←BACK TO BLOG
>

The Fight Futarchy Doesn't Solve

Prediction MarketsMarket DesignGovernance
JUL 2026·5 MIN READ

#THE PITCH

Robin Hanson circulated a paper in 2000 (formally published in the Journal of Political Philosophy in 2013) arguing that government could be split into two clean layers. Vote on values, bet on beliefs. Democracy decides on a single measure of national welfare, and prediction markets then bet on which policy maximizes it. Whichever policy the market prices higher gets implemented, no legislature required to adjudicate the beliefs part ever again.

This stopped being a thought experiment a while ago. MetaDAO, built on Solana, runs proposals through conditional pass-versus-fail token markets instead of one-person-one-vote, and in February 2025 it used its own mechanism to hire Hanson himself as a paid adviser, the system voting in its own inventor. GnosisDAO passed GIP-145, a roughly nine-month advisory futarchy pilot built on the Conditional Token Framework, run alongside its ordinary Snapshot voting. People are running this with real money now, not just citing it in position papers.

#THE PART THAT ACTUALLY WORKS

The market half deserves the credit Hanson gives it. Condition a token on 'policy A passes' and another on 'policy A fails,' let traders price both against a welfare metric, and you get a real-time estimate of which choice people with money on the line expect to do better, stripped of the incentive to grandstand that legislative debate rewards. That's the actual contribution here, and it's not nothing. The problem was never the betting.

#THE METRIC IS THE ELECTION

The problem is the number everyone's betting against. Hanson's own candidate is 'GDP+', augmented GDP meant to fold in health, leisure, happiness, and environmental quality, with the exact formula left to be settled politically. He's floated real specifications: a discounted average over time of the square root of GDP, or a weighted average of augmented consumption sliced by geography, ethnicity, age, gender, and income. Read that list again. Every term in it is a fight, not a footnote.

  • ▸How much weight leisure gets against raw output: a judgment about whether a poorer, less-worked society counts as worse off
  • ▸Whether the metric is discounted, and at what rate: a judgment about how much the future counts against the present
  • ▸How income, age, geography, ethnicity, and gender subgroups get weighted against each other: the entire redistribution argument, encoded in a formula
  • ▸Whether environmental quality is a hard constraint or a soft input that can be traded off against consumption

The Metric Design Tradeoff

illustrative
Blunt metric (e.g. raw GDP)
Composite metric (e.g. CPI, Hanson's GDP+)

Metrics simple enough to resist gaming, like raw GDP, miss what people value, while richer composite metrics, like CPI or Hanson's GDP+, capture more value only by opening more room for manipulation, a stylized illustration of the Rethink Priorities Goodhart's-law critique, not measured data.

#COWEN'S OBJECTION, AND THE GOODHART TRADE

Tyler Cowen has made close to this exact argument against Hanson directly: values and beliefs don't separate cleanly in practice, so treating the vote and the bet as two sequential, independent steps isn't a design feature, it's a category error. What counts as a policy 'succeeding' already smuggles in a theory of what matters.

Rethink Priorities' critique sharpens the trap further: a welfare metric simple enough to resist manipulation is too simple to capture what people actually value, and one rich enough to capture what people value is rich enough to be gamed: Goodhart's law wearing a governance costume. GnosisDAO running its pilot in parallel with Snapshot rather than in place of it looks like an organization that has already priced in this exact tension and isn't ready to let a formula decide alone.

#WHERE THE ARGUMENT ACTUALLY MOVES

My objection to futarchy isn't that it's dishonest; it's that it's honest about the wrong stage. The market genuinely can't lie about which policy it expects to score higher on GDP+. What it can't do is tell you whether GDP+ was the right thing to maximize, and that fight doesn't disappear just because it happens once, upstream, instead of every election cycle. If anything it gets worse: a legislative debate about a tax rate or a war is legible to an ordinary voter. An argument over whether welfare should be a discounted square root of GDP or a subgroup-weighted consumption average is legible mainly to whoever wrote the formula.

Hanson didn't design a system that depoliticizes government. He designed one that relocates the politics into the one room where the fewest people are equipped to walk in and argue back, and called that neutral.

Two Real Futarchy Experiments

MetaDAO governance

Markets replace voting

Fully on-chain conditional pass/fail markets (Solana)

MetaDAO status

Ongoing

Hired Robin Hanson as adviser, Feb 2025

GnosisDAO (GIP-145)

Advisory pilot

Runs parallel to existing Snapshot token voting (Ethereum)

GnosisDAO duration

~9 months

Markets inform, but don't replace, the final vote

MetaDAO lets conditional markets fully replace voting, while GnosisDAO's GIP-145 keeps markets advisory alongside its normal token vote, a real difference in how much each DAO trusts the market alone.

#Further Reading