Who Actually Decides
Rule changes in protocol economies are produced by coalitions, not by communities. Identifying the coalition is the first step in any honest governance analysis.
Keywords: rule change, governance coalition, protocol governance, blockchain governance, reference implementation, repository control, maintainer authority, foundation governance, sponsor influence, exchange coordination, validator adoption, miner signalling, soft fork, hard fork, activation mechanism, BIP, EIP, governance capture, decision rights, coalition formation, institutional economics, Olson, collective action
A protocol does not change its own rules. People change them. The change happens through a sequence of decisions made by identifiable parties, even when the parties are dispersed and the decisions are presented as the output of an emergent community process. The set of parties whose coordinated action makes a rule revision effective is what I will call the rule-changing coalition. Identifying the coalition for any given protocol is the first step in any honest governance analysis. It is also the step the field most often skips, partly because it produces a less flattering picture than the alternative, and partly because the analytical vocabulary for doing it has not been settled.
The argument of this essay is simple to state and uncomfortable to apply: rule change in operational protocols is the output of identifiable coalitions, not of community consensus. The coalition for any given change consists of the parties whose coordinated action is required to move the rule from proposed to effective. Anyone outside this coalition is observing the change, not producing it. Failing to distinguish the producers from the observers is a mistake in description and a mistake in policy. I will set out the structure, work through the components of typical coalitions, address the standard counterarguments, and end with what the analysis implies for governance reporting and for participants who must evaluate protocols.
1. Why coalitions, not communities
The dominant rhetoric of the field describes governance in language that suggests an undifferentiated community of participants whose collective preferences determine outcomes. The vocabulary varies — “the community,” “the network,” “the ecosystem,” “the users” — but the structure is similar. Outcomes are presented as the aggregate of dispersed individual decisions, with no specific actors identified as the proximate cause of the change.
This vocabulary is wrong as a description of how rule changes actually happen. Rule changes happen because specific identifiable parties take specific identifiable actions: maintainers merge code, foundations publish roadmaps, validators run new clients, exchanges support new transaction formats, infrastructure providers update their integrations. The aggregate outcome looks like a community decision because many parties are involved, but the involvement is not symmetric and not interchangeable. A subset of the involved parties is doing work that the others cannot do, and the subset’s coordinated action is what makes the change effective. The other parties are responding to the action of the subset; they are not producing the action.
This distinction matters because the costs and consequences of rule change are not distributed symmetrically across the involved parties. The parties whose action produces the change have authority and bear responsibility. The parties whose participation is observational are exposed to the change but do not produce it. Treating the two groups as if they were the same — as if “the community decided” were a complete description — obscures who has the authority and who bears the responsibility. The obscuring is not accidental. It is the rhetorical work that the community-language is doing.
The proper analytical move is to identify, for any given rule change or for the system in general, the coalition whose coordinated action produced or would produce a change. This requires asking, concretely: if these parties refused to act, would the change occur? If yes, the parties are not necessary to the coalition. If no, they are. The coalition is the set of necessary parties. Everyone else is observing.
The coalition is typically small relative to the population of participants. This is not a defect of the framework; it is a feature of how rule change actually works in any institutional system. The relevant decisions are made by the parties with the relevant authority, and most participants do not have the relevant authority for any given decision. This is true of corporate governance, public administration, regulatory rule-making, and standards setting. It is also true of protocol governance, despite the rhetoric that suggests otherwise.
Olson (1965) made a closely related observation about collective action: groups whose members have heterogeneous interests and capabilities produce outcomes through the action of subgroups, not through the action of the group as a whole. The subgroup that does the work is typically smaller and more concentrated than the nominal membership of the group. Olson’s framework was developed for political and economic associations, but it applies directly to protocol governance, where the work of producing rule change is done by a small subset of nominal participants with the relevant authority.
2. The components of a typical coalition
What does a rule-changing coalition consist of, in operational protocols? The composition varies, but several components recur across systems with sufficient regularity that they can be enumerated.
Reference implementation maintainers. Most operational protocols have a dominant client implementation, and the maintainers of that client exercise effective authority over which proposed changes are merged into the codebase, when releases are tagged, what version designations the releases receive, and what changes are deprioritised or rejected. Repository permissions are formal: there are merge accounts, commit accounts, and review accounts, and the holders of those accounts are identifiable. The maintainers are not the only parties who write code, but they are the parties whose decisions determine which code becomes the reference implementation. Without them, no rule change reaches the canonical client. They are necessary to the coalition.
Sponsors and funders. Maintainers are paid. The entities that pay them — foundations, corporate sponsors, grant programmes, ecosystem treasuries — exercise influence over what gets prioritised, what gets staffed, and what gets sustained. The influence does not require explicit direction in any individual case; it operates through hiring decisions, budget allocations, programme designs, and the strategic plans that funders publish. A change that the funders prioritise gets development resources and attention; a change that the funders do not prioritise often does not. The funders are not always necessary to the coalition for a specific change, but they are typically necessary for sustained efforts to change the system.
Foundations and standards bodies. Many protocols have a formal foundation or comparable entity that publishes specifications, manages improvement-proposal processes (BIPs, EIPs, and their analogues), maintains official documentation, and coordinates events at which substantive governance discussion occurs. The foundation is typically a legal entity with directors, employees, and a charter. The directors and senior employees of the foundation exercise governance authority directly: by deciding which proposals advance through the official process, by interpreting the protocol’s history and direction, by representing the protocol externally, and by curating the standards that other parties treat as authoritative.
Adoption-critical infrastructure. A rule change does not become effective until it is adopted by the parties whose adoption is required for the change to function: the largest validators or miners, the largest staking services, the largest mining pools, the dominant infrastructure providers, the major wallets, and the major exchanges. Each of these has the practical capacity to refuse to adopt a change, and a refusal by enough of them prevents activation. The set of parties whose adoption is necessary is typically small — a handful of exchanges, a handful of staking providers, a handful of pools — and is identifiable through publicly available data on capacity shares.
Exchange and listing authorities. Exchanges decide which transaction formats they support, which forks they list under which ticker, which versions of a protocol’s native asset they recognise as canonical, and what compliance treatment they apply to new features. The decisions are not technically necessary for the protocol to function, but they are economically necessary for the protocol’s native asset to retain liquidity and for participants to transact at scale. An exchange consortium that refuses to support a particular fork can effectively prevent its emergence as a viable economic entity, regardless of the technical merits.
Wallet and infrastructure software. Wallet providers and infrastructure software (block explorers, indexers, archive services, oracle networks) determine which protocol versions are accessible to ordinary participants and which transactions appear in the data feeds that participants use to make decisions. A wallet that supports only one version of a protocol, or an indexer that displays only one fork, exercises effective authority over participant experience even when the underlying protocol nominally supports both.
Narrative and classification authorities. The framing of a change — as a bug fix, an upgrade, an emergency intervention, a hard fork, a soft fork, a routine release — is itself a governance act. The classification structures expectations about what coordination is required, who is expected to follow without active assent, and what dissent looks like. The parties who control the narrative — typically the foundation, the dominant maintainers, and influential commentators with reach into the participant population — exercise authority that does not appear in the formal process but that materially affects whether and how a change occurs.
For any given rule change, the rule-changing coalition is some subset of these components, depending on what the change requires. A change that affects the consensus rules requires adoption by validators or miners, which puts the adoption-critical infrastructure in the coalition. A change that affects the transaction format requires support from wallets and exchanges, which puts those parties in the coalition. A change that requires sustained development effort requires support from sponsors and foundations, which puts those parties in the coalition. The relevant subset is identifiable for each change once the analyst asks the right question: who must act for this to happen?
3. The coalition is identifiable
The coalition is sometimes described as informal, distributed, or emergent in ways that suggest it cannot be identified. This is a partial truth used to draw a fuller conclusion than it supports. The coalition is informal in the sense that its membership is not declared in advance, and it is distributed in the sense that its members hold authority through different mechanisms. But it is not invisible, and identifying it is generally a tractable empirical exercise.
For reference implementation maintainers, identification is direct: repositories are public, merge permissions are visible, release commits are signed, and the parties who hold authority can be named with reasonable confidence. The information is sometimes obscured by indirection — maintainers operating under pseudonyms, organisational identities that conceal individual responsibility — but the operational picture can usually be assembled from public artefacts.
For sponsors and funders, identification depends on the disclosure regime. Where funders publish their support — as some foundations and corporate sponsors do — the picture is clear. Where funding is not publicly disclosed, the picture is harder to assemble but typically not impossible: payment flows leave traces, employment relationships become visible through public profiles, and the corporate filings of larger sponsors disclose grant-making activity. The exercise has been undertaken by independent researchers in several cases and the results, while imperfect, are informative.
For foundations and standards bodies, identification is direct: they are legal entities with public registrations, directors, governance documents, and operating procedures. The decisions they make in their formal capacity are publicly recorded.
For adoption-critical infrastructure, identification follows from concentration data: the largest validators, miners, and staking services are visible in on-chain data; the largest exchanges are visible in market data; the largest wallets are visible in download statistics and integration disclosures. The identification of the few parties whose adoption is decisive can be done with reasonable accuracy.
For narrative and classification authorities, identification is harder because the relevant authority is exercised through public communications and institutional reputation rather than formal decision rights. But the parties whose framings are widely adopted — the dominant developers, the foundations, the influential commentators — are visible to anyone observing the discourse.
Identification across all components is not a single research output; it is a cumulative analytical exercise. The exercise has been done partially for several major protocols, and the results have been more concentrated than the field’s standard descriptions would suggest. Where the exercise has not been done, the absence reflects a choice not to do it rather than a fundamental obstacle to doing it.
The relevant point for analysis is that the coalition exists, the coalition is identifiable in principle, and the costs of identification are bounded. The rhetorical move that treats the coalition as fundamentally invisible is a strategy for avoiding the analysis, not a property of the underlying system.
4. The coalition’s payoff structure
If rule change is produced by coalitions, then rule change occurs when the coalition’s coordinated action is in its members’ joint interest. This requires representing the coalition’s payoff structure explicitly, which is a step beyond what the descriptive analysis alone provides.
Consider a simplified representation. Let the coalition be denoted C, the proposed rule change r, and the coalition’s collective payoff from the change ΠC(r). The payoff has several components.
ΠC(r) = BC(r) + W(r) − K(r) − R(r) − I(r) − F(r)
where:-
BC(r) is the direct benefit to coalition members from the change — gains in influence, profitability, market position, or strategic advantage;
-
W(r) is the welfare benefit accruing to participants generally, to the extent it indirectly benefits coalition members through participant retention and ecosystem growth;
-
K(r) is the coordination cost of forming and sustaining the coalition through to activation;
-
R(r) is the legal, regulatory, professional, and reputational cost imposed on identifiable coalition members through accountability mechanisms;
-
I(r) is the capital at risk — the value of coalition members’ in-system holdings, infrastructure, and reputation that would be impaired if the change reduced confidence or fragmented the system;
-
F(r) is the cost of fragmentation, including the risk that a contentious change produces a fork that divides users, liquidity, and developer attention.
The coalition acts when ΠC(r) > 0. Several features of this expression deserve attention.
First, the welfare term W(r) is not the coalition’s primary motivation. It enters the calculation only insofar as it indirectly benefits coalition members. A change that produces large aggregate welfare gains but does not benefit coalition members directly may not be undertaken, because the coalition’s payoff is dominated by BC rather than W. A change that produces small aggregate welfare gains but large direct benefits to coalition members may be undertaken, because the coalition’s payoff is dominated by BC. The aggregate welfare effect of rule change is therefore not a reliable predictor of which changes occur; the distribution of the welfare effect across coalition members is what matters.
Second, the cost terms K, R, and F are the standard institutional constraints on coalition action. A coalition that is expensive to form, exposed to severe accountability, or likely to produce fragmentation is constrained. A coalition that is cheap to form, faces little accountability, and is unlikely to produce fragmentation is unconstrained. The institutional architecture of the protocol determines the magnitudes of these terms, which is why governance design matters.
Third, the capital-at-risk term I(r) is critical for identifying when coalitions discipline themselves. Where coalition members hold substantial in-system positions, they internalise some of the cost of changes that damage participant confidence. Where coalition members hold little in-system position — perhaps because they are paid in fiat by external sponsors and do not depend on the system’s continued value — they do not internalise that cost. The discipline of capital-at-risk operates only when the coalition has capital at risk.
Fourth, the inequality ΠC(r) > 0 is the coalition’s decision rule, not a normative statement. The fact that the coalition acts when its payoff is positive does not mean the action is welfare-improving for participants. The participants’ welfare enters the coalition’s calculation only through W(r), weighted by the coalition’s pass-through of welfare to its own returns. If that pass-through is small, participant welfare has little weight in the decision. The coalition acts on its own payoff, not on the participants’.
Several patterns of governance design follow from this representation. Where R(r) is small (low accountability), I(r) is small (low capital at risk), and K(r) is small (low coordination cost), coalition action is cheap and frequent. Where any of these is large, coalition action is expensive and rare. Designing protocols so that R, I, or K is large at the points of governance authority is one approach to constraining opportunism; designing so that the coalition has aligned incentives with participants — large I and large pass-through of W to BC — is another. Both approaches have been attempted, with mixed results.
5. Welfare-improving and opportunistic rule change
The coalition framework does not prejudge whether any particular rule change is good or bad. It identifies the parties producing the change and the structure of their incentives. Whether the change is welfare-improving or opportunistic depends on the specifics: whose welfare is affected, in what direction, by how much, and whether the coalition’s internal calculation aligns with participant welfare.
Welfare-improving changes do occur. Bugs are fixed, errors are corrected, capacity is increased, and unforeseen technical problems are addressed. These changes typically meet the coalition’s decision criterion because the welfare improvement passes through to the coalition’s own returns — through participant retention, fee revenue, infrastructure value, and continued relevance. When W is large and pass-through is high, BC approximates W, and the coalition’s interest aligns with the participants’.
Opportunistic changes also occur. A change that disproportionately benefits coalition members at the expense of other participants meets the coalition’s decision criterion when BC is large for coalition members specifically, regardless of what W is for participants generally. The change can occur even when W is negative — that is, even when the change reduces aggregate participant welfare — provided that the coalition’s direct benefit is large enough to compensate for any reduction in W‘s pass-through.
The distinction between the two cases is not always observable from the change itself. A change presented as a routine improvement may in fact concentrate benefits among coalition members; a change presented as controversial may in fact be aggregate-welfare-improving despite distributional consequences. The framing is a governance act, not a description of the change.
Several diagnostic questions help distinguish the cases. Who benefits from the change in measurable terms — fees, capacity, market position, strategic advantage? Are the beneficiaries concentrated in the coalition or distributed across participants? Does the change require the coalition to bear costs proportionate to its benefits, or does it externalise costs onto non-coalition participants? Does the change introduce or remove constraints on the coalition’s future authority? Each of these questions has answers that are observable in principle, even when the rhetoric around the change obscures them.
The diagnostic is not a verdict. Some changes that benefit coalition members are also welfare-improving, and the alignment is not by itself evidence of opportunism. Some changes that benefit coalition members disproportionately are nonetheless reasonable in context, and the disproportion is not by itself evidence of impropriety. The diagnostic is an organising framework for evaluating the change, not a substitute for the evaluation.
What the framework rules out is the conclusion that, because the change went through the official process, or because no formal objection was registered, the change is welfare-improving by virtue of having occurred. The official process is a description of how the coalition coordinated its decision; it is not an assessment of the decision’s welfare consequences. Treating the process as if it were the assessment is one of the more common errors in the discourse around protocol governance.
6. Activation as coalition mechanism
The coalition does not produce a rule change in a single moment. The change is produced through a sequence — proposal, discussion, drafting, code review, merge, release, signalling, adoption, activation — and different components of the coalition exercise authority at different stages. The activation mechanism specifically is a coalition mechanism: it is the procedure through which the coalition’s coordination crystallises into an effective rule change.
Different activation mechanisms allocate coordination costs differently across the coalition and across non-coalition participants.
Miner or validator signalling. A change activates when consensus operators with a defined supermajority of capacity signal support. This places the activation decision in the consensus operators, making them necessary coalition members for changes that use this mechanism. Where consensus capacity is concentrated, the supermajority can be achieved by coordination among a small number of large operators, which lowers K. Where consensus capacity is dispersed, the same threshold requires coordination among many operators, which raises K.
User-activated soft forks. A change activates when a defined set of users (often economic majority by some measure) runs the new client and refuses to recognise blocks that do not comply with the new rules. This places the activation decision in the user population, theoretically expanding the coalition. In practice, the “users” who matter for activation are the adoption-critical infrastructure parties — large exchanges, custodians, services — whose adoption is decisive for the economic majority test. The mechanism nominally expands the coalition; in practice it relocates authority from consensus operators to infrastructure providers.
Hard-coded activation heights. A change activates at a specified block height, with no signalling threshold required. This places the activation decision in the maintainers who include the activation logic in the release. Adoption is implicit: any operator running the release at the activation height implements the change, and any operator who does not is deprecated to a forking minority. This mechanism shifts authority sharply toward the maintainers and reduces the role of explicit consensus-operator coordination.
Token-weighted on-chain governance. A change activates when token-holders vote with sufficient weight in favour. This places the activation decision in the token-holders, with weight proportional to holdings. In practice, the weight tends to be concentrated, because token holdings are concentrated, with the additional concern that custodians often vote on behalf of beneficial owners with their own preferences in mind. The mechanism expands the formal coalition to all token-holders but concentrates effective authority in the largest holders or their custodians.
Emergency procedures. Some protocols have explicit emergency mechanisms for rapid changes in response to security incidents or other urgent conditions. These mechanisms typically vest authority in a small group — foundations, designated security teams, or maintainers — and rely on post-hoc legitimation rather than ex ante coalition coordination. The mechanisms reduce K by design, accepting concentration of authority in exchange for rapid response capacity. Where the emergency procedures are routinely used, the concentration becomes a feature of the system rather than an exception.
Each activation mechanism is itself a governance choice with consequences for who is in the coalition and at what cost. Reporting the activation mechanism alone does not capture the coalition; reporting the actors who are in the coalition for any given change requires identifying who is necessary to the activation in that specific case.
7. Coalition stability and capture
A coalition that has produced rule changes once tends to have the capacity to produce them again. The institutional infrastructure that supported the first change — the maintainer team, the foundation, the funded development effort, the relationships with adoption-critical infrastructure — persists, and is available for subsequent changes. This is not a defect; it is how operational governance works. The infrastructure has to exist, and a one-time coalition is not a coalition that can sustain a protocol.
The persistence of infrastructure has two consequences. First, the marginal cost of subsequent rule changes is lower than the cost of the first, because the coordination structure does not have to be rebuilt. Second, the parties operating the infrastructure accumulate authority over time, both because their familiarity with the process makes them more effective and because the alternatives to them gradually atrophy as participants come to rely on the existing structure.
The accumulation of authority is the precondition for capture. A coalition whose membership has stabilised and whose institutional position is secure faces fewer constraints on its decisions than a coalition that must demonstrate its legitimacy in each round. Capture, in this sense, is not necessarily corruption or self-dealing in the strong sense; it can be the routine consequence of any institutional arrangement persisting long enough to acquire its own interests.
The signs of capture are observable. Coalition members increasingly act in ways that benefit themselves at the expense of participants. The narrative around proposed changes becomes increasingly tilted toward the coalition’s framing. Critical commentary on the coalition’s decisions becomes harder to sustain because dissenters are excluded from the institutional resources that would amplify their views. Internal accountability mechanisms that were notionally present become formal rather than substantive. Each of these is a known dynamic in institutional governance and there is nothing about protocol governance that exempts it from the dynamic.
The countermeasures are also known. External accountability through legal or regulatory channels constrains capture by raising R(r). Term limits or rotation requirements on key positions raise the difficulty of stabilising a coalition. Transparency requirements — public decision records, open funding disclosures, observable conflict-of-interest mechanisms — raise the difficulty of obscuring opportunistic action. Structural separation of authority across multiple institutional centres prevents any single coalition from becoming adoption-critical for all changes. None of these is novel; each has been applied in other governance contexts with varying success. The application to protocol governance has been partial, and where it has been attempted, the durability of the application has been mixed.
The point for analysis is that coalition stability is a feature to be tracked, not assumed away. A protocol that has had the same effective coalition for several years has different governance properties than a protocol whose coalition has rotated through significant changes in membership. The first is more efficient at producing rule changes, but more exposed to capture; the second is less efficient but more constrained. Neither is unambiguously preferable, and the trade-off depends on the system’s other properties.
8. Standard objections
“There is no coalition; the process is open and anyone can participate.” Open participation is not the same as effective authority. Many parties contribute to discussions, file proposals, and write code; few have the authority to merge, release, fund, or activate. The framework distinguishes participation from authority and treats the latter as the relevant analytical object. Saying that anyone can submit a proposal does not address who decides which proposals advance.
“Identifying the coalition is reductive; governance is more complex than that.” The coalition framework is not a complete description of governance; it is a starting point. It identifies the parties whose action is necessary, and analysis can build on that foundation by examining the relationships among those parties, the constraints they operate under, and the patterns of their decisions. Avoiding the starting point does not produce a more sophisticated analysis; it produces no analysis. Complexity is a property of the answer, not a reason to refuse to ask the question.
“Coalition language is hostile and assumes bad faith.” The framework does not assume bad faith. It identifies who has authority and asks what that authority is used for. The use can be welfare-improving, opportunistic, or mixed; the analysis is empirical, not presumptive. The framework would apply equally to a coalition acting in unimpeachable good faith — it would identify the coalition, examine its decisions, and reach favourable conclusions if the decisions warranted them. Hostility is not a property of the framework; assessment is.
“Coalitions form and dissolve dynamically; treating them as stable misrepresents the process.” The framework does not require static coalitions. The coalition for a specific change is identified for that change. The composition can shift across changes, and the analysis tracks the shifts. What the framework rules out is the notion that, because composition shifts, no coalition exists for any given decision. Each decision is produced by some coalition; the dynamics across decisions are themselves an object of analysis.
“Many parties contribute meaningfully without being formally in the coalition.” True, and the framework accommodates this by distinguishing necessary from sufficient contributors. Many parties contribute to outcomes without being necessary for them. The coalition is the set of necessary contributors; the broader set of contributors is the participation network. The two are different objects and the framework respects the difference.
“Rule changes are sometimes blocked by parties outside the formal coalition; the framework cannot represent veto authority.” The framework can represent it. A party that can block a change has effective authority and is therefore a member of the coalition for that change in the negative direction — its non-participation prevents activation. The coalition definition is “parties whose coordinated action is required,” which includes parties whose non-action would prevent the change. The framework distinguishes proactive coalition members (whose action produces the change) from veto-holders (whose non-action permits it), and analysis can address both.
9. What this implies for governance reporting
Honest governance reporting for an operational protocol would include several elements that current standard practice typically omits.
For each significant rule change in the protocol’s history, an identification of the coalition that produced the change. Who held merge authority over the relevant repository? Who funded the development effort? Who exercised foundation-level authority over the proposal process? Whose adoption was decisive for activation? Which parties’ framing of the change shaped expectations? The identification produces a record of how rule change has actually been produced, distinct from the official process narrative.
For the protocol’s current state, an identification of the coalition that would be required to produce a change of typical scope. The same components — maintainers, sponsors, foundation, adoption-critical infrastructure — applied to current conditions. This identifies who currently has the authority that would be exercised in any future change.
An assessment of the coalition’s incentive structure. What capital does the coalition have at risk? What accountability does it operate under? What is the coordination cost of its action? What is the alignment between coalition incentives and participant welfare? The assessment is the basis for evaluating whether the coalition’s decisions are likely to be welfare-improving or opportunistic.
An identification of the changes that have been proposed and not adopted, with attention to who blocked them and why. The set of non-adopted proposals is informative about the coalition’s preferences in a way that the set of adopted changes alone is not. A coalition that consistently blocks changes that would constrain its own authority is a coalition exercising authority opportunistically; a coalition that adopts such changes is one operating under credible internal discipline.
An assessment of coalition stability. How long has the current effective coalition been in place? What changes in membership have occurred, and what triggered them? What would trigger a substantive change in the coalition’s composition under current conditions? The assessment indicates how exposed the protocol is to capture dynamics.
None of these is particularly difficult to produce for an analyst with access to public information and willingness to do the work. The reporting is rare not because it is impossible but because it is uncomfortable. The alternative — describing governance as a community process without identifying the coalition that produces decisions — is more flattering and less informative. Participants who must evaluate the system need the more informative version.
10. Closing
Protocols do not change their own rules. Coalitions of identifiable parties produce rule changes through coordinated action, and the parties whose action is necessary for any given change are the coalition for that change. Identifying the coalition is the first step in any honest analysis of governance, and most operational protocols have not had this work done with the rigour it deserves.
The components of typical coalitions — reference implementation maintainers, sponsors and funders, foundations and standards bodies, adoption-critical infrastructure, exchange and listing authorities, wallet and infrastructure software, narrative and classification authorities — are observable, identifiable, and analysable. The work of identifying them is bounded. The reasons it is not done more frequently are partly difficulty and partly choice; the difficulty is real but not insurmountable, and the choice is between a flattering account that obscures who decides and a more accurate account that names them.
The coalition’s incentive structure determines what kinds of rule changes occur. A coalition with high capital at risk, substantial accountability, and high coordination costs is a coalition constrained to act in ways that align with participant welfare. A coalition with low capital at risk, weak accountability, and low coordination costs is a coalition free to act on its own preferences regardless of participant welfare. The institutional architecture of the protocol determines which kind of coalition operates, and the architecture is a design choice, not an accident.
What follows for participants is that evaluation of a protocol requires identifying the coalition that governs it. The protocol’s nominal characteristics — the consensus mechanism, the validator count, the token distribution — do not capture this. The coalition is a separate object and requires separate analysis. Participants who skip the analysis are evaluating a system whose governance they have not examined.
What follows for designers is that governance architecture is governance design. Constraints on coalition action — accountability requirements, capital-at-risk requirements, coordination requirements — are not external impositions on a self-governing community; they are the mechanisms by which the coalition’s authority is constrained. Designers who treat the community as the governing body and refuse to identify the coalition are not designing constraint-free governance; they are designing governance whose constraints are unspecified and therefore difficult to enforce.
What follows for analysts is that governance description requires coalition identification. The vocabulary of “the community decided” is not a description of how decisions occur; it is a rhetorical move that obscures who decided and how. Replacing the move with the more accurate description does not require any sophisticated theoretical apparatus. It requires the willingness to ask, for any specific decision, who took the actions that made it effective, and to record the answer.
The basic claim of this essay can be stated as a single proposition. Rule change in operational protocols is the output of coalitions, and a coalition is the set of parties whose coordinated action is necessary for the change to occur. Anyone outside this set is observing the change, not producing it. The set is identifiable, the set’s incentives are analysable, and the set’s decisions are the operational meaning of governance. Anything else is a description of who is paying attention, not of who is deciding. Participants need to know the difference. The analytical infrastructure for distinguishing them exists. The remaining task is to use it.
References mentioned in passing: M. Olson, The Logic of Collective Action (1965); R. H. Coase, “The Nature of the Firm,” Economica (1937); D. C. North, Institutions, Institutional Change and Economic Performance (1990); O. E. Williamson, The Economic Institutions of Capitalism (1985); A. O. Hirschman, Exit, Voice, and Loyalty (1970).