Priced Belief

I have a bias I should state first: I think belief is the most under-instrumented quantity in the economy. Everything downstream of it, spending, voting, hiring, panic, is measured obsessively. The thing itself is measured with surveys, which is roughly like measuring temperature by asking people if they feel warm.

The structural shift

A price in a well-formed market is not an opinion. It is an opinion someone paid to hold, updated continuously, aggregated across everyone willing to pay, and punished immediately when wrong. That is a fundamentally better instrument than any survey ever built, and until very recently we only pointed it at a few dozen asset classes.

Three things changed at once. The regulatory position on event contracts in the US softened from prohibition to supervision. The cost of standing up a market on an arbitrary question fell to nearly nothing. And the liquidity problem, the reason niche markets were always useless, became tractable with automated market makers that will quote a price on a question three people care about.

The constraint on measuring belief was never conceptual. It was that thin markets do not produce trustworthy prices. That constraint is being engineered away.

The wedge

The consumer betting angle is the loud, obvious, and mostly uninteresting part. It is a real business and it will be won by whoever has distribution, which is not a technology question.

The interesting business is belief as an input to other systems. A continuously-updating probability is a data product, and there is no established market for it because it has never existed at this resolution. Who pays for it:

• Anyone whose planning cycle is slower than the world’s rate of change, which is now most large institutions.
• Insurers and reinsurers pricing events with no historical base rate.
• Media, which currently reports belief as anecdote.
• Any firm that would like to know what its own employees actually think, which is a genuinely different number from what they will say in a meeting.

The last one is the sleeper. Internal prediction markets have been tried for twenty years and failed on liquidity and on politics. The liquidity problem is now solvable. The politics problem is not, which is why this will be sold as forecasting infrastructure rather than as truth serum.

The framework

A belief market is useful exactly to the degree it satisfies four conditions. Most fail at the second.

1. Resolvable. The question has an unambiguous answer at a known time.
2. Costly. Being wrong takes something from you. Play money produces play beliefs, and every study of this that claims otherwise is measuring a population that is unusually interested in being right.
3. Liquid enough. Not deep, just deep enough that a single participant cannot set the price.
4. Unmanipulable relative to its stakes. The cost of moving the price must exceed what moving it is worth. Most markets on questions that matter fail here, which is the real ceiling on the category.

My position

I am long the infrastructure and the data layer, sceptical of the consumer applications on economics rather than on principle, and I think condition four is the thing that decides whether this becomes market structure or stays a novelty.

The uncomfortable part, which the enthusiasts skip: an instrument that measures belief precisely is also an instrument for manipulating it precisely. A visible price is itself an input to the belief it is measuring. Every serious version of this has to answer for the feedback loop, and “the market will correct it” is not an answer when the market is the mechanism.

The prediction

Within a few years, a prediction-market price gets cited in a mainstream institutional forecast the way a bond yield is cited today: not as a curiosity, as a rate. The signal is not volume on election night. It is the first central bank, insurer, or major index provider referencing an event-contract price in a published document.

← all theses