How prediction markets can create harmful outcomes: a case study

post by B Jacobs (Bob Jacobs) · 2025-04-02T15:37:09.285Z · LW · GW · 2 comments

This is a link post for https://bobjacobs.substack.com/p/how-prediction-markets-can-create

Contents

  What are prediction markets?
  The Good
  The Bad
  Sports betting
  Decreasing trust
  Misinformation
  Conclusion
None
2 comments

What are prediction markets?

Prediction markets are platforms where participants place bets on the outcomes of future events. These events can include anything from election outcomes and economic trends to sports games and even natural disasters.

Here’s a simple example: imagine a prediction market for an upcoming presidential election. Bettors place money on the candidate they think will win. If many people believe Candidate A will win, more money flows towards Candidate A, increasing the price (or odds) of bets in their favor. Similar to horse racing, these odds shift in real-time, reflecting the collective beliefs of all the bettors. As the event approaches, these odds ideally provide accurate, aggregated predictions because participants have a financial incentive to be correct.[1]
This collective forecasting can help individuals, businesses, and even governments make better decisions.

At least, that's the idea.

The Good

I like the idea of prediction markets. For years I was a top 25 predictor on Metaculus, the world’s biggest “prediction platform”, a site akin to a prediction market but without the use of real money. Metaculus gives you a detailed track record of how you predicted in the past. Not only does this sharpen your forecasting abilities by promoting critical thinking and showing you in which areas you’re overconfident or underconfident, it also prevents hindsight bias (the “I knew it all along” effect).

For anyone who has ever watched the news, the appeal is obvious. Public figures regularly get away with making false or misleading claims. Prediction markets could help us here. If every time a bullshit artist starts spouting bullshit people would go: “Okay, bet on it then”, the bullshit artist would either be immediately revealed as a fraud, or go bankrupt (and then be revealed as a fraud).[2]
Conversely, people who may not look or sound like a traditional expert, but are good at making predictions, could make a lot of money calling out fraudsters and can subsequently become influential figures instead. That seems like it would create a much better media landscape.

The Bad

If prediction markets are so good, where are they? Well, governments don’t tend to like them. I’ve previously talked about how it may not always be ethical to require people to bet on their beliefs, and talked about how the interests of rich people could bias certain prediction markets,[3] but that’s not the reason governments don’t like them.

In 2003, the U.S. Defense Advanced Research Projects Agency (DARPA) funded a “Policy Analysis Market” designed to predict geopolitical events, including assassinations, terrorist attacks, and regime changes. When journalists and politicians discovered the market, they were outraged. Headlines labeled it a “terrorism futures market”, saying people could profit from successfully predicting—or even facilitating—harmful outcomes. The program was swiftly canceled within 24 hours of media coverage.

The objection here (beyond the usual moral panic) is that prediction markets can create perverse incentives. The classic scenario illustrating this concern is known as the “assassination market objection”. Suppose there’s a market predicting the assassination of a public figure. Bettors placing money on the assassination have a direct financial incentive to cause that assassination to happen (so they can profit from their prediction).

Proponents of prediction markets often dismiss this concern, arguing that incentives to cause harm exist in conventional financial markets as well, but rarely lead to manipulation or unethical actions because existing legal frameworks prevent it.
Governments aren’t convinced and have largely banned or heavily regulated prediction markets. This might have been a good idea, but it does limit our ability to empirically check whether they would’ve indeed created perverse outcomes.

With one notable exception.

Sports betting

Sports betting is a recently legalized form of prediction market that has grown enormously in a short amount of time. It can serve as a clear case study for the potential dangers of prediction markets. People have pointed out how it incentivizes and generates gambling addiction, which has also grown enormously since it was legalized. I won’t focus on this aspect here, but this is also a very real concern.

For all the ills sports betting has created, it does finally allow us to see whether perverse incentives are indeed created and acted upon in prediction markets. In short, the answer is yes:

Last year NBA player Jontay Porter was playing a game for the Toronto Raptors. Early in the game he grabbed several rebounds, seemingly headed for a solid performance.6However, Porter suddenly asked to leave the game, claiming an injury. Initially, this seemed routine and harmless, players often experience minor injuries and leave games prematurely. However, three months later, the NBA revealed that Porter had deliberately influenced the game’s outcome. Porter had participated in betting markets focused specifically on his own performance.

In Porter's case, here's exactly how the scheme worked:
The market knew Porter was a decent basketball player, so bettors generally expected Porter to perform well, placing bets saying he'd achieve more than 4.5 rebounds. Porter and his associates secretly placed large bets predicting he would have fewer than 4.5 rebounds. After quickly grabbing three rebounds in the game, Porter intentionally left, claiming an injury, thereby ensuring he stayed below the threshold. By leaving the game early, Porter ensured the underperformed bet would win — allowing him and his associates to rake in the profits.

Participants in prediction markets like Porter do not just bet on uncertain outcomes; they can effect/manipulate the outcomes. This is not a one off either, we can find similar scandals in tennis, football, rugby, baseball, cricket, and many more sports.

Sports leagues are now beginning to acknowledge these risks and are internally debating how much they should restrict betting markets, as are regulators.

Decreasing trust

So this is obviously bad if the behavior that’s incentivized by the market is unethical in itself (e.g. assassinations), but I would argue it has a negative effect even beyond that. I’ve previously written about how social trust is decreasing and what kinds of negative effects this has on society. Incidents like Porter’s significantly damaged public trust in the integrity of sporting events. People realize Porter was only caught because the site noticed an absurdly high volume of bets and warned the NBA. If Porter hadn’t coordinated with so many associates he would’ve gotten away with it, which means a lot of athletes are likely doing it and getting away with it.

If even an NBA player like Porter, who’s earning hundreds of thousands of dollars per year, can succumb to these incentives, imagine the risks for lower-paid or unpaid athletes, like college athletes.

People can now no longer fully trust the outcome of sporting events. Imagine what would happen if we widened prediction markets to allow betting on events in the public sphere, or worse, government decisions…

Misinformation

It gets worse. Prediction markets often incentivize the spreading of misinformation.

Here’s a simple example: let’s say there’s a prediction market for the outcome of a presidential election. The market is offering contracts that will pay $1 if Candidate A wins. Right now, the market price for a "Candidate A Wins" contract is $0.40, meaning the market believes Candidate A has a 40% chance of winning.

An individual or group buys a large number of contracts at $0.40. Then they deliberately spread false but convincing rumors online that Candidate A is surging in key states, or that there’s a leaked tape of Candidate B having sex with a minor.

This shifts public perception and causes other market participants to believe Candidate A’s odds are rising. The contract price jumps from $0.40 to $0.70. Those who bought in early (and spread the rumors) can now sell at a 75% profit.

Has something like this happened in real life? Almost certainly yes. It seems, for example, to have happened during the 2020 US presidential election. But of course, it’s nearly impossible to prove whether the people spreading rumors were deliberately lying. Still, even if the creation of falsehoods didn’t happen, the spreading of falsehoods can be just as dangerous.

I haven’t seen the forecasting community engage too much with this criticism. The standard retort seems to be that prediction markets also incentivize the detecting of falsehoods. That’s true enough in most cases, but I find this an insufficient defense for two reasons:

  1. In many cases, it’s more profitable to keep the discovery of a falsehood to yourself, so you can beat the market.
  2. It’s much faster and easier to spread falsehoods than it is to debunk them, so if both are incentivized we should expect the former to happen more often.

A world where prediction markets are common is a world where you will never be able to fully trust what people say, or do.

Conclusion

I appreciate the benefits of prediction markets: aggregating knowledge, holding public figures accountable, sharpening forecasting skills... That said, I'm extremely wary of expanding real-money prediction markets. What we witnessed with the legalization of sports betting is likely a sign of what’s to come if we fully legalized prediction markets: the creation of perverse incentives that people act on, an even steeper decline in public trust, and a sharp increase in gambling addiction.

I don’t think this is worth it. Luckily, platforms like Metaculus demonstrate that we can capture most of the advantages —like accurate forecasts, accountability, and improved reasoning— without financially incentivizing harmful behavior. Instead of pushing for broad legalization of prediction markets, I encourage the forecasting community to focus on promoting these safer, non-money-based alternatives.

 

  1. ^

    There are some further objections to the underlying assumptions here, but that’s beyond the scope of this post

  2. ^

    This doesn’t work for very longterm bets, and it also wouldn’t convince everyone, since conspiracy theorists still exist. Still, I expect it to be helpful on average.

  3. ^

    There are more objections to prediction markets, but those are beyond the scope of this post

2 comments

Comments sorted by top scores.

comment by Viliam · 2025-04-03T12:05:38.975Z · LW(p) · GW(p)

The part about misinformation is very important, and unlike assassination markets it seems to be in a blind spot of the rationalist community.

I suspect that most people here perceive the prediction markets through the mistake theory perspective: "people try to do their best at predicting outcomes, some of them make honest mistakes, prediction markets create a financial incentive to get it right (or to become less confident about your predictions)". A collective effort similar to scientific research, but distributed, open for everyone, and allocating money to the best researchers.

And that is true... kind of. In the same sense as the traditional market is a collective effort to produce goods for everyone at an affordable price, if possible.

That also is kinda true, but it also fundamentally misinterprets the gears of the situation by conflating "what typically happens at a free market" with "what motivates the participants in the market". Some of them might be motivated by the desire to provide cheap goods for everyone, but most of them are probably there in order to make money, preferably lots of money. The fact that free markets often make things cheap is a good news for the customers, but the producers are not necessarily happy about it, any many use various tricks to avoid this outcome.

Free market is great for the new entrepreneurs, because it allows them to join the game. Free market is bad for the established entrepreneurs, because it allows their competitors to join the game. This is why established entrepreneurs often looks for ways to keep their dominance on the market (other than merely providing better products at lower costs), such as regulatory capture. Instead of collective effort, free market would be better described as an impersonal force pushing the participants towards the outcomes they do not really prefer.

Now apply the same logic to the prediction markets.

The fact that prediction markets provide good predictions is nice, but the true motivation of many participants is simply to make lots of money. This will be even more true than today if prediction markets become more popular. If at some moment anyone will be able to spend their savings on prediction markets, we could expect that most people will be there for the money, and from their perspective "providing correct prediction" will be like "providing cheap goods" -- maybe something that can't be avoided, but not their ideal outcome.

On the prediction market, you make money when people vote against you. Therefore, even if you are a super-forecaster and regularly make correct bets, your profit is maximized when other people are wrong. If your budget is large enough so that you could make millions (or even billions) on the prediction market (in the hypothetical future with trillions of dollars in prediction markets), it would be rational for you to spend lots of money to make sure that as many people as possible are wrong.

The tobacco companies tried to make people wrong about the dangers of smoking. Now imagine companies just as rich, and just as evil, trying to make you wrong about... well, everything they can bet on. They don't even have to grow the tobacco and make the cigarettes in order to get rich; they can just as conveniently bet on other people's tobacco, cigarettes, vaccination, autism, whatever.

In theory, prediction markets can push people towards sanity, by rewarding correct predictions and punishing incorrect predictions. In practice, prediction markets would also create huge financial incentives for unethical companies to push people in the opposite direction (so that they can then take their money in the market). It is not obvious to me which force would turn out to be stronger in real life.

"But won't people stop betting on the prediction markets if they lose predictably?" No they won't. First, gamblers already keep losing and they don't stop. Second, if betting on a wrong but strongly held belief will feel emotional good at the moment of betting, even if it loses some money later, it may be worth it, emotionally. Not everything needs to be profitable: whenever you buy a cookie, you lose some money, but you still enjoy the cookie; perhaps the irrational people will enjoy placing their predictably wrong bets the same way.

The current prediction markets are systematically biased, because many forecasters there are motivated by the prestige, instead of / in addition to money. People willing to sacrifice prestige for money would bet slightly differently. They would actually probably often create new accounts, just to reduce the effect of "I know that this person is almost always right, so I hesitate to bet against them". They would want you to perceive them as idiots instead, because that will encourage you to bet against them. They would want you to believe that they are acting completely stupid.

As a hypothetical scenario, imagine a super-forecaster who predicts that something will happen with probability X, and then creates a new account to make that bet with $1000000, adding some insane comment such as: "I prayed to Jesus to make me rich, and Jesus told me to bet that the value is 12.34, because chapter 12 verse 34 in the Bible mentions money, and I trust Jesus enough to put my lifetime savings here". Then they would create another new account that bets $100 dollars against the position, saying "come on, this is so silly that this is practically free money!" And they they would share that screenshot on social networks. In my opinion, as a profit maximizing move, this would be more instrumentally rational than what participants in the prediction markets are doing these days (except for this [LW · GW] guy [LW · GW]).

comment by RobinHanson · 2025-04-02T18:12:20.770Z · LW(p) · GW(p)

Metaculus wouldn't work if it didn't offer incentives for participants. The fact that they aren't monetary doesn't mean they won't induce the same sort of problems you worry about above.

Surely we should compare, for particular topics, the magnitude of actual sabotage to the magnitude of the info value gained. And there are many ways to design markets to reduce the rate of sabotage.