Exit, Voice, Veto


I. Participation without power

Freedom begins where refusal becomes survivable.

Institutions are generous with participation when participation changes nothing. They survey, consult, convene, invite testimony, open a feedback channel, and thank everyone for their candour. The vocabulary rotates—engagement, inclusion, co-design, consultation—while the arrangement stays exactly where it was. People are invited to speak after the decision has acquired a budget, a timetable, a legal opinion, and several pieces of furniture.

Call it Participation Without Power. You may be heard while unable to leave. You may complain while unable to stop the harm. You may appeal to the people who made the decision, who are usually delighted to agree with themselves.

None of that is worthless. It builds records, alliances, and pressure. It is not yet control.

An institution is non-captive only to the extent that the people subject to it can refuse without losing what they need to survive, stay legible, and act with others. The question is not whether they were invited to the meeting. It is what happens when they say no.

The wager of this essay fits on an index card. For a defined class of decisions, whether people hold usable exit, voice, and veto predicts whether a harmful decision gets paused, revised, or reversed. The number of town halls does not. If town halls turn out to predict it just as well, the thesis is wrong, and Section VIII says how you would catch it.

Urrepublika names the test after its three powers: Exit, Voice, Veto.


II. After loyalty

The title borrows, with respect, from Albert O. Hirschman’s Exit, Voice, and Loyalty (1970). Hirschman treated loyalty as the force that delays exit and keeps members invested long enough to use voice. The insight holds; people stay out of care, commitment, and shared history.

But staying is ambiguous evidence. Under dependency, people stay because leaving would cost them housing, income, health care, immigration status, records, safety, or community. There, loyalty is the flattering name dependency gives itself.

Urrepublika does not reject loyalty. It declines to treat endurance as consent, and it replaces Loyalty with Veto, because a captive arrangement needs more than an ethical reason to stay or speak. It needs a way to force a decision back into contestation. Voice puts dissent on the record. Veto makes the record matter.

The triad is not a menu of lifestyle options. It is a diagnostic of institutional power:

PowerQuestion
ExitCan a person leave without being destroyed by leaving?
VoiceCan they contest the decision where it can still change?
VetoCan they stop or return a harmful decision before the harm becomes irreversible?

Each power works only through guarantees. Exit needs Portability and Fork. Voice needs Recall. Veto needs Appeal, plus something no guarantee supplies by itself: a reason for the institution to comply.


III. Exit

Exit is the ability to leave an institution without being destroyed by leaving.

A formal right to leave does not meet that standard. The worker who can resign only into homelessness, the tenant who can move only into a worse market, the patient who loses care by challenging a provider, and the migrant whose status is tied to an employer have each been granted permission to disappear. It is administrative clearance for abandonment, stamped and dated.

Exit starts to mean something when people can leave with enough support and continuity to remain socially real. Two guarantees do that work.

III.1 Portability

Portability protects continuity across departure: records, qualifications, data, contacts, benefits, creative work, case history, and the evidence of what happened.

Privacy, confidentiality, safety, and other people’s rights set real limits. The principle is narrower and harder: an institution may not monopolize your usable past in order to control your future.

The law has found this principle and installed it at roughly knee height. Since September 22, 2024, Quebec’s Law 25 has let people request the computerized personal information collected from them in a structured, commonly used format, but not what the organization created or inferred about them—so the profile that most shapes how you are treated is the part that stays behind. GDPR Article 20 draws a comparable boundary around data a person provided, and neither rule moves a reputation, a body of work, or a community.

Mastodon shows the same gap in software. Move your account and your followers are redirected; your posts stay put “due to technical limitations,” and follows, blocks, and mutes travel only by manual export and import. The audience comes along. The archive does not.

Exit without Portability is exile. Portability without Exit is a record in a locked room.

III.2 Fork

Fork differs from Exit in one decisive respect. Exit is one person leaving. Fork is leaving together, with enough of the group’s shared work to build another arrangement.

A fork can carry code, standards, procedures, archives, or publicly created knowledge—not private data or someone else’s work. It asks whether people who made something together can reorganize without starting from ash.

In September 2010, OpenOffice.org community members formed The Document Foundation and released LibreOffice from the shared codebase, carrying a common inheritance into a new house. That shows a fork can be done. Whether the mere possibility of one keeps incumbents polite is a separate question, and a single case cannot answer it.

Fork without Exit leaves collective work captive. Exit without Fork rescues the individual and surrenders the common inheritance to the institution they fled.


IV. Voice

Voice is protected, collective access to the point where a decision can still change.

A feedback form is not a channel of power. A consultation held after the budget, route, closure, contract, or development is settled is a suggestion box bolted to the back of a moving truck. Voice needs four things:

  1. People can speak, organize, disclose, and criticize—alone and together.
  2. What they say reaches a decision point, not a ceremonial inbox.
  3. They are protected from retaliation.
  4. Their speech can alter the decision, directly or through a route to Veto, Appeal, or Recall.

The second is where most voice goes to die. A person who speaks alone may be brave. A group that can organize and reach the decision point is beginning to be a nuisance, which is what institutions call power when they are being polite.

IV.1 Recall

Recall is the guarantee that makes delegated voice answerable. It is not a passing majority’s right to punish dissent. It means authority is held in trust and can be withdrawn through fair, usable procedures, with due process for the person facing recall. A representative, manager, board, or administrator who cannot be replaced has stopped being a delegate and started to resemble a landlord.

British Columbia’s Recall and Initiative Act is instructive precisely because it exists. A petition needs signatures from more than 40% of those entitled to sign in the member’s district within 60 days, and Elections BC reported in July 2026 that not one of 31 recall attempts since 1995 has met the Act’s requirements. The nearest miss was 1998, when Paul Reitsma resigned before certification; the latest, against Dallas Brodie, needed more than 15,000 signatures and gathered just under 3,700.

Thirty-one attempts and no removals do not prove the route is broken. Some petitions were weak, and a threshold that deters harassment is doing part of its job. They do show a route that is possible on paper and has never once reached the finish line its own rules drew. Recall that exists only as a statute is a door painted on a wall—useful mainly for telling people there is a door.

Voice without Recall produces representation that cannot be corrected. Recall without Voice produces replacement without deliberation.


V. Veto

Veto is a bounded power to halt, delay, or return a decision that imposes concentrated and foreseeable harm.

It is not a private right to protect your comfort against everyone else’s need. It is a rule that certain decisions cannot proceed until the people exposed to their consequences have a meaningful route to stop, revise, or appeal them.

A usable veto has four parts:

ElementInstitutional question
StandingWho may invoke it?
TriggerWhat harm, risk, or irreversibility activates it?
Return pathWhat must the institution do when it is invoked: pause, disclose, negotiate, revise, or abandon?
EnforcementWhat makes the decision-maker comply when compliance becomes inconvenient?

Most accounts of consent specify the first three and assume the fourth, as though formal rules obeyed themselves. They do not. They wait to see who is watching.

Consultation without veto is testimony before a machine that has kept its engine running.

Section 7 of British Columbia’s Declaration on the Rights of Indigenous Peoples Act authorizes agreements under which a statutory decision is exercised jointly with an Indigenous governing body, or only with its consent. Under the first, signed with the Tahltan Central Government in June 2022, the environmental assessment certificate for the Eskay Creek mine restart required Tahltan consent. The first decision came in January 2026, and the Tahltan said yes.

That settles standing, trigger, and return path. It does not yet settle enforcement, because the consent route has not been asked to survive a no. This is not a criticism of the agreement. It is honest bookkeeping: a veto that has never refused is judged on its design and flagged untested. Consent means most where withholding it is survivable, and you only find that out when someone withholds.

V.1 Appeal

Appeal is Veto’s corrective guarantee. It guards against two opposite failures: authority that certifies itself, and blockage that answers to no one.

A real appeal has reasons, accessible evidence, clear timelines, independent review, and remedies with teeth. A procedure in which the original decision-maker admires their own paperwork from another angle is not an appeal. It is a mirror with office hours.

Canadian administrative law already names the first requirement. In Vavilov (2019 SCC 65), the Supreme Court of Canada held that a reasonable decision must be justified, transparent, and intelligible to those it affects. A decision nobody can understand cannot be contested—but reasons are the doorway, not the room. Independence, timelines, and remedies still have to exist, and cost and delay can make a reviewable decision unreviewable in practice.

Veto without Appeal hardens into blind blockage. Appeal without Veto delivers remedy after the harm.

V.2 Capacity to force return

Which leaves the awkward fourth question: what makes anyone comply? The answer offered here is blunt. A veto binds only where refusing it creates credible return pressure—a cost, interruption, reversal risk, or loss of authority the institution cannot safely absorb. Call that the capacity to force return: the ability to drag a decision back into contestation whether or not its authors feel like revisiting it.

That capacity does not come from being right. It comes from organization, reserves, standing, access, votes, and the ability to withhold something the institution needs—the seven proxies first applied in The Domus Account. A strike, a court order, or a budget hold can each do the job by a different road.

The framework separates four things that institutions like to blur together:

  1. a right written into policy;
  2. a person or group authorized to invoke it;
  3. the material capacity to survive invoking it; and
  4. evidence that the institution actually returned to the decision.

A formal veto missing the third and fourth is consultation wearing a hard hat.

V.3 Misuse and standing

Veto serves domination as readily as resistance. The United States Senate filibuster, the Polish–Lithuanian liberum veto, and exclusionary land-use obstruction all warn against mistaking every blockage for democracy. A power that lets the housed stop housing for the unhoused is not a check on power. It is power.

So the design rule, offered as a heuristic rather than a finding, is short enough to carry in a pocket:

Standing follows exposure to harm. Strength follows irreversibility.

Standing follows exposure—not property, title, proximity to a microphone, or the ability to retain counsel. Strength follows irreversibility: the more permanent, bodily, ecological, or livelihood-threatening the harm, the stronger the right to pause.

Exposure runs in both directions. It includes the people a change would harm and the people the status quo already harms, such as those shut out of housing by an existing allocation. An incumbent’s veto over change has to be weighed against everyone the unchanged arrangement is already hurting.

The rule does not settle every dispute. It forces a disciplined argument about whose exposure counts and why, which beats letting the best-resourced incumbent define “affected” as “already in possession.”


VI. The system

The four guarantees do not compete with the triad. They are what make it usable.

PowerGuaranteeWhat the arrangement tests
ExitPortability, ForkCan a person, or a group, depart without erasure or forced atomization?
VoiceRecallCan people contest authority and replace delegates who stop answering to them?
VetoAppealCan exposed people halt or reopen a harmful decision, and have the halt reviewed?

Enforcement is not an eighth capacity. It is the condition under which a veto or an appeal actually changes something, and it comes from the capacity to force return.

Each capacity faces the same four plain tests. Does a real route exist? Can the people affected understand it, trigger it, and get the record? Can they use it without retaliation or ruin? Has using it ever changed a decision? A right that passes the first test and fails the other three is furniture.

The answers are never added up into one flattering number. A generous Exit does not offset an absent Veto where the harm is irreversible; a person who can leave a poisoned watershed has not been protected from the poison. The point is to find where refusal fails, not to award the institution a participation ribbon.

Before the roster goes live

Fiction. The agreement, Service, and events are invented. This follows “The shift that moved” (The Domus Account); it also stands alone.

The new roster was due to go live on Monday. By Friday, thirty-one workers had signed a letter asking the Service to pause it. The director thanked them for speaking plainly and offered a meeting on Tuesday. It was the earliest available slot after the matter was settled.

“After it goes live,” Nadia said.

“After we have information,” said the director. The implementation notice had given him thirty-four pages of it.

The software assigned visits in shorter blocks and moved workers between districts to fill the gaps. The Service said it would waste fewer paid minutes. It had no field for minutes worked by relatives. Nadia showed the director Thursday’s notice on her phone. She did not show him the messages beneath it. Her sister had written about the clinic; her neighbour had written that he could take the boy, but only this once.

Mr. Vale, who received a morning visit, had come with the users’ council. He disliked the proposed roster for another reason. When a regular worker changed, he had to explain again where the bathroom rail came loose and which cupboard held his medication list. He also disliked the workers’ proposal to postpone Monday’s launch. Under the old roster he had missed two visits that month. “You want me to keep a system that already fails me,” he said. The minutes would later describe this as a request for clarification.

He was right. The workers had written restore the old roster because it was shorter than describing a new one.

Their agreement with the Service allowed a safety representative to request a temporary hold on a roster change. It required a review within two working days. It said nothing about who covered the visits during the hold. The director read that sentence aloud and waited. He had found the one clause on which everybody could agree: an unfunded pause could make Mr. Vale pay for a dispute he had not started.

Jo, the workers’ representative, put a second sheet on the table. Twenty workers had offered to keep their existing visits for two days. Four trained relief workers could cover the missed visits, provided the Service paid the relief rate already set aside in its vacancy budget. The users’ council had checked the list for people whose visits could not move. Mr. Vale’s name was on it.

The director said the vacancy money was reserved. Jo asked what vacancy it was reserved for. Nobody at the table knew. For several seconds, the vacancy enjoyed more protection than any person in the room. The budget officer, called from another floor, found that the reserve had been used twice that year and still held enough for three days. The director agreed to the two-day hold while they checked the proposed routes with workers and service users together.

At the end of the meeting Mr. Vale signed the review request. Nadia signed it too. He wanted a dependable morning visit; she wanted a schedule she could live inside. Neither had received a promise that those were the same thing. On Monday morning, his regular worker arrived at eight. The director’s Tuesday meeting remained on the calendar, where it could finally occur before the decision it was meant to discuss.


VII. Decay: Forgetting

Capacities do not stay installed because someone once declared them.

In the Anastomotic Circuit, Forgetting—the process by which a safeguard keeps its name and loses its force—is not a clerical accident. It is a political process with beneficiaries. Recall thresholds never fall to meet a changed electorate. Appeal bodies are captured by the agencies they review. A portability right survives in law while its fee or format makes it unusable.

The wager is that, without periodic re-establishment, each capacity drifts from working toward partial or absent, and that the drift favours the institution over the people subject to it. This is asserted, not observed; the test it would have to pass belongs to the Circuit. No one needs to schedule the erosion. It keeps its own calendar.

The practical rule does not wait for the test. Every safeguard needs a date, an owner, a review route, a public record of use, and a scheduled re-establishment. A right that cannot be found, understood, or enforced has already begun to leave the building.


VIII. How this argument can lose

An essay that praises refusal had better be able to take one. The test runs against a defined decision, not against an engagement report written by the institution being tested. Code the powers before anyone knows how the decision turned out, count the consultations separately, and see which predicts a change of course. One case illustrates. It does not test.

The serious rival is the boring one: resources. Groups with real exit, voice, and veto are usually better resourced, and resources alone may explain why their decisions come back. If reserve depth on its own predicts return as well as the full triad, this essay loses to a simpler materialist account, and it should lose gracefully. Elite quarrels, market shifts, and embarrassment can also turn a decision around without any exposed group lifting a finger. None of these tests has yet been run; the protocol is in the Receipts drawer below.

Nor is this a plea for frictionless institutions. Institutions that affect other people should meet friction. The claim is only that the friction should belong to the people who bear the consequences.

VIII.1 The seven questions

Run against a single decision, the whole apparatus fits in seven questions:

  1. Can people leave without being destroyed?
  2. Can they take enough of themselves to begin again?
  3. Can they build another arrangement from what they made together?
  4. Can they speak where the decision can still change, without punishment?
  5. Can they replace those who govern them?
  6. Can they stop a harmful decision?
  7. Can they have it reviewed by someone other than its authors?

Where the answers are no, participation is decoration placed over dependency.

Where the answers are yes—and have been shown to be yes by someone who refused—freedom has acquired an institutional form.


Receipts: evidence tier, claims and falsifiers

Evidence tier: Falsifiable. Four claims carry status grades; falsifiers are declared before any evidence is coded; Section VIII and the tables below state what would withdraw the argument. No falsifier has yet been run. Evidence Protocol · Theme Explorer

Test status: Not run. Outcome: Unresolved. UR-EVV-001, UR-EVV-002 and UR-EVV-004 remain Asserted; UR-EVV-003 remains Heuristic and is subject to review, not an empirical pass/fail label. Label review: 22 September 2026; no new test or result.

Claim identifiers, status grades, and the falsifier convention follow The Evidence Protocol. Legal, technical, and historical cases on this page are documentary illustrations, not completed comparative tests. Corrections are published as dated errata rather than silent edits.

Claims register and falsifiers

IDGradeClaimFalsifier
UR-EVV-001AssertedFor a defined class of decisions, the presence of usable exit, voice, and veto predicts whether a harmful decision is paused, revised, or reversed. The volume of participation does not.Across comparable cases coded before outcomes, triad codes fail to predict return (pause, revision, or reversal) better than participation volume does; or high participation volume with absent Veto predicts return as reliably as substantial Veto.
UR-EVV-002AssertedA veto binds only where refusing it creates credible return pressure—a cost, interruption, reversal risk, or loss of authority the institution cannot safely absorb. Under the untested four-of-seven hypothesis, it is predicted to bind where four or more of the seven proxies are coded substantial for the group holding it, and to be routinely overridden below that line.Vetoes held by groups with fewer than four proxies coded substantial are honoured under refusal at rates comparable to vetoes held by groups at or above four.
UR-EVV-003HeuristicStanding follows exposure to harm. Strength follows irreversibility.None. A heuristic is not falsified; it is reviewed. Review is triggered if exposure-based standing, applied across a decision class, systematically hands blocking power to groups who are less exposed than those the decision would protect.
UR-EVV-004AssertedWithout periodic re-establishment, each capacity drifts from substantial toward partial or absent, and the drift favours the institution over those subject to it.Capacities coded substantial retain that code across two or more review periods with no re-establishment, across a class of comparable institutions (Falsifier B of the Anastomotic Circuit, which fails the claim if safeguards keep their protective force across generations without being re-established).

Grades at close: UR-EVV-001, 002, 004 — Asserted. UR-EVV-003 — Heuristic. No falsifier has yet been run.

Capacity-to-force-return proxies

Closed before any case was read and first applied in The Domus Account. The proxies are not a scorecard for righteousness; they locate possible sources of enforcement. UR-EVV-002 adopts the untested threshold declared there: a veto is predicted to bind where four or more proxies are coded substantial for the group holding it.

ProxyWhat it measures
Organized membershipCapacity to coordinate affected people into sustained collective action.
Withholding capacityAbility to withhold labour, participation, data, consent, payment, or cooperation.
Reserve depthTime, money, care, legal support, and material slack that make refusal survivable.
Legal standingAccess to enforceable rights, review, remedy, and representation.
Institutional accessA recognized route into the decision process, its records, and its implementation.
Electoral weightAbility to change who holds delegated authority.
Chokepoint controlControl over a dependency whose interruption forces the institution back to negotiation.

Threshold clarification — 22 September 2026. Four of seven is the original, fixed hypothesis to be tested, not an empirically calibrated decision boundary. The published rule supplies no empirical derivation for choosing four rather than three or five. Counting each proxy once is a modelling choice; it does not establish equal causal weight, independence or interchangeability. No person’s standing, right to refuse or entitlement to protection depends on crossing this threshold.

Prospective sensitivity checks — added 22 September 2026. Retain the original four-of-seven rule as the primary hypothesis. Before the next outcome comparison, separately specify checks at three and five, a reserve-depth-only comparison, and a small, justified set of proxy combinations that could reinforce or substitute for one another. Record the decision class, coding rules, observation window and outcome in advance; report every specified comparison. These additions were not part of the original declaration and no results are claimed.

Two different sets of seven. Institutional capacities are Exit, Portability, Fork, Voice, Recall, Veto and Appeal. Enforcement proxies are organized membership, withholding capacity, reserve depth, legal standing, institutional access, electoral weight and chokepoint control. The four-of-seven hypothesis counts enforcement proxies only. The seven institutional capacities are not summed.

Coding scheme

For a defined group, decision class, and period, each of the seven capacities—Exit, Portability, Fork, Voice, Recall, Veto, and Appeal—is graded against four conditions: possible (a real route exists); legible (affected people can understand the route, trigger it, and obtain the relevant record); protected (they can use it without retaliation, ruin, or loss of standing); and consequential (its use has altered, paused, reversed, or materially constrained a decision). Enforcement is the condition under which Veto and Appeal are coded consequential, measured through the proxies above.

  • Absent: no usable route exists.
  • Partial: a route exists but fails one or more conditions in ordinary use.
  • Substantial: the route is possible, legible, protected, and consequential, with an identifiable enforcement basis.

A route that has never been used against the institution’s preference is coded on its design and flagged untested; it cannot be coded substantial on design alone. The Tahltan consent route is coded this way. The BC recall route is coded possible and has never been consequential by its own terms. The seven institutional-capacity codes are never summed into a single score. The separate count of substantial enforcement proxies serves only the declared, untested threshold hypothesis.

Test protocol

  1. Define the case. Name the decision, decision class, institution, period, and exposed group, identifying exposure by who bears the harm (rule in V.3).
  2. Code before outcomes. Before the outcome is known, or before anyone coding has read it, code the seven capacities (absent, partial, substantial, with untested flags) and the seven proxies. Date the codes and record the evidence for each.
  3. Record participation volume separately. Consultations held, submissions received, and participants reached are the rival predictor for UR-EVV-001.
  4. Observe the outcome. Proceeded unchanged, paused, revised, or reversed; whether harm was imposed; whether retaliation followed; whether repair occurred.
  5. Compare. Assemble comparable cases within a decision class—evictions, service closures, platform policy changes, extraction permits—before testing any claim. One case illustrates; it does not test.

Rivals and scope

Resource confound is the strongest rival: groups with substantial exit, voice, and veto are usually better resourced, and resources alone may explain return. The proxy coding exists partly to separate these; a finding that reserve depth alone predicts return as well as the full triad would weaken UR-EVV-001 and strengthen a simpler materialist account. Other rivals: elite conflict (decisions reversed because powerful actors disagreed), market shift (decisions abandoned because they stopped paying), and media salience (decisions returned because they became embarrassing). A case where return is fully explained by a rival is coded as such and does not count in the claim’s favour.

Scope: institutions whose decisions impose consequences on people who did not make them—employers, landlords, platforms, public agencies, schools, unions, associations, and states. The model does not claim that every institution should be frictionless.

Case limits: the fork case shows a community carrying a common technical inheritance into a new institution; it does not show that fork availability disciplines incumbents. The Tahltan case establishes standing, trigger, and return path in statute and agreement, not enforcement under refusal. Vavilov establishes reasons, not the independence, timelines, and remedies an appeal also needs.

Voice revision 2026-09-17. This revision changed register and moved the apparatus into this drawer. It did not change any claim, grade, or falsifier.


Related topics

  • Hirschman’s exit–voice trade-off
  • Veto player theory
  • Ostrom’s design principles for collective-choice arrangements and conflict resolution
  • Free, prior, and informed consent
  • Data portability law
  • Forking in open-source governance
  • Reasonableness review in Canadian administrative law

Sources used

Further reading