I understand nothing about futarchy

I. The word, and what it names

Futarchy was coined by Robin Hanson in 2000, in a text whose title already contains the whole idea: Shall we vote on values, but bet on beliefs?

Let us take it slowly. When we decide something in politics, we permanently mix two very different questions.

The first is a question of values. What do we want? Less unemployment or less inflation? More growth or fewer emissions? There is no correct scientific answer to that. It is a choice, it belongs to those who will live with it, and that is exactly what voting is for.

The second is a question of fact. Given what we want, will the proposed measure get us there? There, there is a right answer. We do not know it yet, but it exists, and we will know it afterwards.

Hanson's proposal fits in one sentence: we keep voting on the first, and we stop voting on the second.

In practice, the people choose at the ballot box a measure of national wellbeing, a sort of official thermometer. Then, for each decision, two bets are opened: what the thermometer will read if the measure is adopted, and what it will read if it is not. If the first ends up above the second, the measure passes.

Nobody bets on what is desirable. You only bet on what will work.

So the headline going round is wrong about half the mechanism. Values stay with the vote. The market never touches the ends, only the means.

II. Why it is resurfacing now

Because the infrastructure exists, and because American law moved. Here I have to be precise, because this is the point where I had it all wrong in my first version.

There was no grand decision by a regulator legalising these markets in early 2026. What exists is more modest and stranger: a letter from one division of the futures markets regulator, in March 2026, saying an event contract is « often » a financial product, and which disavows itself by specifying that it binds only its authors. Then a ruling by a federal appeals court, in April, handed down on an emergency basis and two votes to one, which produced the real effect.

And this detail nobody tells: the case was about football bets. The court found them barely distinguishable from those of an ordinary sports betting site. The dissenting judge spoke of « acts of alchemy ».

There is the founding event of the great democratic tipping point: two judges out of three, in a hurry, about football matches.

III. What these markets are really worth

Two researchers, Joshua Clinton and TzuFeng Huang, analysed around 2,200 political markets over the last nine weeks of the 2024 American presidential election. Version of June 10, 2026.

Result: 86 percent of actively traded markets did better than a tossed coin. By platform: Kalshi 90 percent, Polymarket 85 percent, PredictIt 82 percent.

That is good. It is very good, in fact, and it has to be said before criticising.

Except that you can slice it another way, and that is where everything is decided.

IV. The figure that changes everything

The authors sorted the contracts by price. A contract priced at 0.95 means the market is nearly certain. A contract priced at 0.50 means it has no idea. The first group is the questions whose answer we already know. The second is the real questions.

The interval contains 50.

Which means, in plain language: on questions whose outcome was genuinely uncertain, these markets cannot be statistically distinguished from a coin.

The prediction market is dazzling at predicting what everybody already knows.

And there is nothing scandalous about that: it is precisely what a market is supposed to do. It aggregates available information. When the information does not yet exist, it has nothing to aggregate.

But governing means deciding precisely in the zone around 50. A public policy whose result is known in advance does not call for deliberation, it calls for a rubber stamp.

V. The coup de grâce, and it is unintentional

Now look at what people actually bet on. Since July 2024, sport, politics and cryptocurrencies account for around 90 percent of the volume on the two big platforms. Sport alone makes up 80 percent at Kalshi.

But the most telling thing is the movement over time.

The infrastructure presented to us as a threat to democracy moved away from politics as it grew. It is not becoming a parliament. It is becoming a betting shop.

And futarchy, to work, would need markets on public policies: on vocational school reform, on the threshold of a benefit, on a map of low emission zones. Those markets do not exist. And if they did, they would be tiny and deserted, which is precisely the regime where we have just seen that prediction is worth nothing.

To this day, no government, anywhere, has ever tied a decision to a market's verdict. The only real applications are decentralised finance collectives, which vote their treasury that way, with real money and a computer program that executes the decision without any human being able to object.

VI. And what do we measure?

This is where I had planned my conclusion, and this is where I was most badly wrong.

I was going to write: refusing a mechanism that is right one time in two implies that we do better, and nobody measures a ministry's success rate.

That is false. We do measure it.

The French public action results barometer has been online since January 2021. It tracks around thirty priority policies, each with an indicator, a starting value, a current value and a dated target, nationally and department by department. It is all open data, updated weekly.

I downloaded the file and counted for myself.

Forty-seven percent. Against the market's 55 percent on genuinely contested questions, whose confidence interval contains 50.

The two mechanisms are roughly level, and that level sits roughly at chance.

There is the real story. Not « the gamblers are going to govern us ». Rather: when you finally take the trouble to count, neither the market nor the administration does much better than a coin toss on what is uncertain.

And if you find the barometer too generous, the harshest figure comes from research itself: out of some ninety educational programmes evaluated by random assignment, the most rigorous method that exists, eleven produced the expected effect. One in eight.

VII. What really should be kept from Hanson

It is neither the market, nor the gamblers, nor the thermometer.

It is this: his mechanism forces you to write down in advance what will count as success.

You cannot open a bet on « will this reform be a success ». You have to say which figure, measured how, by what date. Without that, there is nothing to bet on.

And that is exactly what ordinary public decision-making almost never does. A reform is announced with a verbal objective, and its success is then debated in whatever terms suit whoever is speaking.

The barometer, precisely, is the French version of that requirement, and with no gamblers at all. Numerical target, date, published gap, weekly update. You may find it imperfect. It has the merit of existing and of being something you can hold people to.

A decision announced with no success criterion is a decision that can never be evaluated, and therefore never corrected. It costs nothing to institute, it requires no market, and it is more serious than players betting on a thermometer.

What is established, and what I got wrong

Verified by me, in the documents: Robin Hanson's authorship and dates · the Clinton and Huang paper, whose June 10, 2026 PDF I downloaded and read, its 86 percent, its 90-85-82 ranking and its 0.40-0.60 band at 55.2 percent with the interval [49.0; 61.2] · the barometer, whose national summary file I downloaded and whose lines I counted.

Established by documentary research, not replayed by me: the breakdown of volume by theme and its movement, according to a May 2026 study by an American research institute · the figures on market governance in decentralised finance · the figure on educational programmes evaluated by random assignment · the detail of the American court decision.

Unverified, therefore absent: the press article that started the subject, still behind a paywall, which I have not read. So I am commenting on the way futarchy is presented in general, and on nobody's text in particular.