Mental model
Prediction Markets
Markets where people trade on future events, turning scattered information into a shared forecast when participation and incentives are strong.
Discover
A tech company needs to forecast how many units their new product will sell in Q1. The CEO asks the sales VP for a projection, but you wonder if there's a smarter approach to tap into knowledge across the company.
Which method is most likely to gather useful knowledge from across the company and reward people for being right?
Let's explore how a company can combine many partial clues into one better forecast when people are rewarded for accuracy.
Understand
Understand
Prediction markets let people trade on whether future events will happen, like an election result or a product launch. As people buy and sell based on what they know, the price can become a rough summary of the group's forecast. The key idea is that people have a reason to act on information, not just share opinions. Check this: When new information appears, does the market price move?
Full explanation
Full explanation
Prediction markets turn a question about the future into something people can trade on. As people buy and sell based on their information, the price becomes a rough signal of the group's current forecast. For example, if a contract is trading around 70 cents, people often read that as the market estimating about a 70% chance of the event.
This works best when different people hold different useful clues and have a reason to act on them. One well-known example comes from Hewlett-Packard, where employees used an internal market to forecast sales, and that market beat a more traditional forecast in one setting.
Prediction markets are not magic: they work less well when few people participate, when nobody has better information, or when the outcome is hard to judge clearly. Their main advantage is that they reward people for acting on what they know rather than only sharing opinions.
Research
Research
Prediction markets use rewards to help combine scattered knowledge. When people can gain by spotting a price that seems off, trading can push the market toward a better shared forecast, especially when the market is well designed and enough informed people take part.
- Chen and Plott (2002): Corporate prediction markets at Hewlett-Packard improved sales forecast accuracy relative to traditional methods by aggregating dispersed information from participants inside the firm. [2]
- Hanson (2003): Market designs that let traders connect related outcomes can improve how information is represented across linked questions, though this source does not by itself establish a general claim that manipulation attempts often fail. [3]
- Wolfers and Zitzewitz (2004): Prediction markets are often treated as information-aggregation tools, but their accuracy depends on participation, incentives, market structure, and reliable contract resolution. [1]
Limitations
Limitations
Prediction markets face several practical and theoretical constraints. They can perform poorly when participation is thin, incentives are weak, market design is poor, or traders lack relevant information. Stronger claims about volatility, manipulability, calibration, or outcome verification need direct sourcing. Most fundamentally, markets can only aggregate the information that participants actually possess.
Try it
Synthesize
Choose a pattern from the guide, then pick an action to try with it.
Which pattern stands out?
What will you try?
Choose a pattern above to select an action.
Sources
Sources
- [1] Prediction MarketsJustin Wolfers and Eric Zitzewitz - 2004
- [2] Information Aggregation Mechanisms: Concept, Design and Implementation for a Sales Forecasting ProblemKay-Yut Chen and Charles R. Plott - 2002
- [3] Combinatorial Information Market DesignRobin Hanson - 2003
Try it
Check your understanding
Your company wants to predict whether a new software feature will launch on time. The engineering team is optimistic, while the QA team is skeptical. The product manager suggests averaging everyone's estimates. Based on what you know about prediction markets, which alternative approach best aligns incentives to reveal dispersed knowledge for forecasting?
Show the guide's explanation
Answer: Create an internal prediction market with rewards tied to accuracy
Prediction markets aggregate diverse information across departments and reward accuracy rather than confidence or status. Engineering knows technical realities, QA knows hidden risks, and other teams have complementary knowledge. When participants have something at stake, they're incentivized to share what they know through trading rather than echoing what sounds good. In one Hewlett-Packard sales-forecasting setting, Chen and Plott (2002) reported improved accuracy relative to a traditional approach.
A prediction market shows a 72-cent price for the contract 'Candidate X wins the election,' which traders often read as about a 72% chance. An experienced trader thinks the real chance is closer to 65%, so they sell because they think the price is too high. What does that action communicate to the market?
Show the guide's explanation
Answer: The trader believes the market is overestimating X's chances based on their information or analysis
The trader is putting money behind the view that the current price is too high. In prediction markets, this is how people push the forecast toward their best estimate: they trade when they think the market has not fully reflected the available information.
True or False: Prediction markets consistently outperform polls because polling only captures what people say they'll do, while markets capture what people actually believe will happen based on their actions.
Show the guide's explanation
Answer: False
Prediction markets are not guaranteed to outperform other forecasting methods. Their performance depends on participation, incentives, market design, and whether traders hold relevant information. Markets can fail when participants share similar blind spots or lack relevant information, causing the market price to reflect those omissions. The key advantage is incentive structure, not guaranteed superiority.
Keep exploring
Find another idea for the decision in front of you.
The complete Reframo library is free to read. Explore another guide whenever you are ready.