Written by: Scott Duke Kominers , Research Partner at a16z crypto
Compiled by: Chopper, Foresight News
Prediction markets allow users to trade on the outcomes of various events. These platforms began to expand rapidly in the United States last year, and now track events ranging from geopolitics to entertainment award winners. But what exactly is a prediction market?
As an economist who has long studied market mechanisms and incentive systems, my answer is simple: predictive markets are essentially ordinary markets. Markets are fundamental tools for allocating resources, allowing goods and services to flow to those who need them most. In this process, markets also possess the ability to aggregate information: the supply and demand clearing process integrates the information held by all participants and transforms it into signals such as prices.
Prediction market platforms and related products directly leverage this information aggregation capability to predict the trajectory of specific future events. The platform launches assets corresponding to specific events; if the predicted outcome comes true, holders receive returns, and users trade these assets based on their judgment of the event's probability. For a long time, many companies have used prediction markets to uncover tacit information held by employees to determine whether key products can be launched on schedule. Researchers also use this tool to assess the reproducibility of experimental conclusions. Today, many media organizations are also choosing to collaborate with prediction markets, using collective wisdom to supplement firsthand interviews and traditional reporting, enriching their content.
Prediction markets aggregate all participants' personal judgments about the future, then integrate these views into a trading market to calculate the probability of various events occurring. Users bet on event outcomes in these markets, logically no different from predicting stock prices in the stock market or trading oil prices in the commodity market. The difference is that the price of assets like oil is influenced by multiple complex factors, while the underlying asset in a prediction market only generates returns when the specified event occurs.
When oil prices rise, we can determine that demand exceeds supply, but we may not know the underlying reasons: is it market concerns about escalating tensions in the Middle East, or has oil found new applications? Predictive markets, however, can set up trading instruments for individual possibilities, allowing for precise breakdowns of predictions. For example, if a market is established for "whether the Strait of Hormuz will be open to navigation at a specified time," the corresponding contract rules could be set as follows: if the event occurs, each contract pays out $1. As users continuously buy and sell, the market price becomes a probability indicator, reflecting the overall assessment of the likelihood of the event by all traders.
The operational logic is as follows: Assuming the current price of each unit is $0.50, it means the market believes the probability of the event occurring is 50/50. If you judge the probability of the flight being completed to be higher than 50%, for example, reaching 67%, you can buy the unit. If your judgment is correct, the unit you bought at $0.50 will ultimately yield a profit of $0.67. This purchase will further push up the market price and the estimated probability, indicating that some traders believe the market has previously underestimated the likelihood of the event occurring. Conversely, if someone feels the current pricing is too high, they will sell at a lower price or short the unit, thereby lowering the market's probability valuation.
Compared to other forecasting methods, a well-functioning forecasting market has significant advantages. First, it can directly output quantifiable probability results, a core strength. Polls and surveys can only statistically determine the percentage of opinions; to extrapolate the probability of an event, statistical methods are needed to analyze the correlation between sample data and the overall population. Furthermore, poll results are mostly static data at a specific point in time, while forecasting markets update their judgments in real time as new participants enter and new information emerges.
More importantly, prediction markets inherently possess incentive and constraint mechanisms. Both buyers and sellers invest real money, and losses result from misjudgments. This forces participants to carefully analyze the information they possess, prioritizing transactions in areas where they are familiar with the information and have a greater informational advantage. Conversely, the desire to profit from information and expertise also motivates people to proactively conduct research and delve deeper into relevant clues. A well-known example is that, in the lead-up to the 2024 US presidential election, some prediction market participants employed unconventional methods to conduct polls, attempting to obtain information that traditional polling agencies could not access.
Finally, prediction markets have an extremely broad reach. Theoretically, traders with information about the oil industry can express their judgments by going long or short on crude oil contracts. However, in reality, many event outcomes cannot be predicted using mainstream commodity markets or the stock market. These scenarios are precisely where prediction markets come in. For example, many prediction markets have recently launched related products to comprehensively evaluate the performance of various artificial intelligence models on different tasks. Trends in these niche areas are difficult to reflect in traditional commodity markets. Anyone can build, fund, and operate a prediction market to answer these specialized questions.
Prediction markets are not a new concept; their earliest prototypes can be traced back to 16th-century Europe, where they were used to predict the next pope. Modern prediction markets integrate knowledge from multiple fields, including economics, statistics, market design, and computer science. In the 1980s, Charles Plott and Shyam Sand pioneered the formal academic framework for this mechanism. Shortly thereafter, the world's first modern prediction market—the Iowa Electronic Marketplace—was launched. Leveraging internet technology, this model has been able to integrate scattered information from around the world and continues to grow and expand.
However, to fully unleash the potential of prediction markets, several challenges remain to be addressed. First, there's the infrastructure aspect: how to determine the final outcome of events and reach a consensus; how to ensure market transparency and transaction traceability; and how to implement large-scale adjudication mechanisms when contract payment results are disputed or even manipulated.
Secondly, there are challenges at the market design level. First, those with core information must be involved. If all participants are ignorant, market price signals are worthless. Conversely, if informed parties are unwilling to participate, predictions will be biased. I proposed as early as 2016 that the market underestimated the probability of Brexit and Trump's first election as US president because participants at that time failed to foresee the rise of populism.
Furthermore, the entry of individuals with insider information into the market also poses risks, especially if they have the ability to manipulate the outcome. Imagine if insiders at the papal election meeting placed bets on the "new pope" in advance, using insider information to preemptively trade; or even secretly interfered with the election results to protect their own holdings—the consequences would be unimaginable. Once participants widely believe that insider trading exists, they will choose to leave the market, ultimately causing the entire market to collapse.
Another risk is that some might deliberately manipulate prediction market prices to guide public perception of event probabilities. In this case, the prediction market transforms from a tool for aggregating opinions into a means of manipulating public opinion. For example, campaign teams could use campaign funds to artificially inflate their market probability of winning, creating a false sense of lead. However, prediction markets possess a degree of self-correction: whenever prices significantly deviate from a reasonable range, traders will make contrarian bets to hedge against irrational pricing.
All these issues indicate that prediction markets need further rule improvements, clarifying standards for participant eligibility, contract design, and overall operation. However, if industry practitioners can overcome these challenges one by one, prediction markets will ultimately become an important tool for humanity to predict the future and cope with uncertainty.
