Futarchy

Most of this series has asked what a market can tell you. This one asks something stranger: what if a market could decide for you? Not advise, not inform — govern. In 2000, the economist Robin Hanson proposed exactly that, and gave it a name that still makes people flinch: futarchy. Government by prediction market.¹

It sounds like a libertarian fever dream, and the reaction it gets is usually a reflex — you can't put a price on the public good. But the actual proposal is more careful than the slogan, and the careful version is one of the sharpest ideas about collective decision-making I know. It's worth understanding even if — especially if — you'd never vote for it.

Vote on values, bet on beliefs

Hanson's design rests on a single distinction, and it's the whole thing. Every political choice braids together two very different kinds of question. There's what we want — should we weigh growth over equality, liberty over safety, this generation over the next? That's a question of values, and there's no expert who can answer it for you; it belongs to everyone. And there's what works — given what we want, which policy will actually get us there? That's an empirical question, a forecast about a tangled causal world, and being a forecast it has a true answer we just don't know yet.

Democracy, Hanson argues, is good at the first and terrible at the second. A vote is a wonderful instrument for surfacing what a society values; it's a dismal instrument for predicting whether a tariff will raise wages, because voters have neither the information nor any reason to acquire it. So his proposal — the title of his essay says it all — is to split them: vote on values, but bet on beliefs.¹

Concretely: elected representatives define a single welfare metric — a measurable proxy for national well-being the public agrees it wants to maximize. Call it GDP-plus: a basket that might fold in income, leisure, lifespan, environment, inequality, whatever the voters' representatives choose to weight. That's the values question, and it stays firmly in democratic hands. Everything downstream — which policy raises that number — gets routed to a prediction market.

The mechanism: conditional markets

Here's where it gets clever, because "let a market pick the policy" isn't obvious — a price tells you a probability, not a choice. The trick is the conditional prediction market: a contract that pays out based on one outcome given that some other thing happens, and is voided — every trade refunded — if it doesn't. You're not betting on what will happen; you're betting on what would happen under a specific scenario.

So for any proposed policy A, you run two markets side by side. One asks: if we adopt A, where does the welfare metric land a year out? The other: if we reject A, where does it land? Both are conditional — if the policy is adopted, the "reject" market voids and nobody who traded it is out a cent, and vice versa. Each market converges on the crowd's best forecast of the metric in its own branch of the future. Then the rule is mechanical: adopt A only if the "adopt" price is higher than the "reject" price. If the market expects the policy to raise welfare, it passes. If not, it doesn't.

That's the engine. No debate about A's merits, no floor vote on A itself — just two prices and a comparison. The market isn't forecasting the future passively; it's being asked to forecast two futures and the decision falls out of which one scores higher.

Why it's a serious idea

Strip away the strangeness and the pitch is genuinely elegant. It takes the two questions that politics constantly confuses — what we want and what works — and sends each to the tool that's actually good at it. Values go to a vote, where they belong, because no market can or should price what a society ought to care about. Empirics go to a market, because on empirical questions markets reliably beat pundits, pollsters, and politicians — which is not my opinion but the forecasting record this series has already walked through.¹

The deeper move is about incentives. A pundit who's wrong on television pays nothing; a politician who promises a policy will work faces voters who can't verify the counterfactual. Futarchy routes the empirical question to the one forecaster that goes broke when it's wrong — and forces the prediction to be explicit, numeric, and gradable before the decision is made. You don't get to argue the tariff will help in vague terms; you have to put a price on it, and someone richer than you gets to take the other side if you're bluffing.

What it actually is today

So has anyone tried it? Be clear: no government runs on futarchy. Not a city, not an agency, not a committee of any consequence. It remains a proposal — provocative, much-discussed, undeployed. Anyone who tells you otherwise is selling something.

What does exist are small, voluntary experiments, almost all of them in crypto. DAOs — on-chain organizations that govern a shared treasury — have piloted decision markets and conditional funding markets: before a proposal to spend the treasury passes, you run paired conditional markets on the token's price (or some metric) under "fund" versus "don't fund," and let the spread gate the decision.³ It's futarchy in miniature, with a token price standing in for the welfare metric. The results so far are interesting and inconclusive — thin participation, a handful of proposals, more proof-of-concept than proof. Useful as a lab. Not yet evidence that it scales to a polity.

What's genuinely hard

I find futarchy clarifying, and I would not want to be governed by it — at least not as written. Four problems, in rough order of how much they keep me up.

The metric is the whole ballgame

Everything rides on the welfare metric, and choosing it isn't a technicality — it's the entire act of government, smuggled into one definition. Worse, any metric you pick becomes a target, and Goodhart's law is merciless: when a measure becomes a target, it stops being a good measure.⁴ Optimize hard enough for measured GDP and the market will find the policies that pump the number while hollowing out everything the number was supposed to stand for. Voting on the metric instead of the policy doesn't escape this — it just moves the entire fight to one ferociously high-stakes definition.

The markets are thin and pushable

Conditional markets are structurally fragile. They fragment liquidity in half by construction — every question splits into an "adopt" book and a "reject" book — and one side often voids, which dampens the incentive to trade it carefully. Thin markets are exactly the ones a motivated actor can shove, and the prize for shoving here isn't a payout, it's control of the policy. That's a far larger bounty than a normal manipulation, which is precisely why someone would pay to move the price — a worry I took apart in what prediction markets can't do.²

The verdict changes the outcome

Then there's reflexivity, and in futarchy it bites harder than anywhere else in this series. A normal market forecasts a world it doesn't control. A futarchy market's verdict becomes the decision, which changes the world, which is what the market was trying to forecast. The thing being predicted is downstream of the prediction. That loop can be self-fulfilling or self-defeating, and it makes the conditional prices harder to trust the moment they actually bind — the same reflexive trap, sharpened.²

The legitimacy problem

The last one isn't technical at all. Even if every mechanical problem were solved — calibrated metric, deep markets, no manipulation — would anyone consent to it? A democracy's authority comes from the felt sense that we chose, together, and can throw the bastards out. "The price said so" carries none of that. People will accept a bad decision they voted for over a better one handed down by a market they don't understand and can't argue with. Legitimacy is not an implementation detail; for a system of government it may be the entire point, and a price doesn't supply it.

So I hold futarchy the way I think it's most useful: as a lens, not a blueprint. The lens is genuinely sharp. The next time a politician insists a policy will work, futarchy invites the right question back — would you bet on it, at the metric you claim to care about, against someone who's read the evidence? Most political claims dissolve under that question, and that's the contribution. You don't have to be willing to be governed by a price to want every empirical promise dragged toward one. Separating what we want from what works is the durable idea here. Letting a market cast the final vote is the part I'd leave on the whiteboard.