape that shifts depending on the interaction of multiple parameters. For scholars and practitioners of prediction markets, this metaphor illustrates a critical point: understanding uncertainty requires not just measuring probabilities, but also studying how those probabilities bend, twist, and amplify under external conditions.

In this sense, Shannon’s information entropy and thermodynamic entropy are not only mathematically analogous, but visually resonant. Both invite us to map uncertainty onto surfaces and study its geometry. Just as physicists analyze these plots to anticipate phase changes or instabilities, market observers can use similar thinking to anticipate when collective wisdom might stabilize or collapse.

Social learning and the wisdom of crowds

Underlying all of this is the role of social learning. Prediction markets are not just mechanisms for individual bets but rather they are collective learning processes. Each trader observes the market, updates their beliefs, and acts accordingly. Sometimes this produces genuine wisdom, as diverse perspectives refine the price into an accurate forecast. But it can also produce herding behavior, where people ignore their own signals and simply mimic what the crowd is doing.

We see this clearly in other social trading environments, like eToro, where traders literally copy the strategies of popular investors. Research shows that this kind of imitation can amplify good information, but it can just as easily magnify noise. Leaderboards and reputation systems can steer flows toward a handful of prominent traders, making the market less diverse and more fragile.

For prediction markets to truly harness the wisdom of crowds, they need to design against these pitfalls. That means ensuring diversity of participants, rewarding contrarian information when the market becomes too one-sided, and carefully structuring questions so that outcomes are clear and resistant to manipulation.

Conclusion: wisdom and its limits

The history of prediction markets is a cycle of fascination and unease. From papal conclaves in Renaissance Rome to Ethereum-based platforms today, people have always wanted to turn uncertainty into tradable probabilities. Sometimes these markets have been astonishingly accurate; other times they have stumbled badly, either through poor design, manipulation, or ethical controversy.

The lesson is not that prediction markets are inherently wise or inherently foolish. They are tools, and like all tools, they work only when designed and used well. They require diversity, liquidity, clear rules, and trustworthy resolution mechanisms. They must also navigate the difficult paradox of insider information, welcoming the right kind of knowledge without creating perverse incentives to influence or corrupt outcomes.

When these conditions are met, prediction markets can provide remarkable insight into the future, often outperforming traditional experts and polls. When they are not, they reveal the darker side of collective behavior, where social learning becomes herding and wisdom devolves into noise.

In that sense, prediction markets mirror society itself: capable of flashes of brilliance, prone to episodes of folly, and always balanced on the knife edge between information and manipulation. Their story is not just about markets. It is about how we, as humans, learn together in the face of uncertainty. 

Sources and further reading

Academic & policy foundation

Wolfers & Zitzewitz (2004), “Prediction Markets,” JEP. (American Economic Association) Arrow et al. (2008),

“The Promise of Prediction Markets,” Science; AEI–Brookings Statement (2007). (Mason) Berg, Nelson & Rietz (2008),

“Prediction market accuracy in the long run” (IEM vs polls). (ScienceDirect) Cowgill et al. (2008), Google internal markets. (NBER Users)

Blockchain venues & regulation

Augur launch and evolution. (CoinDesk)

CFTC order against Polymarket (2022). (CFTC)

Reports on Polymarket’s 2025 U.S. return. (Reuters)

Election-season scale and manipulation concerns. (Vox)

Insiders, manipulation, and market design

Hanson & Oprea (2009), “A Manipulator Can Aid Prediction Market Accuracy.” (IDEAS/RePEc) Hanson,

“Insider Trading and Prediction Markets.” (Mason)

Outcome/manipulation models and robustness. (IJCAI)

SEC enforcement (shadow trading, high-profile cases). (Venable)

Social learning & herding

Banerjee (1992), Bikhchandani-Hirshleifer-Welch (1992); surveys and updates. (OUP Academic)

Empirical work on social/copy trading (eToro). (ScienceDirect)