Competitive Topologies and the Monopoly Problem

The Decay Function of Monopolies Without Exit

Hirschman’s insight was that firms and states recover from declining performance through the joint operation of exit and voice. Each mechanism relies in part on the other. A monopoly whose customers cannot leave is not disciplined by their complaints alone. Managers can ignore complaints when customers cannot credibly threaten to leave. Institutions then favor internal convenience, tolerate insider control, and retain outdated processes because they bear little visible cost for doing so.

Democratic institutions are, in their territorial dimension, monopolies of this kind. The well-documented symptoms are consistent with a monopoly whose customers cannot easily leave: declining trust (Pew Research Center, Americans’ Views of Government, 2024; Edelman Trust Barometer, 2024); chronic procurement failures; the U.S. Department of Defense’s inability to earn a clean audit opinion since its first full agency-wide audit in 2018, across successive reports covering trillions of dollars in reported assets (DoD Office of Inspector General, annual financial statement audits); legacy software running critical systems (Government Accountability Office, Federal Agencies Need to Address Aging Legacy Systems, GAO-16-468). They are consistent with what Hirschman’s model predicts once territorial exclusivity forecloses exit and the electoral cycle stretches voice across four-year intervals.

Polarization and the sense that contemporary politics has become negative-sum are documented by political-science work (McCarty, Poole, and Rosenthal, Polarized America, 2006; Mason, Uncivil Agreement, 2018; Levitsky and Ziblatt, How Democracies Die, 2018), and by concrete measures of share of global output. At market exchange rates, the G7’s share of world GDP has fallen from two-thirds in the early 1990s to under half today; at purchasing-power parity the shift is sharper, with the G7 share below a third and a BRICS bloc whose PPP output now exceeds it (World Bank World Development Indicators; IMF World Economic Outlook, October 2024). The shift raises the stakes of internal political conflict. The causal direction runs both ways. When relative fortunes tighten and exit is foreclosed, incumbents have fewer reasons to invest in substantive reform.

This is the condition Schmidt and Sorota correctly identify. Populations that cannot leave and whose voice cycles are long can, when institutions underperform, turn to whatever alternative promises effectiveness: authoritarian strongmen, technocratic programs, or, increasingly, algorithmic ones. Schmidt and Sorota propose better tools for voice. Voice alone cannot discipline a provider whose customers cannot leave. Credible exit would add an incentive to respond, so the two proposals address different parts of the same problem.

The Prior Literature on Jurisdictional Competition

The claim that jurisdictions can be disciplined by the mobility of their citizens has a long history. Charles Tiebout formalized it (Journal of Political Economy, 64:5, 1956) as a pure theory of local public goods: if households can sort across jurisdictions costlessly and with full information, and if the jurisdictions capture no externalities, then local governments will be disciplined to provide the bundles their residents want. Tiebout’s seven assumptions have been the centerpiece of a sixty-year debate. They are unrealistic for ordinary citizens. Truman Bewley (1981), in a general-equilibrium critique in Econometrica, showed that Tiebout equilibria may fail to exist or fail to be efficient, which has been the main reason economists have treated the Tiebout model as suggestive rather than operational for human beings.

Corporate law shows why the identity of the person choosing a jurisdiction matters. Roberta Romano (“Law as a Product: Some Pieces of the Incorporation Puzzle,” Journal of Law, Economics, and Organization, 1:2, 1985) argued that jurisdictional competition among U.S. states for corporate charters produced a race to the top: Delaware’s dominance reflects specialized courts, a developed body of precedent, and efficient rule-production. William Cary had earlier framed the opposite charge (“Federalism and Corporate Law: Reflections upon Delaware,” Yale Law Journal, 83, 1974): the same competition rewards states that cater to managers at the expense of shareholders and outsiders, producing a race to the bottom on dimensions that the choosing party does not fully internalize. Lucian Bebchuk (“Federalism and the Corporation,” Harvard Law Review, 105, 1992) and later work with Ferrell refined the critique: where the gains from chartering accrue to insiders while costs are externalized onto outsiders, competition systematically misaligns.

Frey and Eichenberger’s FOCJ proposal (1996) separated sovereignty into functional, overlapping jurisdictions so that providers could compete over individual public goods. Elinor Ostrom’s Governing the Commons (Cambridge, 1990) showed, from a different direction, that polycentric orders can sustain common-pool resources without collapsing into either the tragedy of the commons or the rigidity of centralized planning, provided certain design principles are met.

The literature makes the choice more specific than “competition good” or “competition bad.” Jurisdictional competition races to the top where the chooser internalizes costs and benefits (dispute resolution speed, clarity of chartering rules, quality of court opinions), and races to the bottom on dimensions where the chooser does not (third-party externalities, distributional consequences, systemic risk). Any proposal for expanding exit must therefore say which dimensions it expects to improve and which it requires floor standards to protect.

The examples used to support or oppose jurisdictional competition are themselves contested. Delaware’s primacy has generated a large body of work on both sides. China’s Special Economic Zones are cited sometimes as a triumph of regional experimentation (Naughton, The Chinese Economy, 2007) and sometimes as evidence that the most dynamic growth occurred outside the zoned structures, driven by township-and-village enterprises and rural entrepreneurship the center did not plan (Huang, Capitalism with Chinese Characteristics, 2008). The SEZ case also sits uneasily with the competition thesis in a more basic sense: economic decentralization was permitted within a framework of unchallenged political centralization. The example shows that local experimentation can generate growth. It does not show that exit disciplines a central monopolist who can, at any moment, withdraw the experiment.

Digital infrastructure changes some of the constraints in Tiebout’s model. A software-operated entity is a legal entity whose software executes its compliance behavior. Such entities can approximate assumptions that were implausible for human households: low physical relocation costs, accessible information about jurisdictional rules, and freedom from social ties to incumbent regimes. Human citizens still face those costs. A mobile entity can be built, but its mobility still poses the race-to-the-bottom problem. The question is which dimensions improve with mobility and which require shared floor standards.

Network Topology as a Frame

The distinction between voice and exit gives the topology two parts. One graph records who can address whom within a governance system (information and authority flow). The other records which services a citizen can choose, drop, or combine without relocating (substitutability). The first determines how information moves. The second determines whether a dissatisfied citizen can choose another provider.

Modern democracies operate with an authority graph that is largely hub-and-spoke and with a substitutability graph that is largely trivial. National governments sit at a hub; citizens sit at the leaves; decisions flow downward through bureaucracy; voice flows upward through intermittent elections. For most public services, citizens have no alternative provider for schooling, insurance, or dispute resolution. This topology was adequate when coordination across distance was costly and when citizens’ public-service choices were in practice tied to a single geography. Deliberation tools improve communication within this structure while leaving its monopoly over service provision in place.

FOCJ-style proposals (Frey and Eichenberger, 1996; Vaubel’s later extensions) propose an alternative: overlapping jurisdictions whose boundaries are functional rather than territorial, so that citizens can be members of one jurisdiction for schooling, another for dispute resolution, another for environmental protection, each of which competes to retain its domain. This is the academic ancestor of the phrase “higher-dimensional governance,” and we use the term in that sense here. The substantive test for any such proposal is whether the functional jurisdictions can be cleanly separated, so that exit on one axis does not collapse provision on another, and whether floor standards can be held across them.

vTaiwan is a Pol.is-based deliberation platform developed with Audrey Tang and collaborators, designed to surface cross-cleavage consensus on specific policy questions within the Taiwanese state (see Small et al., “Polis: Scaling Deliberation by Mapping High-Dimensional Opinion Spaces,” Recerca, 2021). It improves deliberation within the state, as Schmidt and Sorota intend. Its design leaves open our question: whether better deliberation gives a monopolistic provider sufficient reason to respond.

Exit, Voice, and the Lazy-Monopoly Warning

Hirschman’s 1970 argument is specific about the relation between exit and voice. Exit and voice are not pure complements; they are partial substitutes mediated by loyalty. When exit becomes easy, the first to leave are precisely the members whose complaints carry diagnostic information. Their departure can deteriorate the voice pool, the “lazy monopoly” outcome, leaving the institution with the apathetic and the loyal, and accelerating its decline. This is the strongest objection to our argument.

The lazy-monopoly outcome requires two conditions in the democratic case. First, the exit-capable population would have to be disproportionately drawn from the voice-active. Second, the voice channel would have to depend on the presence of that population to produce signal. Both are plausible, and digital infrastructure does not remove them. Our argument therefore depends on a judgment about current conditions. We judge that current voice channels are already saturated with well-designed mechanisms that institutions ignore, so further improvements have low marginal value. Credible exit has high marginal value because its absence lets institutions disregard those channels. If that judgment about their relative responsiveness is wrong, Hirschman’s warning applies.

The distribution of exit costs creates a further problem. The early movers in any newly available exit channel are typically the capital-rich and voice-capable: those with the resources, legal advice, and optionality to leave. Their departure can leave the remaining population bearing the jurisdiction’s fixed costs. It is the basis for the floor-standards requirement developed below. A defense of expanded exit must specify which externalities a departing entity carries with it (carbon, financial-systemic risk, money-laundering exposure, distributive burden) and which public-good contributions it must continue to make even once it has exited administratively. Without those obligations, expanded exit permits departing entities to pass their costs to those who remain.

For the least mobile citizens, the state’s ability to retain them regardless of performance weakens its incentive to respond. Even a perfectly deliberative legislature will ignore preferences it is under no structural pressure to honor. Mechanisms that reduce the cost of exit along specific axes, among them business registration, dispute resolution, capital formation, professional licensing, and research funding, put structural pressure on the territorial monopolist without requiring any citizen to emigrate. That pressure operates continuously between elections, creating competition for services that states provide. Jurisdictions that provide those competencies badly would begin to lose the activity that fills their treasuries, and legislatures could assess their bureaucracies against that loss of activity. This is a Tiebout-style argument with two changes: the mobile unit is a software-operated legal entity rather than a human household, and the externality-heavy axes are explicitly excluded from the competition.

Infrastructure and Institutional Choice

Representative democracy emerged under binding logistical constraints. Gathering a million people, counting a million ballots in a day, or faithfully transmitting preferences across a continent was infeasible. Those constraints shaped much of what we now call “democratic deliberation”. As the constraints soften, the institutions that were shaped by them should be subject to scrutiny rather than preservation for its own sake. Schmidt and Sorota’s argument and ours share this premise.

Two families of voting mechanisms illustrate what lower coordination costs make possible. The first, delegated or “liquid” democracy, lets citizens delegate voting power issue by issue and revoke the delegation at will (Ford, “Delegative Democracy,” 2002; Green-Armytage, 2015). The empirical record is mixed. LiquidFeedback inside the German Pirate Party and related experiments consolidated voting weight into a small number of super-delegates, producing a new concentration pattern rather than dissolving the old one (Kling et al., “Voting Behaviour and Power in Online Democracy,” Policy & Internet, 2015). That experience leaves open design problems in sybil-resistance, delegate accountability, and default settings.

The second family comprises preference-intensity aggregation rules, most prominently quadratic voting (Lalley and Weyl, “Quadratic Voting: How Mechanism Design Can Radicalize Democracy,” AEA Papers and Proceedings, 108, 2018). Quadratic voting has been piloted, most notably by the Colorado House Democratic caucus for internal budget prioritization, and in field experiments by the RxC community. These pilots provide evidence, but leave the mechanism’s broader suitability unresolved. Taiwan’s civic-tech stack is often cited in this connection. Its mechanism is Pol.is, which clusters opinions into a two-dimensional map. Pol.is and quadratic voting are distinct mechanisms. Both face unresolved questions about collusion resistance, quasi-linear utility assumptions, and base-rate effects in small populations.

Their use at broader scale still requires research.

Parallel Experimentation and Its Limits

The scale of experimentation also affects institutional choice. A national policy exposes the entire population to the cost of failure, giving incumbents reason to favor incremental changes. Decentralized experiments can contain failure while allowing successful variants to spread. This is the core attraction of Ostrom’s polycentric analysis (Governing the Commons, Cambridge, 1990) and of federalist “laboratories of democracy” arguments going back to Brandeis. Both rest on the premise that the knowledge required to govern well is distributed, local, and revised by experience.

The American federal system exhibits elements of this logic. States differ on healthcare financing, cannabis policy, occupational licensing, and criminal justice. Observable outcomes allow policies to spread, though slowly. The literature on policy diffusion (Walker 1969; Berry and Berry 1990; Shipan and Volden 2012) documents both the mechanism and its frictions: imitation is common but slow, and is often mediated by party networks rather than by impartial performance comparison. Interstate competition has produced clear gains on some axes (dispute resolution in Delaware courts, unemployment-insurance design) and clear losses on others (corporate-tax base erosion, regulatory arbitrage in consumer finance). The pattern is the one the prior section anticipated: gains where the chooser internalizes costs and benefits, losses where the chooser does not.

Huang (2008) and Rozelle and Hell (Invisible China, 2020) argue that much of China’s growth happened outside the zones and that SEZ policy concentrated investment at the cost of regional inequality. Even to the extent the zones functioned as competitive laboratories, they did so inside a political system in which the central authority could and did override their terms at will. The central authority’s power to end an experiment leaves the underlying monopoly intact.

The practical change concerns comparison and movement costs, rather than general “software-speed governance.” Research in the legal-informatics literature, notably Merigoux, Chataing, and Protzenko’s Catala (Proceedings of the ACM on Programming Languages, ICFP 2021) and the L4 project at Singapore’s Centre for Computational Law, is developing ways to write rules in a form that is both human-readable and machine-executable. When that becomes feasible, three frictions fall together: the cost of comparing one jurisdiction’s rules against another’s, the cost of operating across jurisdictions, and the cost of observing outcomes. Calling this “software speed” obscures the time that people and institutions need to implement a policy and the years its effects can take to emerge. What changes is the cost of comparison and of movement, and those are the costs that discipline a territorial monopolist.

Policy revision also differs from software deployment in what it can undo. Lessig’s Code and Other Laws of Cyberspace (1999; revised Code v2, 2006) and Winner’s The Whale and the Reactor (1986) identified differences between code-as-law and law-as-law that subsequent work has developed: reversibility, legitimacy, consent, exception-handling, and the treatment of lives already altered by a rule. A software deploy can be rolled back; a policy whose consequences have been lived through cannot be. In domains where these differences bind hard, among them fundamental rights, criminal justice, and civil liberties, no reasonable reading of digital infrastructure authorizes faster iteration as a norm. In narrower domains such as business registration, tax compliance, dispute resolution, and professional licensing, the disanalogy softens and rapid adjustment is compatible with the substantive stakes. In these domains, rendering rules in software can reduce the cost of comparing and exiting jurisdictions.

Programmable Organizations and the Nominal-Effective Gap

One strand of Schmidt and Sorota’s worry is that automated governance, by displacing discretionary human judgment, leaves nobody to hold accountable. An organization can implement its rules in code or use on-chain votes while concentrating effective control in a few people. Research on the difference between nominal and effective decentralization requires our proposal to account for who exercises that control.

Angela Walch’s “Deconstructing ‘Decentralization’: Exploring the Core Claim of Crypto Systems” (in Brummer, ed., Cryptoassets: Legal, Regulatory, and Monetary Perspectives, Oxford, 2019) catalogues the gap between the rhetoric of decentralization and its operational reality: core developer concentration, validator concentration, governance-token concentration, dependency on oracles and off-chain inputs, and the reliance of nominally trustless systems on trusted parties at their edges. De Filippi and Wright, Blockchain and the Law: The Rule of Code (Harvard, 2018), extend the point into legal theory: “code is law” replaces one opacity with another, shifting the interpretive burden from the lawyer to the developer and leaving most non-technical users with less capacity to scrutinize the rules that bind them, not more. Empirical studies of governance-token distribution (Barbereau et al., “DeFi, Not So Decentralized: The Measured Distribution of Voting Rights,” HICSS, 2023) typically find majority voting power concentrated in a handful of addresses. The historical cases include TheDAO (2016), the Parity wallet freeze, the Tornado Cash OFAC sanction, the Uniswap fee-switch debate. These cases are evidence about how decentralized organizations govern.

This evidence separates two claims about programmable organizations. The first is that rules written in code, where appropriate, can be more legible, more comparable, and more portable than the same rules written in unstructured natural language. Catala, L4, and the broader literature on formal compliance investigate this testable improvement. The second claim is stronger: that on-chain organizations are, as a consequence of being on-chain, governed by their stated rules rather than by whoever controls their privileged operational levers. Being on-chain provides no general guarantee of that kind. Any proposal that relies on programmable organizations inherits the burden of specifying who operates the substrate, how upgrade authority is distributed, how key-loss and capture are handled, and what accountability mechanism exists when effective control departs from nominal control. That burden applies to the proposal sketched in this essay.

A few jurisdictions have experimented with legal wrappers for such organizations. Wyoming’s DAO LLC statute (2021, amended 2022) and the Republic of the Marshall Islands’ DAO Act (2022) recognize these entities for limited purposes; both are best read as regulatory-arbitrage offerings for crypto-adjacent activity rather than as experiments in sovereign governance. Estonia’s e-residency program is narrow in scope: it enables non-residents to access Estonian digital services, register businesses, and use digital signatures under Estonian law, without conferring voting rights, physical residency, or EU citizenship (Kotka, Alvarez del Castillo, and Korjus, “Estonian e-Residency: Redefining the Nation-State in the Digital Era,” 2015). These programs establish the technical and legal feasibility of digital participation within their limited purposes. They leave digital sovereignty unresolved.

Prediction Markets and Their Actual Record

Prediction markets offer a different input to governance: forecasts from dispersed participants. Digital infrastructure lowers the cost of operating them. Wolfers and Zitzewitz’s survey (“Prediction Markets,” Journal of Economic Perspectives, 18:2, 2004) and Arrow et al.’s Science piece “The Promise of Prediction Markets” (320, 2008) report that well-designed markets aggregate information from dispersed participants with calibration that typically meets or exceeds expert forecasts, especially on questions where base rates are stable and participation is broad. They also document the regime in which this claim has most force (high liquidity, well-specified questions, absence of thin-market manipulation) and the regime in which it has least force (sparse markets, whale concentration, and questions entangled with the market participants’ own behavior).

Polymarket is the most visible current venue and a useful case. It priced the 2024 U.S. presidential outcome close to the ultimate result ahead of election day, and it misfired on several smaller contests in the same cycle where volume was thinner. The pattern, informative where liquid and unreliable where thin, is what theory predicts. Prediction markets, where designed and populated well, carry information that survives comparison with expert forecasts. Their reliability depends on the narrow conditions described above.

Using those forecasts in governance requires a separate choice of what policy should achieve. Robin Hanson’s “futarchy” proposal (“Shall We Vote on Values But Bet on Beliefs?” Journal of Political Philosophy, 21:2, 2013) argues that citizens vote on the metric by which policies are judged and markets forecast which policies will best advance that metric. The proposal has known objections, among them goodharting of the chosen metric, manipulation by well-capitalized actors, and the distributional question of who can participate, and it has not been implemented at any governance scale that would settle them. It illustrates a useful separation: decision mechanisms that separate the normative question (what do we value?) from the empirical one (which policy advances it?), each handled by an institution suited to it. Manipulation and unequal participation leave the viability of such designs open.

Verification Without Surveillance

Schmidt and Sorota are right that digital-first democratic participation requires a verification substrate, and right that any such substrate, done badly, becomes surveillance infrastructure that outlasts the legitimacy that authorized it. Zero-knowledge proofs and other privacy-preserving verification methods address parts of this problem. Their cryptographic and institutional components have different levels of maturity.

The underlying cryptographic primitives are well-established. Zk-SNARKs and zk-STARKs now run in production for rollup systems processing significant transaction volume; the cryptography is no longer speculative. What has not been solved, and what Buterin himself treats as unresolved (“What do I think about biometric proof of personhood?”, 2023), is the identity layer on which such proofs depend. Proof-of-personhood systems, including Worldcoin, Sismo, Anon Aadhaar, and the BrightID family, each confront the same set of tradeoffs: sybil resistance versus biometric data concentration; key-loss catastrophe versus custodial recovery; state endorsement versus independence; regulatory fit versus technical purity. None of these tradeoffs is resolved in a way that would permit the substrate Schmidt and Sorota describe to be built today at population scale. The question “who is a legitimate participant?” is upstream of every voting or petitioning mechanism, and it is not answered by the availability of zero-knowledge proofs downstream of it.

Bounded institutional settings offer firmer evidence. For member verification inside a chartered organization, eligibility for a specific benefit, or accreditation for a licensed role, the mathematics and institutional arrangements are now adequate to verify a property without disclosing the underlying credential. Extending it to national democratic participation requires institutional design as well as cryptography. Schmidt and Sorota’s verification proposal and our exit-rights proposal both depend on resolving that problem.

Functional Jurisdictions and Their Limits

The political theory of the modern state holds that governance is geographically contiguous and jurisdictionally comprehensive: a single authority controls a defined territory and regulates all activities within it, and citizenship is inherited or resident rather than chosen. Frey and Eichenberger’s FOCJ proposal was, in effect, a re-examination of that premise. They argued that some public goods are naturally bounded by function rather than by territory (school districts, water districts, and emergency-service districts already display this structure at sub-national scale), and that their governance could be organized around the functional constituency rather than the territorial one. Exit, in such a scheme, is exit from a functional jurisdiction, not from a country.

The substantive question it raises is whether such jurisdictions, which already exist in narrow forms, can be extended without running into four failure modes that the FOCJ literature itself has documented. First, cherry-picking: functional jurisdictions that attract low-cost-to-serve constituents while shedding high-cost-to-serve ones produce distributional pathologies identical to those the race-to-the-bottom critique identified in corporate chartering (Bebchuk 1992). Second, coordination: some public goods cannot be disaggregated without collapsing the public good itself; national defense and macroeconomic policy are the clearest cases, but the boundary is broader than it first appears. Third, legitimacy: a citizen who has opted into multiple functional jurisdictions and out of others is subject to rules she did not consent to only in the domains where she has not exited, and the coherence of a political community under such conditions is the topic of a real debate (Pettit’s Republicanism, 1997, and Shapiro’s Politics Against Domination, 2016, bear on this directly). Fourth, enforcement: functional jurisdictions still need coercive power to enforce their rules, and in the absence of a general territorial state that power either concentrates in a secondary monopoly or disappears.

A functional-jurisdictional order must therefore specify which domains are plausibly disaggregable, which require shared floor standards across jurisdictions, and which are irreducibly territorial. Carbon regulation, financial-systemic stability, anti-money-laundering coordination, and civil-rights floors are the clearest candidates for the last category, with externalities too diffuse or stakes too high to leave to local exit. Opening a domain to competition requires those conditions to be specified. They remain unresolved by a general appeal to “higher dimensions.”

Within those limits, digital infrastructure offers a reduction in the transaction cost of operating across functional jurisdictions, a reduction in the information cost of comparing them, and a reduction in the switching cost of moving between them. In domains where the cherry-picking, coordination, legitimacy, and enforcement failures do not dominate, including business registration, many forms of civil dispute resolution, certain kinds of professional credentialing, and some research-funding structures, these reductions are meaningful. In domains where those failures do dominate, no amount of reduced friction cures them.

Bundling and the Design of Choice

One feature of contemporary democratic systems is that they bundle preferences at coarse granularity. A vote for a party is a vote for a platform, and the platform will cover fiscal policy, social policy, foreign policy, environmental regulation, and criminal justice in a single package that few individual voters agree with in all its parts. Arrow’s impossibility theorem and the subsequent social-choice literature (Sen, Collective Choice and Social Welfare, 1970) tell us that no aggregation rule recovers a coherent social preference from diverse individual preferences without some loss; the bundling pattern of party politics is one particular way to handle that loss, and its costs are well-studied.

The functional approach also permits unbundling in principle: citizens can, in a technical sense, express preferences on multiple dimensions independently. Several mechanisms have been proposed for this, including direct democracy on specific questions, liquid democracy in Ford’s sense (2002) and Green-Armytage’s refinement (2015), and quadratic voting as Lalley and Weyl formalize it (2018). Pilots exist: LiquidFeedback inside the German Pirate Party; quadratic-voting experiments in the Colorado House Democratic caucus; Pol.is clustering in Taiwan’s civic-tech stack, which we noted earlier is a different mechanism from quadratic voting and should not be conflated with it. The pilots show how independent choices can work.

They also reveal recurring failure modes. The Pirate Party’s LiquidFeedback experience produced strong concentration of voting weight in a small number of “super-delegates” (Kling et al., 2015), reproducing in a new form the concentration pattern the experiment was meant to dissolve. Quadratic-voting pilots have been small enough and short enough that the collusion-resistance, quasi-linear-utility, and bribery-vulnerability concerns discussed in the theoretical literature (Weyl, “The Robustness of Quadratic Voting,” Public Choice, 172, 2017) have not been fully stressed. Alternative aggregation rules show promise in these settings, but their evidence does not yet support use in national elections. They may instead complement bundled partisan voting where bundling costs are high and collusion risks are manageable, including local budgeting, intra-party priority-setting, and participatory allocation of specific funds, where the pilots have most evidence.

The reduction of polarization is often invoked as a benefit of unbundling. The evidence on what causes polarization, surveyed by McCarty, Poole, and Rosenthal (2006), Mason (2018), and the subsequent literature on sorting and negative partisanship, does not support the simple claim that better voting mechanisms would resolve it. Polarization is partly identity, partly elite cueing, partly media structure, and partly sorting. Aggregation rules are a smaller lever than the polarization problem requires.

Where Faster Decisions Help

The most common version of the argument for digital-first governance rests on the claim that institutions must adapt faster to keep pace with technological and social change. Many regulatory frameworks were built for stable industrial economies and handle distributed digital services, gig labor, and cross-border data flows poorly. Those observable failures justify reform, but do not establish that governance should generally run at software speed.

Stability is itself a public good. The legal certainty that induces investment and long-horizon decisions depends on rules that change at an observable and predictable pace. Rights protections matter in part because they cannot be altered on the monthly cadence of a software deploy. James Scott’s Seeing Like a State (1998) and Andy Stirling’s work on directionality in innovation argue that rapid feedback loops can produce oscillation, overshooting, and manipulation by the feedback signal instead of convergence on good outcomes. Markets can race to bad equilibria at the same speed they race to good ones. Reform remains necessary. The right argument distinguishes domains where cycle time is the binding constraint from domains where the binding constraint is legitimacy, rights-protection, or shared expectation of stability.

Administrative procedures provide the stronger case for speed. Registration, licensing, compliance filings, dispute-resolution for specific high-frequency classes of dispute, and the administrative architecture of tax and subsidy delivery are domains in which users experience cycle time as a binding constraint, rights-protection stakes are moderate, and the consequences of iteration are reversible. These are also the domains in which machine-readable rule systems of the Catala and L4 type are most tractable. The literature supports faster administration in these domains. Our claim is limited to them. We do not make the general prescription that “governance should operate at software speed,” beyond that scope.

The Claim and Its Open Questions

The mechanisms considered here return us to Schmidt and Sorota’s prerequisites: trustworthy identity without surveillance, checkable content provenance, and deliberation beyond small assemblies. Both their proposal and ours depend on that infrastructure.

Our disagreement is with the implicit claim that voice-augmentation will be sufficient. Hirschman’s argument is that institutions without credible exit decay regardless of the quality of their voice mechanisms. The contemporary pattern of declining institutional trust, chronic administrative failure, and the turn toward authoritarian or technocratic alternatives is consistent with the monopoly-without-exit decay pattern Hirschman described. We hold, conditionally and within specific domains, that digital infrastructure lowers the cost of exit in ways that can restore an accountability mechanism voice alone often cannot provide.

The connection between that infrastructure and effective competition rests on four claims. Tiebout’s mobile-citizen model was implausible for human households because the underlying frictions were high; those frictions are substantially lower for software-operated legal entities whose compliance behavior is executed by software. Where rules can be written in machine-readable form, the cost of comparing jurisdictions and moving between them falls. Where functional jurisdictions can be cleanly disaggregated, competition along one axis is compatible with coordination along another. Where programmable organizations are operated with attention to the nominal-effective decentralization gap, they can provide substrate for pluralistic governance rather than a new center of concentration.

Four open problems remain. The race-to-the-bottom critique of Cary and Bebchuk applies to any regime of expanded exit and is not answered by the fact that exit is cheaper; it requires floor standards on externality-heavy dimensions, among them carbon, financial-systemic risk, anti-money-laundering, and civil-rights floors, and the specification of those standards is an open policy problem rather than a solved one. Hirschman’s lazy-monopoly warning remains: if exit selectively pulls the voice-active out of the voice pool, the institutions they leave decay faster, and the selection effect remains unresolved. The infrastructure that makes exit cheap is itself a point of potential concentration; Schmidt and Sorota’s concentration worry applies in full force, and the argument survives only if the substrate is genuinely pluralistic. Finally, whether voluntary association across functional domains sustains the shared political identity required by republican freedom in Pettit’s sense needs a fuller treatment of democratic theory.

Voice and exit need one another. Institutions can ignore voice when departure is impossible, while exit without effective voice can accelerate the decline Hirschman described. Democratic institutions currently have too little of either. Schmidt and Sorota seek infrastructure that makes both possible. We add that, within the domains and conditions set out here, citizens and entities need a credible alternative provider if better deliberation is to hold a territorial monopoly to account.