Decision Markets in Governance

Speculative If a conditional market decides at least as well as the committee, board or token vote it replaces — which no head-to-head trial has yet shown, and section 7 says what would — the prize is a governance rule that turns a standing disagreement into one tradeable price and then binds the organisation to whatever that price says. This is a different object from a forecasting market. A forecasting market advises; a decision market decides. The claim is not that a crowd predicts well — that part is measured — but that you can wire the prediction directly to the lever.

Established The aggregation edge the rule would inherit is real and has a size. Over the U.S. presidential race markets of 1988–2004 the Iowa Electronic Markets carried a mean absolute error of 1.6 percentage points against the Gallup poll’s 1.9. Inside firms, Ford’s internal markets cut mean squared forecast error by roughly 25% against the company’s own expert baseline (a mean-squared-error ratio of 0.742). Those are the numbers a decision rule built on markets would be spending.

Frontier What the binding rule adds, and the reason to care, is that it removes a specific failure of committees: the option nobody can be seen to favour never reaches a vote, and the option a senior person favours is rarely killed. A market has no such reticence; it will short a bad proposal that a room of subordinates would wave through. Ben Hawkins’ attempt to have MetaDAO mint him 1,500 governance tokens at roughly a third of spot price was voted down not by a board but by traders taking the other side of his own $250,000, and it stayed down.

Speculative The upside is narrower than the field’s rhetoric, and it is unevenly distributed. Whoever writes the welfare metric the market settles against holds most of the power the rule appears to give away; where markets are thin the wealthy set the price; and the roughly ninety-six binding decisions run on the largest live deployment to mid-2026 are mostly capital-allocation and token-launch calls, not the contested policy questions the theory advertises. The condition for any of it is that a conditional price means what its designers claim — the causal effect of acting, not a correlation with acting — which section 5 argues is the whole ballgame.

Established Futarchy is a decision rule, proposed by the economist Robin Hanson, that separates two questions an organisation usually answers together: what it wants, and which action will get it there. Hanson’s slogan is “vote on values, but bet on beliefs.” Voters or members choose a welfare measure — a number the organisation agrees it wants to go up. Speculators then trade conditional markets, and the action the markets say will most raise that number is the one adopted. The formal statement is his 2013 paper in the Journal of Political Philosophy, “Shall We Vote on Values, But Bet on Beliefs?”, extending a 1999 sketch he called decision markets.

Established The mechanical heart is the conditional, or decision, market: a pair of markets that price the welfare measure under each option. One market pays out on the world in which the proposal passes; the other on the world in which it fails. If the pass-conditional price for the welfare measure is enough above the fail-conditional price, the proposal is enacted. The market outcome is not read to a committee that then decides; it is the decision. That binding step is what makes futarchy a governance mechanism rather than an intelligence-gathering one, and it is exactly the property that distinguishes this brief from ordinary forecasting.

Established This is a different decision rule from the ones the neighbouring briefs cover. A forecasting or opinion market — the subject of the corpus treatment of collective judgement — produces an estimate that a human is free to ignore. Futarchy removes the human’s freedom to ignore it by construction. The related brief on Distributed Governance catalogues token-weighted voting and its concentration problems; futarchy is the proposal to replace the vote on consequences with a market while keeping the vote on ends. The difference is the entire subject here.

Frontier Until 2023 the idea was almost entirely on paper; since then it has been executed at small scale on blockchains, which supply the one thing the proposal always lacked — a way to bind a real organisation to a market verdict automatically. A smart contract can hold a treasury and release it only if the pass market clears a threshold, with no official able to overrule it. That is why the live evidence in this brief comes from Solana and Polkadot rather than from a government or a firm: code, not law, is currently the only thing that will actually execute a market’s verdict against the wishes of the people it governs.

Established The blockchain revival was not a spontaneous invention: Vitalik Buterin wrote up futarchy for Ethereum in a 2014 essay, and Michael Abramowicz’s 2008 book Predictocracy had already worked through a book-length version of governance-by-market. The idea sat dormant for roughly a decade because it lacked an executor, not because it lacked elaboration.

Frontier The name signals both the ambition and the risk: futarchy is government by prediction of the future, and the neologism has always run ahead of the evidence. Naming a form of government before it has governed anything is the founding tension this brief tracks.

Speculative The seductive part, and the part to hold at arm’s length, is the analogy to a market economy: prices already allocate capital across a whole economy without anyone in charge, so why not policy? The analogy smuggles in the claim the brief spends the rest of its length testing — that a conditional price on a rare, one-off, self-referential decision behaves like a price in a deep, repeated, exogenous market. It may not, and the reasons it may not are specific and named below.

Established The evidence for the ingredient — that markets aggregate dispersed information into accurate forecasts — is strong, old, and independent of futarchy. Wolfers and Zitzewitz’s surveys document the Iowa Electronic Markets beating the Gallup poll across four presidential elections, with a final-week absolute error near 1.5 percentage points against the poll’s 2.1. Inside firms, an internal Hewlett-Packard market forecast printer sales better than the company’s own process, and a Siemens market correctly called a software project as certain to miss its deadline that management still officially expected to meet.

Established The single most relevant measurement for this brief is Cowgill and Zitzewitz’s study of corporate prediction markets at Google, Ford and an anonymous “Firm X”, because it is large, independent of any protocol, and it measured both the accuracy and the biases. Across 270 markets at Google, 149 at Ford and 1,345 at Firm X, prices were close to well-calibrated and beat expert baselines; Ford’s markets ran a mean-squared error about 25% below the expert forecast (ratio 0.742, p about 0.10). This is the empirical high-water mark for the claim that a market can outdo the incumbent forecaster.

Frontier The same study is also the clearest measurement of the way these markets go wrong, and the failure is directly relevant to a rule that would bind on the price: an optimism bias. Securities tracking favourable outcomes at Google were systematically overpriced — a bias coefficient of about −0.103 (p<0.05), worst in markets on whether internal projects would finish on time (about −0.239) and mild in demand forecasts (about −0.039). The bias moved with company mood: a 2% rise in Google’s share price, roughly one standard deviation, lifted the price of optimism-tracking securities by three to four percentage points on the day.

Frontier The one piece of good news in that study for futarchy is that the biases shrank as the markets matured. Cowgill and Zitzewitz report that the optimism bias was small early in a market’s life, grew, and then essentially disappeared by the end of the sample, and that “the inefficiencies that do exist generally become smaller over time.” That is an argument that a standing, repeated market learns; it is not an argument about a rare one-off decision market, which never gets the repetitions.

Frontier The live governance evidence — markets that actually bound a real organisation — is thin, recent, and comes almost entirely from MetaDAO on Solana, founded in November 2023 by a pseudonymous developer known as Proph3t. MetaDAO runs a proposal by minting conditional-on-pass and conditional-on-fail tokens; if the time-weighted average price in the pass market sits more than about 5% above the fail market over a trading window of several days, a smart contract executes the proposal with no human veto. By mid-2026 one ecosystem tracker counted about 96 proposals decided this way across roughly 14 organisations, with daily conditional-market volume of $2–8 million and protocol value locked near $13 million.

Frontier Those figures are the best available and they should be read with the vintage and the source in mind, because they do not support the strongest framing in circulation. An earlier independent write-up (analytics firm; not an operator) counted about 21 MetaDAO markets and 32 across all its DAOs on roughly $700,000 of total volume; a mid-2024 explainer counted about ten proposals. The count has grown but the claim of “hundreds of binding decisions” overstates it: tens to about one hundred is what the record shows, and a large share of those are token launches and capital calls run through MetaDAO’s launchpad rather than contested policy. The commission’s anchor is directionally right — futarchy has moved from paper to binding execution — and numerically high.

Handwave The first peer-reviewed assessment, published in 2025, is real but is weaker evidence than its existence suggests, and this brief treats it as a description of the state of play rather than as support for the mechanism. Weidener and Shilina, in Frontiers in Blockchain, ran a retrospective, deterministic simulation over ten past VitaDAO proposals and found that the higher expected-value branch matched the action token-holders actually took in every case. The authors themselves stack the caveats: the outcomes were specified after the fact (hindsight bias), VitaDAO was chosen for its maturity and so is “an upper bound”, the simulation “cannot replicate the endogenous dynamics of live prediction markets” and omits price discovery and belief updating entirely, and the authors disclose holdings in the very DAOs studied and could not blind themselves to the results. They call the finding “suggestive, rather than definitive.”

Established The corporate markets ran on genuinely small trader populations, which matters for any claim about scaling the rule down to a real organisation. Google’s markets drew about 1,465 traders across roughly 70,706 trades and Ford’s about 902 traders across 9,258 trades, while Firm X reached near-total participation — 57 of 58 invitees — and still produced calibrated prices. Aggregation did not need a crowd; it needed the right handful of informed people.

Frontier Not every corporate market was well-behaved, and the exception is instructive. Ford’s features markets showed substantial overpricing of securities trading below 0.7 and a constant term near −0.106, a combined optimism-and-buying-preference distortion, so the “markets beat experts” headline conceals real, measured, market-specific failures.

Frontier The optimism bias matters here in a way it does not for a pure forecaster: a rule that binds on an optimism-inflated pass price will enact projects the market privately expects to slip. The bias Cowgill and Zitzewitz measured runs in exactly the direction that would make a bound decision market too eager to say yes.

Established So the position, stated flatly: that markets forecast well is established; that conditional markets have been built and have bound real organisations to their verdicts is established; that they produce better decisions than the rule they replace has not been measured even once in a live setting. The gap between the second and third sentences is the subject of the rest of this brief.

Frontier The open scientific question is not whether markets aggregate information but whether a conditional price identifies the thing a decision-maker needs: the causal effect of taking an action, rather than the correlation between the action and the outcome. A conditional market on “welfare if we pass” versus “welfare if we fail” reveals P(welfare | passed), and P(welfare | passed) is not in general the effect of passing. This is the identification problem, and it is the live theoretical frontier of the whole subject.

Frontier The problem bites hardest exactly when the decision is informative about hidden states of the world, which is when decisions matter most. The standard illustration: a market may price a firm’s shares lower conditional on a chief executive being fired, not because firing hurts the firm but because a board only fires a good executive when it privately knows something has already gone badly wrong. Condition on the firing and you read the bad news, not the effect of the firing. A rule that bound on that price would keep a failing executive precisely when the market “knew” the most.

Frontier Hanson accepts the problem is real and locates it in a specific sequence — price, then information, then decision — which he calls decision selection. His proposed repairs are concrete: let people with the decision-maker’s information trade; announce the timing of the decision immediately before it is made, so speculators are not pricing in future news; and, most cleanly, condition prices on decisions that are sometimes made at random, so that the market is pricing an intervention rather than an observation. He treats the flaw as one of design, avoidable, rather than fatal.

Frontier The randomisation fix is theoretically sound and practically corrosive, and the tension between those two facts is unresolved. If the organisation must occasionally take the action the market rated worse, purely to keep the price causal, then it is not fully bound by the market after all, and the frequency of randomisation trades off cleanly against the very decision quality the mechanism promises. Critics argue this makes futarchy “parasitic” on an outside decision process that defines the metric and injects the randomness — the market harvests the value of that process without being able to supply it.

Frontier A second open question is reflexivity: once decision-makers reliably follow the price, the price starts pricing in the fact that it will be followed, which can create self-fulfilling equilibria unrelated to the underlying merits. A policy can be approved in good times because approval signals confidence, not because it helps, and the belief sustains itself because trading on the signal is individually rational. Whether real markets fall into these traps or arbitrage them away is not known, because no deployment has been both binding and deep enough to test it.

Speculative A third question is whether a welfare metric robust enough to bind on can be written at all. Any published target invites Goodharting: a metric of well-being can be gamed, and even GDP can be inflated by mandating pointless transactions. Hanson’s answer is to let the values-vote revise the metric over time, but a metric that is frequently revised is a metric traders cannot price far ahead, and one that is fixed is one adversaries can farm. Nobody has demonstrated a metric that is both stable and unfarmable for a contested political question.

Frontier A prior puzzle sits underneath all of this: standard theory says rational traders with common priors should not trade at all — the Milgrom–Stokey no-trade result — so every working market already leans on something outside the clean model, whether noise traders, differing priors, or hedging. Futarchy inherits that unexplained dependence and makes it load-bearing, because the rule only works if trade actually happens on the decisive leg.

Speculative Tyler Cowen’s standing objection is that the separation of values from beliefs is not clean in practice, because how you measure welfare is itself a value judgement. If the metric encodes contested values, the “bet on beliefs” half is quietly deciding values too, and the slogan’s tidy division does not survive a real political question.

Frontier The binding bottleneck is not technical; it is that the one property futarchy needs from its markets — a causal conditional price — is the one property markets do not automatically supply, and every fix for it weakens the binding that is the point. This is the identification problem of section 4 restated as an engineering constraint: you can have a market that is genuinely binding, or a price that is genuinely causal, and holding both at once requires randomising decisions you claimed the market would make.

Frontier Thinness is the second bottleneck, and it is worse for decision markets than for forecasting markets because decisions are rare, lumpy and non-repeating. The Iowa and corporate markets that work are standing markets with many contracts and continuous news; a decision market on a single proposal opens, trades for days, and closes forever. Wolfers and Zitzewitz note that Tradesports contracts on genuinely one-off events — the next Supreme Court retirement, the future of the papacy — drew almost no trade. A market nobody trades cannot aggregate anything, and a rule that binds on it binds on noise.

Frontier MetaDAO has already met the thin-market failure in a named form: the fail-conditional market goes quiet. Once a proposal looks likely to pass, the conditional-on-fail world becomes a counterfactual few will ever be paid on, so it attracts little serious trading and can go, in one operator’s description, nihilistic and noisy — which corrupts precisely the price the pass/fail comparison depends on. The mechanism’s accuracy is set by its least traded leg, not its most.

Frontier Manipulation is a real bottleneck but the evidence says it is a smaller one than intuition expects, which is the strongest point in futarchy’s favour. In a controlled experiment by Oprea, Porter and colleagues, traders paid to push the price could raise it by about seven points when aiming high (p about 0.001) but failed to move it down (p about 0.34) — and, crucially, uninformed observers watching the manipulated price were not misled: forecast accuracy did not differ between manipulated and clean markets (a Chow test statistic of 1.5, p about 0.22). Hanson has argued the general case in a paper whose title is its thesis, that a manipulator can aid prediction market accuracy, by acting as a loss-making noise trader who subsidises the informed.

Frontier But the experimental reassurance was measured on advisory markets, and binding changes the payoff to manipulation, which is why this brief does not carry the finding across unedited. When the price merely informs, a manipulator pays to move a number; when the price decides, a manipulator pays to buy an outcome, and the profit from the resulting policy can dwarf the trading loss. MetaDAO’s own record cuts both ways: Ben Hawkins’ $250,000 push and a $50,000 Pantera-linked bid both failed, which is real evidence of resistance at those budgets — and no evidence at all about a budget ten times larger against a thinner market.

Frontier Wealth-weighting is a structural bottleneck the mechanism cannot design away: a market aggregates beliefs weighted by the capital behind them, so the rich vote their beliefs more loudly. In a forecasting market that is tolerable because accuracy is the only goal; in a governance rule it means the conditional price encodes a wealth-weighted view of consequences, and if wealth correlates with interest the “prediction” carries a preference smuggled in as a belief. This is distinct from manipulation — no one breaks a rule — and it is why critics treat futarchy’s neutrality claim with suspicion.

Frontier Decision markets also lack myopic incentive compatibility, a technical bottleneck with a plain-English form: it can be rational for a trader to bet against what they believe in order to shift the decision and profit elsewhere. In a pure forecasting market honest trading is the sound strategy on average; wiring the price to a decision breaks that guarantee, because the decision changes the world the other bets pay out in.

Speculative Risk premia are a quieter distortion: a conditional price is an expected value struck by risk-averse traders, so it is shaded by how risky each branch is, not only by how good it is. A safe mediocre option can out-price a risky excellent one, and a rule that reads the shaded price as a pure welfare estimate will lean systematically toward the timid choice.

Frontier Hedging is a further contaminant: a trader with a real stake in the outcome trades to insure that stake, not to reveal a belief, and those trades move the decisive price. The more a decision matters to the people nearest it, the more their orders carry hedging rather than information — the opposite of what the rule assumes.

Speculative The longest-horizon bottleneck is that bets do not resolve if the entity ceases to exist, so a market systematically under-prices ruin. A policy that raises welfare for two decades and then collapses the organisation can still clear a market dominated by traders who will have cashed out, and existential outcomes pay nothing to a trader who is dead. A decision rule that cannot price its own tail risk is a poor rule for exactly the decisions — safety, survival, the long term — where getting it wrong is unrecoverable.

Frontier Futarchy depends first on a settled science of the welfare metric it will bind on, and that science does not exist for contested public questions. The mechanism inherits every unsolved problem of social-welfare measurement and adds a live adversary with capital. For a firm the dependency is milder — share price or a revenue KPI is a defensible proxy — which is why the only working deployments bind on token price or treasury value, not on anything a polity would recognise as the public good.

Frontier It depends, second, on a resolution of the identification question that survives contact with self-selected traders. Hanson’s randomisation repair is a research programme, not a settled result: nobody has run a decision market with randomised enactment at a scale that lets you check whether the conditional prices come out causal. Until that is done, the mechanism rests on a theoretical patch whose cost — occasionally doing the thing the market rated worse — has never been priced empirically.

Established It depends, third, on a substrate that will execute a verdict against the wishes of the governed, and that dependency is currently met only by public blockchains. The reason the evidence base is Solana and Polkadot rather than a legislature is not fashion; it is that a smart contract is the only agent that has so far been willing to move a treasury on a market’s say-so with no human able to intervene. A legal or corporate futarchy needs a body that binds itself in advance, and no such body exists at scale.

Frontier It depends, fourth, on liquidity that decisions do not naturally attract, which means it depends on subsidy. Because a decision market is rare and one-off, someone must pay to make it deep — a market-maker subsidy or an automated market maker seeded from the treasury, which is what MetaDAO’s design borrows spot liquidity to provide. The size of subsidy needed to make the price trustworthy, relative to the size of the decision, is unmeasured and may be prohibitive for small organisations.

Established One dependency is already satisfied and worth crediting: the pricing engine itself, Hanson’s logarithmic market scoring rule, is a solved, widely deployed piece of mechanism design that supplies bounded-loss liquidity for exactly the thin markets futarchy needs. The hard parts are everything wrapped around that engine, not the engine.

Frontier The dependency on a willing executor is loosening slightly beyond MetaDAO: Optimism ran a Season 7 futarchy grants experiment and Zeitgeist shipped a live futarchy pallet, so the substrate is diversifying even though none of these has yet produced a decision-quality measurement. A wider pool of comparable decisions is precisely what the eventual paired trial of section 7 will need.

Frontier Finally it depends on empirical work the neighbouring literatures have started but not finished. The corpus brief on Forecasting & Replication covers what is known about calibrated human and market forecasting; the brief on Collective Intelligence covers aggregation of dispersed judgement. Futarchy is the claim that you can take those aggregates and make them binding, and the binding step is the part none of those literatures tests.

Frontier The one test that would move this assessment is a pre-registered paired trial that alternates comparable decisions between a conditional market and the incumbent rule and scores both about eighteen months later, run alongside seeded adversarial trading at announced budgets to price manipulation resistance directly. This is the experiment the whole subject is missing: not another demonstration that a market can bind a decision, which is established, but the first measurement of whether the bound decisions are better than the ones the old rule would have made.

Frontier The design has to defeat the identification problem, not sidestep it, which is why alternation and randomisation are load-bearing rather than cosmetic. Comparable decisions are drawn from a single stream and assigned — ideally at random for a subset — either to the market or to the committee, so that the causal effect of the rule is identified even though the causal effect of any single decision is not. Scoring both arms on the same welfare metric eighteen months out turns Weidener and Shilina’s retrospective, deterministic concordance into a prospective, live-market comparison with the endogenous dynamics they had to omit.

Frontier The manipulation arm is the second half, and it is cheap because it can be adversarial by design: pay traders a known budget to move a live decision and see whether the verdict flips. This is the Oprea–Porter experiment upgraded from advisory to binding — the earlier study found manipulators could push an advisory price about seven points without misleading observers, and the open question is whether that resistance survives when the price decides and the manipulator is buying a policy rather than a number. A budgeted, pre-announced attack answers it at a stated price.

Frontier A natural experiment is already running and should be mined even though it cannot settle the question: the accumulating binding decisions on MetaDAO and the nascent Zeitgeist deployment. Every executed proposal generates a conditional price and, later, a realised outcome, so the calibration of the pass/fail spread against what actually happened can be measured now, without new hardware. What it cannot deliver is the counterfactual — what the committee would have done — because the organisations self-selected into the market, which is exactly the confound the paired trial removes.

Speculative A cleaner but harder observation would be a decision market with genuinely randomised enactment, the direct test of Hanson’s causal repair. Enact the market’s preferred option most of the time but a known fraction of the time flip a coin, publish which decisions were randomised, and check whether the conditional prices on the randomised subset are unbiased for the outcome. No deployment has done this; it is the experiment that would tell you whether a conditional price can be made to mean the causal thing the theory needs it to mean.

Frontier The corporate archive is an under-used near-term source, because a handful of firms have run internal markets for years and hold the paired data privately. Cowgill and Zitzewitz’s Google, Ford and Firm X markets already recorded a market price and, often, a separate managerial forecast and a realised outcome for the same question; the analysis that has not been published is the head-to-head decision-quality comparison where management sometimes followed the market and sometimes did not. Those firms could answer part of this question from data they already own.

Frontier Even a well-run paired trial faces a power problem the design must confront in advance: decisions are few, outcomes are noisy, and eighteen months yields a small sample, so the trial needs many comparable decisions or a pooled multi-organisation design to detect anything short of a large effect. A single organisation’s handful of proposals cannot settle the question, which is itself a finding about how hard the question is to settle.

Handwave The experiment nobody can run is the one that would matter most for the political case: a binding futarchy over a genuine public-welfare metric at the scale of a city or a state. The DARPA Policy Analysis Market, proposed in July 2003 to trade geopolitical risk, was killed within days as “terrorism futures” before any price formed — a governance market ended by politics, not by evidence. Until a polity is willing to bind itself to a metric it cannot easily revise, the public-sector case for futarchy rests on argument, not measurement.

Established The engineering that exists is the conditional-vault-and-automated-market-maker stack MetaDAO built, and it works: it will hold a treasury and execute a verdict with no human in the loop. A proposal mints paired conditional tokens against deposited assets; the pass and fail markets trade on an automated market maker seeded from spot liquidity; a time-weighted average price over a window of several days — five days in one version, ten in another, the parameter has changed across releases — decides enactment if the pass market clears the fail market by about 5%. The time-weighting is a deliberate anti-manipulation device, forcing an attacker to hold a distorted price rather than spike it at the close.

Frontier The unsolved engineering is liquidity provision for the fail leg and for rare decisions, because the current design lets the decisive price be the thinnest one. A robust system needs a market-maker subsidy sized to the stakes and structured so the counterfactual leg stays honest — a problem MetaDAO’s “nihilistic fail market” shows is not yet solved. Proposals to keep both legs liquid until resolution, or to reward accuracy on the losing leg, are on the table and untested.

Frontier The welfare-metric oracle is the hardest component and the least built, because binding requires a number a contract can read and settle on without dispute. Token price is trivially readable, which is why deployments use it; anything richer — a research output, a public-health outcome, a citizen-well-being index — needs a trusted, tamper-resistant oracle, and the moment the metric is legible to a contract it is legible to a farmer of the metric. This is the same dependency the brief on Standards as Governance treats for measurement standards generally, sharpened by a live adversary.

Established MetaDAO’s most-used application is not policy at all but a token launchpad, where a futarchy market decides whether and how to raise capital, and it has real throughput. Roughly $33 million has been raised across about ten launchpad projects; Mountain Capital was the first launch and the first to return capital to holders through the mechanism’s liquidation process. This is where most of the “binding decision” count actually comes from, which is why this brief separates capital calls from contested policy.

Frontier The system started tiny and its per-decision costs are non-trivial, which bounds who can use it. MetaDAO launched with about 65 participants and a $10,000 treasury; running a proposal cost roughly $150,000 before the automated-market-maker redesign, which now borrows about half of spot liquidity to seed each market. A rule whose overhead is measured in six figures per decision is not yet a general-purpose governance tool.

Speculative The listing of MetaDAO’s governance token on a major exchange in May 2026 is a mixed signal for the mechanism. Deeper, more liquid markets help the pricing layer, but a governance token that also trades heavily for speculative reasons imports price noise unrelated to any proposal into the very asset the welfare metric is denominated in.

Frontier Off-chain, the missing engineering is legal, not computational: a body that binds itself in advance to execute a market verdict. A firm could in principle write a charter that delegates a class of decisions to a market, but none has, and the fiduciary and liability questions — who is responsible when the market decides badly — are unresolved. The related brief on Institutional Design covers the general problem of pre-commitment devices; futarchy is an unusually literal one, and the law has not caught up to it.

Established The nearest adjacent technology is the ordinary forecasting or prediction market, from which futarchy borrows its engine and to which it adds the binding rule. Everything known about prediction-market calibration, the Iowa markets and the corporate markets applies to the pricing layer; nothing known about them applies to the enactment layer, which is what makes futarchy a distinct subject rather than a prediction-market application.

Established Automated market makers and conditional-token standards are the enabling substrate, developed for decentralised finance and repurposed here. The logarithmic market scoring rule Hanson himself invented, and the conditional-token frameworks of the Ethereum ecosystem, are what let a rare market bootstrap liquidity and pay out cleanly on one of two worlds. Futarchy is in this sense a governance application riding on financial plumbing built for other reasons.

Frontier Quadratic voting and other mechanism-design tools are siblings rather than components, addressing the same failure — preference intensity that a one-person-one-vote rule cannot see — by a different route. The brief on Distributed Governance records quadratic voting’s one real legislative trial and its collapse on open-meeting law; futarchy shares the ambition to price intensity and the vulnerability to wealth, and differs in that it prices consequences rather than preferences.

Frontier Combinatorial market designs, such as Zeitgeist’s pool that resolves four joint outcomes at once, are the adjacent engineering most likely to make richer welfare metrics tradeable. Pricing welfare jointly under enactment and non-enactment, rather than as two separate markets, is a cleaner way to read the conditional comparison — and it is new enough that no decision record exists to judge it.

Speculative AI forecasters are the adjacent technology most likely to change the picture, because they attack the thin-market bottleneck directly. If calibrated models can trade or seed the counterfactual leg, the rare-decision liquidity problem softens; if they merely predict, they compete with the market rather than filling it. The corpus brief on Collective Intelligence and the one on Forecasting & Replication both bear on whether machine forecasters are good enough to bind on, which is a live and unsettled question.

Frontier Futarchy’s institutional requirement is a body willing to surrender discretion in advance, and the historical record is that such bodies are rare and that they claw the discretion back the moment a market verdict hurts. The pattern is visible in the neighbouring brief on Distributed Governance: DAOs that suffered governance attacks responded by replacing on-chain rule with a multisig of humans. A rule is only as binding as the willingness to be bound, and that willingness is scarcest exactly when the rule matters.

Frontier The mechanism needs an institution that owns the values-vote — the metric — separately from the market, and keeping those two clean is an institutional problem, not a technical one. If the same interests that trade also write and revise the metric, the separation Hanson’s slogan promises collapses. No deployment has an independent metric-setting body; on MetaDAO the metric is token price and the token-holders are the traders, which is convenient and exactly the conflation the theory warns against.

Frontier Regulatory standing is unresolved and has already been fatal once at the level of a public body. The DARPA market died of politics; the corporate case runs into securities and gambling law wherever employees trade on company outcomes with real money, which is why Cowgill and Zitzewitz’s markets used play or limited stakes. A futarchy binding real capital on real policy sits in a regulatory space no jurisdiction has cleanly zoned, and the blockchain deployments operate there precisely because the law has not yet arrived.

Frontier The one institution that has actually surrendered discretion is a smart contract, and its operators are candid that the results are modest. MetaDAO’s co-founder describes the mechanism as running at roughly “80 IQ” — enough to stop a catastrophic treasury decision, not enough for executive judgement — which is a more honest self-assessment than most governance experiments offer and sets the realistic near-term bar.

Speculative A recurring institutional pattern is that experiments keep a human backstop, and every backstop is a partial admission that the rule is not yet trusted to bind. Zeitgeist restricts proposal submission to a governance-approved origin; MetaDAO’s operators intervene in practice; the fully autonomous, fully bound futarchy the theory describes has essentially never been left to run unattended on a consequential decision.

Speculative The institution that could make the strongest honest use of futarchy is a small, well-defined organisation with a clean, hard-to-game metric — a research funder scored on replicated outputs, say — rather than a polity. The corpus briefs on Scientific Governance Models and Institutional Design both point at where a pre-commitment-to-a-market rule might beat a committee, and it is not the contested public square the founding rhetoric aimed at.

Frontier The central ethical objection is that futarchy launders a preference as a belief and then weights it by wealth. Because the conditional price aggregates beliefs in proportion to capital, a wealthy trader with an interest in the outcome can move the “prediction” without breaking any rule, and the mechanism presents the result as a neutral forecast. This is not a bug a better contract fixes; it is what a market is. In a firm that may be acceptable; in the governance of people it is a distributional choice dressed as an epistemics.

Frontier A second objection is that the welfare metric, once fixed, becomes the whole of ethics for the system, and any value not in the number is invisible to it. Goodhart’s problem is not merely that the metric gets gamed but that binding on it makes the un-measured — dignity, fairness, the interests of those who cannot trade — formally irrelevant to every decision. The values-vote is meant to guard this, but a vote that must reduce the good to a single tradeable index is already a strong ethical commitment.

Speculative A third concern is intergenerational and existential: a rule that cannot price outcomes in which the traders are gone will systematically under-weight ruin and the far future. A market discounts what does not pay out, and extinction pays nothing, so the decisions where error is unrecoverable are precisely the ones the mechanism handles worst. For an organisation whose choices carry catastrophic tails, this is a decisive ethical reason to keep a human veto — which is to say, not to be fully bound.

Speculative The Goodhart failure is concrete, not abstract: a well-being metric can be gamed by manipulating the polls that feed it, and even a hard number like GDP can be inflated by mandating pointless transactions — the textbook illustration is a rule requiring citizens to exchange banknotes to book activity. Binding on the metric turns every such gap into an attack surface with money behind it.

Frontier Set against these, the honest ethical case for futarchy is accountability: a market verdict is legible, dated and attributable in a way a committee’s compromise is not. When a bound market makes a bad call the price history says who was wrong and by how much, which is more than most governance rules offer. That transparency is a real good, and it is the part of the founding argument that survives the critiques intact.

Speculative The civilizational version of the claim is that societies could route more of their choices through markets and fewer through committees, treating governance as an information-aggregation problem the way an economy treats resource allocation. If it worked at scale it would shift power from those who deliberate to those who trade, and from institutions that hold discretion to rules that execute automatically. That is a large change and this brief judges its probability low, because the identification, thinness and metric problems of section 5 do not shrink with scale — several get worse.

Speculative The more plausible civilizational path is narrow adoption: specific, well-scoped decisions inside firms, funders and protocols that have a clean metric and can tolerate randomisation, leaving contested public policy to the messy human processes it has always used. On that path futarchy is a useful specialised tool, not a new form of government, and its main effect is to make a class of technical allocation decisions auditable and fast rather than to remake the polity.

Frontier The failure mode worth naming is the one where the tool is adopted for its speed and legibility before the identification problem is solved, and organisations end up bound to causally meaningless prices with a veneer of rigour. A confidently wrong number that executes automatically is more dangerous than an obviously subjective committee, because it is harder to argue with. The corpus treatment of Technocracy and Democracy covers this tension — the substitution of an expert or mechanical process for a political one — and futarchy is its sharpest instance.

Speculative A wealth-weighted decision rule adopted broadly could feed back on inequality itself, because those with capital would set more of the consequential choices and could steer them toward outcomes that preserve capital. This is a slow structural risk rather than a manipulation event, and it is the reason critics treat “let the market decide” as a distributional programme rather than a neutral one.

Speculative If the causal-price problem is genuinely solved — the randomised-enactment result of section 7 — the ceiling rises, but even then the wealth-weighting and existential-underpricing objections cap it below the founding vision. The most defensible long-run picture is a civilization that uses conditional markets widely for bounded, repeatable, metricised choices and refuses to use them for its irreversible ones, which is a real and worthwhile change and a smaller one than “vote on values, bet on beliefs” promised.

These horizons track when the binding-decision-market rule could move from small deployments to tested, then trusted, governance — and where it plateaus.

• 10 yr: Frontier Continued growth of on-chain deployments (MetaDAO, DAO treasuries, Zeitgeist) into the hundreds of binding decisions, and, plausibly, the first pre-registered paired trial against an incumbent rule; the calibration of live conditional prices against realised outcomes becomes measurable even if decision-quality does not.
• 25 yr: Speculative Narrow institutional adoption for well-scoped, clean-metric decisions in firms and research funders, contingent on a randomised-enactment result that shows conditional prices can be made causal; no adoption for contested public policy.
• 50 yr: Speculative Conditional markets a standard tool in the governance toolkit for bounded allocation choices, with AI forecasters supplying liquidity to the thin counterfactual legs; the wealth-weighting and metric problems keep it out of high-stakes political decisions.
• 100 / 250+ yr: Handwave A society that routes a large share of collective choices through markets, the founding vision — possible only if the identification, existential-underpricing and metric-gaming problems are all solved, none of which is on any current research path.

• Depends on the forecasting and aggregation results catalogued in Forecasting & Replication and Collective Intelligence, and on the token-voting substrate and its concentration failures documented in Distributed Governance; nothing else on this map blocks it, because the binding step is what this brief owns.
• Requires (not on this map) a pre-registered randomised trial comparing a conditional market against the rule it would replace; an organisation that binds itself in advance to execute the verdict; a welfare metric a contract can settle on that a funded adversary cannot farm; and a standing liquidity subsidy large enough, relative to any plausible attack budget, to keep the decisive leg honest.
• Enables auditable, dated, attributable decisions for bounded allocation problems; a research funder or protocol treasury that executes on a market verdict rather than a committee compromise; and a measurable calibration record for governance choices that today leave none.
• Adjacent to quadratic voting and mechanism design in Distributed Governance, to expert-versus-mechanical rule in Technocracy and Democracy, and to pre-commitment devices in Institutional Design.

Established “Prediction markets do not work / are always manipulated.” False as stated: the Iowa markets beat the Gallup poll across four elections and corporate markets beat expert baselines by about 25% on mean squared error, and controlled experiments find that even paid manipulators fail to mislead observers about the underlying value. The real problem is narrower and is about binding, thinness and causation, not about whether markets forecast.

Speculative “Futarchy has already been proven to make better decisions.” No. The 2025 peer-reviewed assessment is a retrospective, deterministic simulation with the authors’ own caveats — hindsight-specified outcomes, a hand-picked DAO, no live-market dynamics, disclosed conflicts — and it calls itself “suggestive, rather than definitive.” No live head-to-head against an incumbent rule has ever been run.

Frontier “Hundreds of binding futarchy decisions have been made.” Overstated. The largest deployment shows tens to about one hundred binding proposals by mid-2026, many of them token launches and capital calls rather than contested policy; the mechanism has genuinely moved from paper to execution, but the count is smaller and narrower than the headline suggests.

Frontier “A conditional market tells you the effect of a decision.” Not in general. It tells you the welfare expected in the world where the decision happens, which mixes the effect of the decision with whatever the decision reveals about hidden states. Making the price causal requires randomising enactment, which weakens the binding that is the point — the central unresolved tension of the whole subject.

Frontier “Manipulation is the fatal flaw.” Probably not the fatal one. The experimental and field evidence says budgeted manipulation is surprisingly hard to make stick in a traded market, and a manipulator can even improve accuracy by subsidising informed traders; the deeper problems are identification, thin one-off markets, and wealth-weighting, which need no rule-breaking at all.

Established “Futarchy is a fringe idea with no serious pedigree.” False: it was proposed by an academic economist in a peer-reviewed philosophy journal, elaborated in a 2008 scholarly book (Predictocracy), and taken up by the founder of Ethereum. The criticism it deserves is empirical thinness, not intellectual unseriousness.

Speculative “It is neutral because the market, not a person, decides.” The neutrality is only as good as the metric and the wealth distribution behind the price. Whoever writes the welfare measure sets what “good” means, and capital sets whose beliefs count; the mechanism moves the political choice from the decision to the metric and the entry rules, it does not remove it.