A Primer on Prediction Markets


Last week we discussed how often elites when making predictions have neither skin-in-the-game nor accountability. This week we’ll talk about a technology aiming to introduce both.

Prediction Markets

In his seminal 1945 essay, The Use of Knowledge in Society, Economist Friedrich Hayek argued that market prices are the means by which disparate pieces of information are aggregated.

Hayek wrote that what’s in a single person’s head is only a small fraction of the sum of total knowledge held by people in society. Quoting Hayek:

“The economic problem of society is thus not merely a problem of how to allocate “given” resources — if “given” is taken to mean given to a single mind which deliberately solves the problem set by these “data.” It is rather a problem of how to secure the best use of resources known to any of the members of society, for ends whose relative importance only these individuals know. Or, to put it briefly, it is a problem of the utilization of knowledge which is not given to anyone in its totality.”

But what if we could unlock that information?

What if we could leverage the wisdom of crowds in a systematic way — if we could separate the experts from the charlatans — and then have the experts weigh in on crucial decisions?

What if we could aggregate people who are working on the front lines of national security, public health, drug development, movies, government funded projects, trade agreements, banks — and ask them whether their initiatives are on track?

For example, what if, instead of asking, say, Elon Musk when we’re getting self-driving cars, we could ask all Tesla employees working on them directly?

What if, more broadly, everyone had skin in the game for their opinions? What if people had a financial incentive to be diligent in their predictions? What if people put their money where their mouths are? Could this help us make better decisions?

That is the hope and promise of prediction markets.

Prediction markets are effectively betting markets for the purpose of predicting something we want to know — to discover what people truly believe.

For example:

  • Will X happen?
  • If X happens, will Y happen?
  • If X happens, what will the change in Y be?

The most popular prediction markets exist in sports and politics, but effectively, futures contracts, bounties, and insurance agreements are all prediction markets in that they use people’s desire to make money to predict the future.

Stocks, in contrast, are memes that have some cash flow (via dividends) but really are priced via speculation. Prediction markets always resolve back to some objective thing in the world.

Prediction markets purchase information from people who know the future, or at least are better at forecasting it. The market represents the public’s best guess as to what will happen. And people with information about the future are given money for divulging it.

Why are Prediction Markets important?

“The pace of scientific progress may be hindered by the tendency of our academic institutions to reward being popular, rather than being right.” Robin Hanson

At the highest level, prediction markets are important because they can lead to better decision making. If we believe that more accurate information is net positive an improvement in accurately valuing certain possibilities can lead to stronger governance and management.

As we discussed last week, there is currently little accountability for predictions. Politicians make baseless predictions with no accountability, while the media profits from sensationalist journalism. Pundits of all stripes have no skin in the game. Even when they get things wrong, they typically don’t go back to correct themselves. Experts don’t have incentives to speak up. Too much to lose.

Charlatans, however, make baseless predictions to build an audience. If they’re wrong, their tribe still supports them. Celebrities are winning The War of Ideas. Tribalism above truth. Entertainment over everything.

In an era of fake news, prediction markets can make a big impact.

Indeed, prediction markets are truth-seeking machines. By forcing people to put their money where their mouths are, people now focus on being correct, rather than being liked, popular, or diplomatic. If they’re unwilling to bet, they’ll be discredited. If they’re wrong, they’ll lose money and reputation. Vice versa, if they win, expertise will be elevated, humility will be appreciated, and charlatanism will be eliminated.

It’s not entirely wisdom of crowds — it’s wisdom of the “right” crowds — the experts. And, just as important, it silences (or cripples) the blowhards. If you don’t know what you’re talking about, you’ll abstain from voting, because, if you don’t, you’ll lose all your money.

Gambling incentives are not perfect, to be sure, but by helping aggregate more information, decision markets might allow us to more accurately estimate the consequences of important decisions, and in the process elevate experts and drown out post-truth charlatans.

How could this be used in the wild?

Consider a board determining whether it should hire or fire a CEO. It could decide based on its own instincts, or it could aggregate insights from the employees. A corporation could ask, “What will our Q1 revenue be if we fire our CEO?” and conversely, “What will our Q1 revenue be if we don’t fire our CEO?”

Another example: Movies. Movies have huge fixed costs and are difficult to MVP. Will X new movie do well? Employees can assess whether the movie will be ready in time, as good as initially envisioned, and whether people would like it.

More applications include the:

  • Likelihood of a homeland security threat
  • Global extent of a pandemic
  • Success of a drug treatment (and its massive fixed cost)
  • Sales revenue of an existing product
  • Cost of a government funded project
  • Effects of a trade agreement or government bailout

Prediction markets combine the Wisdom of Crowds with the Efficient Capital Markets hypothesis (stating asset prices reflect all available information providing the best estimate of intrinsic value).

Over the past 20 years, these concepts have been applied more broadly to such diverse topics as business project deadlines (Project Xanadu), intelligence analysis (DARPA’s experimentation), and box office success (Hollywood Stock Exchange).

Sports betting, democratic elections, and the stock market are three examples of prediction markets that exist today. Augur, Stox, Gnosis, Hivemind, et al are examples of more general purpose prediction market platforms.


One of the most interesting applications of prediction markets popularized by Robin Hanson is Futarchy: Government run by prediction markets.

Futarchy in action is the example of hiring/firing a CEO applied to the government: "If we fund this project, what will our GDP be in 5 years?" vs. “If we do not fund this project, what will our GDP be in 5 years?” Assuming GDP optimization is the goal, and the crowd believes GDP will be higher with funding, then fund the project.

In a Futarchy, governments can do the following: 1) allow citizens to democratically vote on which metrics to optimize for and 2) create markets to let the wisdom of the crowd inform how to reach those goals. Democracy tells us what we want while speculators bet on how to get it.

The Government could ask its citizens to rank order their priorities (GDP, education, and healthcare, for example), and then optimize against those priorities. Then, the Futarchy optimizes on that vector.

When democracy was invented, the world was very different. Interests were more regional than they are today, and the cost of communicating was much higher. These factors made sending regional representatives to a central legislature an obvious democratic strategy.

Nowadays, politicians don't win by telling the truth or making reasonable arguments. Under a Futarchy, there would at least be some market for truth-telling politicians.

Not only does a Futarchy help us make better decisions, it helps us make quicker, more scalable solutions by automating the process & reducing overhead. Right now, our government is bloated and its main business model is entrenching itself. Perhaps a Futarchy could fix that.

According to Robin Hanson, Futarchy seems promising if we accept the following 3 assumptions:

  1. Democracies fail largely by not aggregating available info.
  2. It is not hard to tell rich happy nations from poor miserable ones.
  3. Betting markets are our best known institution for aggregating info.

But of course, this concept doesn’t come without contention: Primary barriers to Futarchy adoption are lack of real world case studies or general purpose Futarchy solutions and entrenched institutions that are resistant to new models.

While Futarchy may provide a good theory to make decisions, at this time, it doesn't sufficiently address complex and important values like fairness and integrity and justice.

Common Criticisms

There are several criticisms and challenges against the wide use of prediction markets. We can divide these into three main buckets:

  • The powerful aren’t incentivized to use prediction markets
  • Implementation challenges
  • Causal relationships

The powerful aren’t incentivized to use prediction markets:

If we think about why things do or do not happen in our society, the typical drivers of change are the powerful (e.g., wealthy families, leading corporations, and political leaders).

The powerful typically act in their own self interest to hold on to their power in whatever form it is in. Robin Hanson’s Elephant in the Brain goes into depth on how people are adept at rationalizing why the interests of the powerful are also the collective’s best interest.

Prediction markets, however, threaten the hierarchical control of top managers, by demonstrating that most managers can’t predict the future. If prediction markets work as they should, they would diminish the power of those in control and give more of a voice to everyone else. While this could improve society as a whole, it could worsen the standing of the powerful.

There is also a view that without centralized moderation, communities inevitably collapse into mediocrity and chaos (essentially, Eternal September), so having a small number of users in charge could be a necessary evil.

Implementation challenges

In terms of implementing prediction markets, there are two major factors to consider: (a) the cost of running & maintaining the markets and (b) the clarity of the markets & associated outcomes.

Currently, most major prediction markets are run by centralized teams and platforms. These teams are important as they create new markets, maintain their integrity, handle disputes, etc.

As a consequence, there are fees associated with these types of markets, resulting in negative sum games. Prediction markets currently are more like gambling (either I win and you lose or vice versa) versus investing in most other markets (with a positive overall growth rate).

For prediction markets to truly take off, we need to establish more positive sum situations by either substantially lowering the costs associated with these platforms or by proving their positive sum nature.

Outside of cost, the clarity of the markets and outcomes can be a major challenge. Regarding the markets themselves, we need explicit language without any possibility of misinterpretation. If the market says that Steph Curry will hit a three in tonight’s game on June 20th, 2021, what happens if the game is delayed and he doesn’t hit a three until 1am EST (the next day). If the market is for a project to be completed by a certain date, how do you define “completed”? What if everyone thinks the project is done but then needs follow up work the next month?

And if prediction markets are run completely decentralized, how do you prove that an outcome truly occurred (i.e., the Oracle Problem)? If the market is a sports game, do you trust ESPN even though ESPN is centrally run?

Causal relationships

The last major bucket of challenges is when betting performance impacts the actual results and can become self fulfilling prophecies. The proliferation of markets might mean everyone is operating under monetary incentives and the future becomes effectively deterministic because the crowd can not just predict, but also affect the future.

This can lead to both severe negative consequences or positive ones. On the extreme negative side, this can lead to assassination markets or terrorist attacks. If there is a market for an individual to die on a certain day, one could be incentivized to make sure that occurs.

Yet on the positive side, if there is a market for a certain project to be finished by a certain time, those involved could be financially motivated to ensure that it does get completed.

In Conclusion

When it comes to implementing prediction markets, we’re still in the first inning. I’m excited to see prediction markets emerge in the years to come, and the effect they have on the quality of our discourse and our abilities to make decisions about the future. The Wisdom of Crowds in its truest form.

Until next week,