Editor's Note: Prediction markets are moving from niche trading tools into the broader public information sphere.
Their logic is not complicated: Design a future event into a tradable contract, allow participants to express judgments with real capital, and then form an approximate probability through price. Compared to opinion polls, expert forecasts, or traditional asset prices, the advantage of prediction markets lies in their ability to aggregate decentralized information in real-time and, through a "lose money if wrong" mechanism, incentivize those who truly possess information to participate.
This is also the most noteworthy aspect of the article. The author does not deify prediction markets as "prophecy machines" but rather understands them within the context of market mechanisms themselves: markets not only allocate resources but also aggregate information; prediction markets directly apply this information aggregation capability to judge whether a specific event will occur. From geopolitics and election results to AI model performance and scientific experiment replication, many nuanced issues that were previously difficult to express through traditional financial assets can now be transformed into a set of tradable probabilities.
However, the effectiveness of prediction markets is not automatic. It depends on who is trading, how the contracts are designed, how outcomes are adjudicated, and whether the market is susceptible to manipulation by insiders or interested parties. If those with genuine information do not participate, prices may just be noise; if insiders place bets early, the market loses fairness; if political teams or project parties use funds to push up the probability of a certain outcome, prediction markets can also turn from "information aggregation tools" into "public opinion manipulation tools."
Therefore, the next step for prediction markets is not merely to increase trading volume, but to establish more trustworthy market infrastructure: transparent participation rules, clear contract design, auditable settlement mechanisms, and constraints on manipulative behavior. Their real value does not lie in letting people "bet on the future," but in providing a new type of public probability signal in highly uncertain environments.
The following is the original text:
Prediction markets allow people to trade around event outcomes. Last year, such markets in the United States began entering the public consciousness on a large scale and are now being used to track everything from geopolitics to entertainment award winners. But what exactly are prediction markets?
As an economist who has long studied markets and incentive mechanisms, my answer is simple: prediction markets are essentially markets. Markets are fundamental tools for allocating resources, ensuring that goods and services flow to those who value them most. In this process, markets also aggregate information: the market clearing mechanism absorbs the various information held by participants and condenses it into a signal, such as price.
Prediction market platforms and products directly leverage this information aggregation capability in an attempt to predict specific future events. They introduce an asset tied to a specific event: this asset pays out if a particular outcome occurs; subsequently, people trade this asset based on their judgment of whether the event will happen. Businesses have long used prediction markets, for example, to gather tacit information from employees to forecast whether a key product will be released on time. The scientific community has also used prediction markets to assess which experiments are more likely to be replicated. Today, we also see several media organizations partnering with prediction markets, using this "wisdom of the crowd" information to supplement reporting from sources and traditional journalists.
Prediction markets attempt to answer questions about the probability of different events occurring by directly collecting information from market participants—their personal judgments about the future—and aggregating this information into a market. People can "bet on" these events just as they might "bet on" a company's future value in the stock market or the future value of oil in the commodity market. What's different about prediction markets is that they are not tied to assets like oil, which are influenced by multiple factors, but instead introduce an asset that pays only if a specific event occurs.
If we see oil prices rise, we know that demand has increased relative to supply. But we don't necessarily know the exact reason behind it: is it because people expect a conflict in the Middle East to escalate, or because someone has invented a new use for oil? In contrast, prediction markets can separate each possibility. For example, a prediction market on "whether the Strait of Hormuz will remain open on a specific date and time" could revolve around a contract that pays $1 per unit if that event occurs. As people continuously buy and sell this asset, the market price can be interpreted as a kind of "probability sensor": it estimates the collective judgment of traders regarding the likelihood of the event happening.
Here's how it works specifically: Suppose the market price for a particular outcome is $0.50 per unit, representing a 50% probability. If you believe the probability of the Strait of Hormuz being open is higher than 50%, say 67%, you would buy. If your judgment is correct, you gain an expected value of $0.67 for a price of $0.50. Your buying action then pushes up the market price and its corresponding probability estimate, reflecting that someone believes the market previously underestimated the likelihood of the event. The reverse is also true: if someone believes the market price is too high, they will sell at a lower price or short the contract, thereby pulling down the overall market probability estimate.
When prediction markets function well, they have significant advantages over other forecasting methods. First, the mere ability to provide a probability estimate is itself a powerful capability. In contrast, polls and questionnaires usually only provide proportions of opinions; converting this proportion into a probability requires statistical inference about the relationship between the measured proportion and the overall population. Polls are also often just snapshots at a point in time, whereas prediction markets can update in real-time as new participants and new information enter.
More crucially, prediction markets have incentive mechanisms. Buyers and sellers have "skin in the game," suffering losses if their judgment is wrong. This incentivizes potential participants to seriously consider what information they possess and to put their capital into questions where they believe they have the most informational advantage. Conversely, prediction markets also give people an opportunity to utilize their information and expertise, incentivizing them to conduct research proactively and gain deeper understanding of relevant issues. A famous case involves a prediction market participant who, before the 2024 U.S. presidential election, even conducted their own poll using an atypical method in an attempt to uncover information not captured by traditional pollsters.
Finally, prediction markets have another important advantage: wide coverage. Someone aware of events that might affect oil demand can, in principle, express their view by shorting or going long on oil. But for many outcomes we wish to predict, there is no large commodity or stock market to effectively carry that information. Prediction markets could be ideal tools for such problems. For example, recent prediction markets have emerged to aggregate people's judgments on which AI models perform better on various tasks. Such questions are too niche to be reflected in traditional commodity markets. Anyone can create and fund a prediction market to answer such specialized questions.
These ideas are not new. They have existed in some form since at least 16th-century Europe, when similar mechanisms were used to predict the next Pope. Modern prediction markets are rooted in economics, statistics, market design, and computer science. In the 1980s, Charles Plott and Shyam Sunder proposed the first formal academic frameworks. Subsequently, the first modern prediction market—the Iowa Electronic Markets—was launched. Thanks to the internet, this model can absorb decentralized information from around the globe.
At the same time, many issues need to be resolved for prediction markets to truly realize their potential. These include infrastructure problems, such as how to verify and reach consensus on whether an event has occurred, how to ensure market operations are transparent and auditable; and also how to handle contract settlement at scale, as contract outcomes can be disputed or manipulated.
Beyond these, there are market design challenges. First, participants with relevant information must actually enter the market. If no one has an informational advantage, the price signals from prediction markets don't actually tell us much. Conversely, various people with relevant information must also be willing to participate; otherwise, the estimates from prediction markets will be biased. I noted in 2016 that prediction markets might have underestimated the likelihood of Brexit and Trump's first election victory because the people participating in prediction markets at that time lacked sufficient awareness of the reality of the populist rise.
Simultaneously, if someone with "perfect information" enters the market—for example, someone who knows in advance what the true outcome will be—this can also be problematic, especially if that person can influence the event's trajectory. Imagine if someone inside the papal conclave placed bets in a "next Pope" prediction market, trading before Pope Leo's public announcement, or even attempting to sway the conclave toward the candidate they bet on? If potential participants anticipate that insiders will trade in the market, the rational choice is to stay away, ultimately causing the market mechanism to break down.
Finally, there is the possibility that someone attempts to distort prediction market prices to influence public perception of the probability of a certain outcome. In this way, prediction markets could turn from tools for aggregating beliefs into tools for manipulating beliefs. If a candidate's communications team wants the outside world to believe they are winning, they could simply allocate part of their campaign funds to try to influence the relevant prediction market. However, prediction markets have some self-correcting ability because when a contract's probability estimate is pushed to an obviously unreasonable position, someone will always choose to take the other side of the trade.
All these issues point to the same need: prediction markets must establish greater transparency and clearer rules regarding participant management, contract design, and market operation. But if prediction market designers can successfully address these challenges, they may become a core tool for our understanding and navigation of the future.
