12 May 2026 Search

In On Prediction Market Regulation, Prof. Robin Hanson covers some arguments mostly used against public prediction markets on news events. Like news, gossip, and academia, prediction markets could:

  1. reveal info better kept secret

  2. reveal secrets people promised to keep

  3. waste time and money that could be used productively

  4. make misleading contributions to get favorable treatment

  5. change the world to get favorable treatment

  6. reward participants unequally.

Yes, public prediction markets are so hot right now—see Prediction markets are breaking the news and becoming their own beat by Neel Dhanesha—but the real interesting uses for them are still behind the corporate firewall. Remember Hail Jeffrey Wernick and Corporate Prediction Markets: Evidence from Google, Ford, and Firm X. The main problem with prediction markets is that they’re too often right, and accurately predicting the failure of the wrong executive’s pet project is an unforgivable offense.

A shareholder would prefer a prediction market to many other corporate information sharing and incentivization mechanisms, but the shareholders sit on top of a tall stack of principal-agent problem, a stack where managers with other priorities are in decision-making roles in the middle. But what if prediction markets were adopted in a more corporate-politics-resistant way? For example, a majority investor or a committee of the board of directors could sign the contracts with the prediction market administrator and oracle (referee), and protect the market from its natural predators. Do the criticisms of prediction markets that Prof. Hanson addresses still apply?

A and B: for most companies, there are going to be some issues that can’t be widely tradeable because too few people know about them. For example, customer confidentiality might require that a contract related to a a customer support issue must be limited to anyone actually cc-ed on the ticket. This is a good case for integrating corporate prediction markets with existing issue trackers.

C: This might be the best argument for internal prediction markets. Free-form Slack chats are usually something that would be better off in some other application: some threads would be better off as a bug report or support ticket, some in a shared calendar, some as edits to a shared document or Wiki page. A prediction market, for users who already know it, would be a time saver when compared to free-form discussion of the likelihood of a future event.

D: An internal market trader who had been successful for a while might end up building enough of a balance to shift the apparent probability of an event. Better-informed but less well bankrolled employees would be able to buy a position that would become profitable eventually, but in the time before maturity, the event would appear unlikely, making the market misleading as a source of management decision-making information. (For example, a group of new employees identify a flaw in a product and buy YES on “Will [product] be recalled for a safety issue before [date]?” The product manager, a successful trader in the past, buys a large NO position until just looking at the price, the product looks safe.) Two possible ways to address this: have more speculators/peer reviewers in the market, and schedule bonus evaluation dates after prediction market maturity dates.

E: There’s no real bright line between prediction markets and incentivization markets. (Some ways that bug futures markets differ from prediction markets). As long as no employee can take a position where they profit from their own failure (the best argument for integrating prediction markets with issue trackers) for internal corporate markets the incentivization is a feature, not a bug.

F: Companies generally want to do this, and prediction markets can be a fairer way to do it than other incentivization methods—as long as all employees get good training and practice in the market.

Consultants

How do you compensate a consultant, when a slick deliverable can be cranked out by an LLM, and the quality of the consultant’s advice can’t be realistically measured until long after their invoice is paid? Proving that a consultant’s deliverable is better than what could have been produced by an LLM is a sufficiently interesting problem that a prediction market could be useful.

One likely (and efficient) possibility is that “consulting” work becomes split into two kinds of contracts, with no consultant doing both for the same client:

  1. Independent referee (oracle) for the client company’s internal prediction market. Since the prediction market has to be administered by someone who does not report to someone who is allowed to trade, for a company-wide prediction market the oracle role would be filled by a consultant or panel of consultants.

  2. Informational/advisory deliverables could have a portion paid to the consultant as a base rate and the rest as a stake to be used in prediction market contracts based on the content of the deliverable. The consultant would not have to wait for maturity—gains could be withdrawn earlier.

Right now LLMs are weak at prediction market (and other market) trading (Money bots talk and bullshit bots walk?) which makes it a useful skill for signaling human competence.