Economic Mechanism — From Productive Surplus to Distributed Income

A productive surplus is not the same thing as a distributed income. The history of the last two centuries is full of moments at which output rose sharply and the median person did not, in the short run, share in the rise. The question this essay tries to be precise about is the engineering question: by what mechanism does an autonomously-produced surplus become an income that reaches a person, and what is the design space of credible answers.

Why “mechanism” is the load-bearing word

When advocates and critics of post-scarcity arguments speak past each other, the disagreement is almost always about a confusion of two distinct questions. The first is the production question: how much value does the economy produce, and at what marginal cost. The second is the allocation question: by what protocol does the value reach particular persons. These are mathematically distinct. They have different evidence bases, different policy levers, and different failure modes. A century of debate about whether automation will or will not deliver shared prosperity has been systematically muddled by treating them as the same question. They are not.

The clearest articulation of why allocation deserves to be treated as a problem in its own right is Friedrich Hayek’s 1945 essay “The Use of Knowledge in Society.”¹ Hayek’s argument was, in form, an argument about mechanism. He observed that the relevant knowledge for allocation — which farmer’s field has been parched, which dockmaster’s crane is unreliable in cold weather, which household’s income has just dropped — exists only in the heads of millions of dispersed actors and cannot be collected centrally fast enough to be used. The price system, he argued, is the only known protocol that aggregates this dispersed knowledge into a single, locally-actionable signal. Whatever one’s view on the broader Hayekian programme, the framing was right: allocation is a mechanism question.

That framing was sharpened, technically, by the mechanism-design tradition that began with Vickrey’s 1961 paper on auctions, was extended by Clarke (1971) and Groves (1973) into the family of pivotal mechanisms that bear their names, and matured through the algorithmic-game-theory literature of the 2000s.²³⁴ What Vickrey-Clarke-Groves added, beyond Hayek, was a constructive proof that there exist allocation protocols which are simultaneously incentive-compatible (telling the truth is a dominant strategy), efficient (the chosen allocation maximises total welfare under the reported preferences), and individually rational (no participant is worse off for participating). The price system is one such protocol under restrictive assumptions; VCG-family mechanisms generalise the result to settings where prices alone underspecify. The mechanism-design literature is, in that sense, the formal vocabulary in which the allocation question is now discussed.

The technical substrate has continued to deepen. Roughgarden’s Twenty Lectures on Algorithmic Game Theory (2016) and Nisan, Roughgarden, Tardos, and Vazirani’s Algorithmic Game Theory (2007) lay out the computational complexity of mechanism design at scale: many of the mechanisms that are clean in theory become intractable when the action space is combinatorial, when participants reason strategically about each other’s reasoning, and when the allocator itself faces computational constraints.⁵⁶ These are not abstract concerns. The protocols by which a planetary economy might route a productive surplus to billions of recipients are, to a first approximation, mechanism-design problems at exactly this scale. The Economic Orchestration research pillar treats mechanism design at scale as one of its central technical problems, and the systems-level treatment describes how the corresponding deployment substrate is structured.

The point we want to be precise about, before we enter the taxonomy, is this: the abundance debate has historically conflated “is there enough” with “does it reach people,” and the conflation is corrosive in both directions. People who believe the production problem is solved sometimes assume the allocation problem will solve itself, which it will not. People who believe the allocation problem is too hard sometimes assume the production problem must therefore be unsolved, which is a non-sequitur. The mechanism question is the load-bearing one, because once production exceeds need at the level of a society — as it already does, on most aggregates, at the level of the global economy — the binding constraint is no longer extractive. It is allocative. And the design of allocation protocols is a domain in which there is, surprisingly, more to say than the public conversation has yet absorbed. The essays in the /abundance section, including this one, are an attempt to say it.

A taxonomy of distribution mechanisms

What follows is a six-part taxonomy of mechanisms by which a productive surplus can be turned into a distributed income. Each has been tried, in some form, at some scale. Each has a literature, a set of operational virtues, a set of failure modes, and a relationship to the others that is more often complementary than substitutive. We try to be even-handed about which is “best,” because we think the evidence is that the right design composes elements from several.

Sovereign wealth funds and citizen dividends

The Alaska Permanent Fund, established in 1976, is the longest-running real-world experiment in routing a non-renewable productive surplus to citizens through a sovereign-wealth-style intermediary. Funded initially from oil-lease revenues on Alaska’s North Slope, the Fund had grown to approximately $80 billion in assets under management as of 2024, and pays out an annual Permanent Fund Dividend (PFD) to every Alaska resident who meets a residency test. Recent dividend amounts include $1,312 in 2023 and $1,702 in 2024.⁷ The dividend is paid in October. It goes to roughly 600,000 to 650,000 residents annually. Children receive the same per-capita amount as adults.

The literature on Alaska as a near-universal-basic-income natural experiment is substantial. Goldsmith’s The Alaska Permanent Fund Dividend: A Case Study in Implementation of a Basic Income Guarantee (2010) and his subsequent 2012 paper lay out the fund’s mechanics, distributional effects, and political resilience.⁸ The edited volume Exporting the Alaska Model (Widerquist & Howard 2012) collects analyses of whether the design generalises — to other resource-rich jurisdictions, to non-resource economies, to a federal U.S. context.⁹ The empirical findings on Alaska are notable for what they did not find. Jones and Marinescu’s 2022 paper in AEJ: Economic Policy found no detectable effect of the PFD on aggregate Alaskan employment.¹⁰ The dividend has been politically resilient across decades and across changes in state administration, including under fiscal stress.

Norway’s Government Pension Fund Global is the contrast case. With approximately $1.7 trillion in assets under management as of 2024, it is the largest sovereign wealth fund in the world by a substantial margin. It is also, deliberately, not a citizen-dividend fund. Its returns are routed into the Norwegian state budget, which then provisions services — healthcare, education, pensions — to citizens. The two funds are structurally similar in their resource-rent funding model and structurally different in their distribution mechanism: Alaska pays cash to households; Norway pays for services. The two designs imply different things about state capacity, recipient discretion, and political dynamics, and we return to that contrast in the section on universal basic services below.

The reason to put sovereign-wealth dividend funds first in the taxonomy is that they are the most-tested, lowest-novelty option for routing rent from a productive surplus to citizens. The institutional machinery — the fund, the trustees, the indexing rules, the residency-verification system, the annual dividend — has been operational for nearly fifty years in Alaska and longer in Norway. The mechanism does not require novel technology, novel legal forms, or novel theory. It requires a credible source of recurring rent, a credible custodian, and a credible distribution rule. Each of those is hard. None of them is unprecedented. The precedents essay treats Alaska, Norway, and several smaller jurisdictional experiments at greater length, including the Iranian household-payments programme launched in 2010 and Macao’s Wealth Partaking Scheme. For the purposes of this taxonomy, the relevant claim is that “sovereign wealth fund plus citizen dividend” is the closest thing the world has to a tested mechanism for translating a recurring resource rent into a recurring per-person income, and that the design generalises in principle to rents from other classes of asset, including productive ones.

Direct cash transfers (UBI/UCT) at scale

The second mechanism is the unconditional cash transfer, paid directly to individuals on a recurring basis. Unlike the sovereign-wealth model, the transfer is not necessarily tied to a particular asset; it can be funded from general revenue, from a dedicated tax, or from a fund. The defining feature is that the recipient receives money rather than a service, and that the transfer is unconditional in the sense that no behavioural test (work, school enrolment, training participation) gates eligibility. The empirical literature on what happens when one runs this experiment has, over the past decade, become unusually rich.

GiveDirectly’s twelve-year randomised controlled trial in rural Kenya is, on most accounts, the largest and most carefully studied programme of its kind. Haushofer and Shapiro’s 2016 paper in the Quarterly Journal of Economics reported on the short-run effects of one-time transfers averaging roughly $700 per household, finding substantial effects on consumption, asset accumulation, food security, and psychological wellbeing, with limited evidence of expenditure on what economists call “temptation goods.”¹¹ The follow-up paper by Egger, Haushofer, Miguel, Niehaus, and Walker, “General Equilibrium Effects of Cash Transfers” (2022, Econometrica), is the more important one for the abundance argument.¹² By randomising at the village level rather than the household level and tracking spillovers across geographies, the authors estimated a fiscal multiplier of approximately 2.5 in recipient communities, with limited and category-specific local price effects. The paper is the strongest empirical answer to the standard inflation objection: the dominant effect of recurring transfers in a partially-tradeable local economy was an increase in real economic activity, not an increase in nominal prices.

The Stockton Economic Empowerment Demonstration (SEED) is the cleanest U.S. result. Run from 2019 to 2021, SEED paid $500 per month for 24 months to 125 randomly-selected Stockton residents. The Center for Guaranteed Income Research evaluation (West & Castro 2021) reported that full-time employment among recipients rose by 12 percentage points relative to the control group over the trial period — a result that surprised commentators who had expected a labour-supply reduction.¹³ The mechanism, by the authors’ interpretation, was that the transfer reduced job-search frictions: recipients could turn down low-quality work, take on training, and afford the small fixed costs of better job matches.

Finland’s Basic Income Experiment ran from 2017 to 2018, with 2,000 unemployed Finns randomly assigned to receive €560 per month unconditionally. The primary outcome on employment was null: there was no detectable difference between the treatment and control groups on employment days, although treatment-group respondents reported higher levels of life satisfaction and trust in institutions. The Kangas et al. (2020) final evaluation is the canonical reference.¹⁴ The Finland experiment is sometimes mis-cited as showing a negative employment effect; the published primary outcome is null, and the secondary wellbeing outcomes are positive.

OpenResearch’s Unconditional Cash Study — Vivalt, Rhodes, Bartik, Miller, and Karlan, with results reported in 2024 — paid $1,000 per month for 36 months to 1,000 participants in Illinois and Texas, against a 2,000-person control group receiving $50 per month.¹⁵ The headline labour-supply finding was a roughly two-hour-per-week reduction in labour supply among treatment participants, with modest welfare gains across several dimensions. The result is consistent with what economic theory would predict for a transfer of that size — a small income effect on labour supply, partially offset by reduced job-lock and improved match quality.

Banerjee, Niehaus, and Suri are running an ongoing twelve-year UBI study in Kenya, with the long-run results not yet reported. The design is significant for its time horizon: most cash-transfer studies have time horizons measured in months or a small number of years, and the long-run questions — about household-formation decisions, fertility, intergenerational human capital, migration — require longer panels.

The convergence across these studies is informative. The empirical question of “what happens when you just give people money on a recurring basis” has been studied at multiple scales, in multiple economies, with multiple identification strategies, and the answers cluster around a narrow set of facts: small or null effects on aggregate labour supply; modest but real welfare and wellbeing gains; consumption that flows disproportionately to food, education, and durables rather than to “temptation goods”; local price effects that are small and category-specific; fiscal multipliers in the same range as those for other forms of fiscal stimulus. None of these results, by themselves, settles the question of whether a planetary-scale dividend is workable. They do, however, decisively undermine the strong-form versions of the standard objections — that the recipients will stop working, or that the prices will absorb the transfer. Neither has happened in the studied cases.

Universal Basic Services (UBS)

The third mechanism takes the opposite side of the cash-versus-services debate. Universal Basic Services, in the formulation associated with Anna Coote and Andrew Percy’s 2020 book The Case for Universal Basic Services and the 2017 UCL Institute for Global Prosperity report, is the proposition that some classes of goods are better provisioned collectively than individually.¹⁶¹⁷ The candidate categories are housing, transit, healthcare, broadband, childcare, and (in some formulations) basic foodstuffs. The argument has three parts.

The first part is the supply-elasticity argument. In markets with low short-run supply elasticity — housing in major cities being the canonical example — a uniform per-capita cash transfer is partially captured by suppliers in the form of higher rents and prices, rather than fully accruing to recipients in the form of higher real consumption. This is the standard incidence result for an inelastic-supply market. The UBS response is to bypass the price mechanism for those particular goods: provide them directly, at a quantity calibrated to need, financed from general revenue. The collective provisioning sidesteps the supply-side bottleneck because the provisioner is, by construction, also able to expand capacity.

The second part is the insurance argument. A cash transfer provides nominal insurance against income shocks; it does not provide real insurance against price shocks in essential categories. If the price of insulin doubles, a cash transfer of constant nominal value loses real purchasing power in the category that matters most. UBS, by provisioning the good directly, provides insurance in real terms.

The third part is the dignity-and-discretion argument, and it cuts both ways. Defenders of UBS argue that universal access to essential services without a price barrier is more dignifying than receiving a cash transfer that the recipient must then spend back into the same provisioning system. Critics argue the opposite: that a cash transfer respects the recipient’s discretion in a way that paternalistic service-provision does not. Both arguments are right within their own framings. The honest reading is that the dignity-and-discretion question depends on the specific service: people generally prefer cash to housing-vouchers-restricted-to-specific-buildings, and they generally prefer collectively-provisioned emergency-room care to a cash transfer earmarked for medical use.

The empirical exemplars are well-known. The UK’s National Health Service has, with all its operational difficulties, demonstrated for nearly eighty years that universal healthcare can be provisioned collectively at a per-capita cost substantially lower than the U.S. multi-payer alternative, with health outcomes that are comparable on most measures.¹⁸ Vienna’s social-housing system, with roughly 60 percent of the city’s housing stock either municipally-owned or limited-profit, is the most-studied example of a UBS-style housing provision in a wealthy democracy.¹⁹ Singapore’s Housing & Development Board — covering more than three-quarters of the city-state’s residents — is another. Each of these systems has critics; none of them has collapsed; all of them have produced outcomes substantially different from what an unrestricted cash-transfer-plus-private-market system would have produced.

The honest tension between UBI and UBS is that UBI preserves more recipient discretion and less state capacity, while UBS preserves more state capacity and less recipient discretion. The trade-off is real and is unlikely to resolve in favour of one extreme. The reference architecture later in this essay is one that explicitly composes them. The deeper essay on the floor design — what set of guarantees should be combined to yield a credible “no-worse-than” — is at /abundance/the-floor.

Automation taxes, AI compute taxes, and data dividends

The fourth mechanism is on the funding side rather than the distribution side. The question it answers is: if the productive surplus is generated increasingly by automated systems rather than by labour, what is the tax base that funds the dividend.

The intuition that an automation-generated surplus should pay an automation-specific tax was popularised by Bill Gates in a 2017 Quartz interview, in which he argued that “if a human worker does $50,000 worth of work in a factory, that income is taxed; if a robot comes in to do the same thing, you’d think we’d tax the robot at a similar level.”²⁰ The argument is intuitive. The technical literature complicates it considerably, in instructive ways.

Acemoglu, Manera, and Restrepo’s 2020 NBER working paper “Does the U.S. Tax Code Favor Automation?” is the most-cited economic analysis of the question.²¹ The paper’s core finding is not that the U.S. tax code fails to tax automation — it is that the U.S. tax code subsidises labour-replacing automation, primarily through accelerated depreciation schedules for capital that effectively reduce the after-tax cost of capital below the after-tax cost of labour. The implication is that the relevant policy intervention is not a new automation tax in addition to the existing structure, but a correction of the existing differential between capital and labour taxation. The paper is, in effect, an argument that the modal “robot tax” proposal underestimates how far the existing tax base is already tilted in favour of capital.

The more comprehensive treatment of AI specifically as a redistribution problem is the body of work by Anton Korinek and Joseph Stiglitz, beginning with their 2018 paper “Artificial Intelligence and Its Implications for Income Distribution and Unemployment” and continuing through subsequent papers in 2021 and 2024.²² Korinek’s 2024 paper “Scenarios for the Transition to AGI” lays out the case that the distribution of AI-generated surplus depends sensitively on the elasticity of substitution between labour and capital, on the bargaining position of workers in industries undergoing rapid AI adoption, and on the tax instruments available to redistribute.²³ The paper is the closest thing the field has to a rigorous treatment of AI as a redistribution problem rather than a productivity problem.

Lanier and Weyl’s Radical Markets (2018) and Lanier’s earlier work on data dignity propose a different funding mechanism: a “data dividend” in which platforms that derive economic value from user-generated data are required to pay royalties to the contributors of that data, individually or collectively.²⁴ California’s 2019 Data Dividend proposal under Governor Newsom was the highest-profile policy translation of the idea, though it did not progress to legislation.²⁵ The data-dividend mechanism has theoretical attractions — it ties the funding stream directly to the productive activity it taxes — and substantial implementation difficulties around valuation and verification.

The standard objection to all three mechanisms (automation tax, AI compute tax, data dividend) is the elasticity objection: capital and intellectual property are mobile across borders in ways that labour is not, so a unilateral tax in any single jurisdiction will be undermined by the relocation of the taxed activity. The objection is real. It is also not decisive. The OECD/G20 Pillar Two agreement on a global minimum corporate tax — agreed in 2021 and now being implemented across most major economies — is a precedent for cross-border tax coordination that, until very recently, was widely considered impossible.²⁶ The mechanism by which Pillar Two was reached is itself instructive: a small group of large jurisdictions agreed to a minimum, and the participation cost of staying outside the regime exceeded the participation cost of staying inside. The same logic could, in principle, apply to coordinated taxation of AI compute or of automated-production surplus. Whether it will, in practice, is open. But the elasticity objection is no longer the conversation-stopper it once was.

The natural pairing is to use revenue from automation, AI-compute, and data-related taxes as the funding source for an Alaska/Norway-style dividend mechanism. The funding side and the distribution side compose. The reference architecture later in this essay assembles a particular composition. There is nothing magic about it.

Negative income tax and Earned Income Tax Credit

The fifth mechanism is the negative income tax, in the lineage going back to Milton Friedman’s Capitalism and Freedom (1962).²⁷ The intuition is structurally simple: rather than running an unconditional transfer programme alongside the existing tax-and-transfer system, fold the transfer into the tax system as a tax with a negative rate at low incomes. Above a threshold, the household pays positive tax; below the threshold, the household receives a payment that phases out as earned income rises.

The U.S. ran four large negative-income-tax field experiments between 1968 and 1980 — in New Jersey/Pennsylvania, rural Iowa/North Carolina, Gary (Indiana), and Seattle/Denver — collectively the largest social-policy experiments of the period. Munnell’s 1986 review of the four experiments is the standard summary.²⁸ The headline findings were modest reductions in labour supply, particularly among secondary earners, with effects too small to support either the strong claims of NIT advocates or the strong claims of NIT critics. The political reception of the experiments — particularly the Seattle/Denver Income Maintenance Experiment, which initially appeared to show effects on marital stability that were later attributed to data-handling artefacts — was disproportionate to the empirical findings, and the experiments are an instructive case study in how social-science evidence is consumed politically.

The U.S. did, however, ultimately adopt a modified form of the negative income tax: the Earned Income Tax Credit, established in 1975 and substantially expanded in 1986, 1990, 1993, and 2009. The EITC is the single largest cash-transfer programme in the U.S. by household reach. According to IRS Statistics of Income data for tax year 2023, approximately $57 billion was distributed to roughly 23 million recipient households, with the maximum credit for a household with three or more qualifying children at approximately $7,430.²⁹ The EITC differs from a pure NIT in that it is conditional on earned income — the credit phases in from zero, peaks at a plateau, and phases out at higher incomes — and so it is not a universal floor in the way that an NIT or a UBI would be. It is a wage subsidy with strong work-incentive properties.

The reason to include EITC in the taxonomy is that it is a working, deployed, near-cash-transfer programme at the scale of tens of millions of households and tens of billions of dollars per year. Whatever one thinks of the limits of the work-conditionality, the operational machinery — the eligibility tests, the household-composition rules, the disbursement infrastructure, the audit regime — exists and works. The argument we want to make is that an abundance-era mechanism can preserve this scaffolding while removing the work-conditionality binding constraint, by combining the EITC’s existing infrastructure with a universal floor that does not phase in from zero. The transition essay at /abundance/transition treats the institutional path of that combination at length. The point here is that the institutional infrastructure for a universal floor in the U.S. is closer to existence than is sometimes appreciated; what is missing is not the disbursement system but the political coalition for unconditionality.

Token / crypto UBI proposals

The sixth and final mechanism is the most novel and the least technically settled. A cluster of proposals over the past decade have argued that cryptographic identity systems combined with token issuance can implement a universal basic income at a scale and with a verification cost that is impractical for state-administered programmes. Worldcoin (now World ID) — launched by Sam Altman and the Tools for Humanity team in 2023 — is the most prominent, with approximately 10 million unique humans verified via the Orb biometric device by mid-2024 and the $WLD token distributed periodically to verified users.³⁰ GoodDollar, launched in 2018, is the longer-running variant.³¹

The honest assessment of this class of proposals has two parts. The first is that the identity-verification component is genuinely novel as infrastructure. The cryptographic problem of “proof of unique humanity” — verifying that a participant in a distribution scheme is a distinct biological human, exactly once, without leaking unnecessary information about that human’s identity — is non-trivial, and the deployment of a working system at the scale of tens of millions of users is a meaningful engineering accomplishment. Whether the specific Worldcoin design is the right design is contested on privacy, biometric-data, and centralisation grounds, but the existence of working “proof of personhood” infrastructure is a precondition for any planet-scale identity-conditioned distribution, including a non-token one.

The second part of the assessment is that the monetary component is weaker. A token-based UBI is, ultimately, only as valuable as the underlying token, and the value of the underlying token is contingent on its acceptance as a medium of exchange or a store of value. Tokens whose value depends on speculative inflows, on the willingness of existing holders to absorb new issuance without selling, or on a network effect that is not yet established are not credible sources of recurring real income. The standard cryptocurrency-economics critique applies: a transfer denominated in $WLD is a transfer of $WLD, not a transfer of dollars or of consumption goods, and the recipient’s welfare depends on the spot exchange rate at the moment of conversion. This is a substantive limitation, and we do not think it is the load-bearing mechanism for an abundance-era distribution.

It is, however, worth treating because it is where many of the public conversations about UBI are now happening, particularly outside the formal economics literature. A reader who has encountered “UBI” primarily through Worldcoin coverage has encountered a particular and partial version of the idea. The taxonomy above is offered partly as a corrective to that partiality.

Mechanism design under AI participation

The mechanisms in the taxonomy were designed under an implicit assumption of computational symmetry: that the participants in an allocation protocol have, to a first approximation, comparable inference capacity. Vickrey-Clarke-Groves analysis, the standard incentive-compatibility proofs, and most of the algorithmic-game-theory literature treat participants as agents who can compute their dominant strategies, evaluate their utility under various reported preferences, and reason about the mechanism’s incentive structure. They do not, typically, treat the case in which one of the participants has orders of magnitude more inference capacity than the others — including, possibly, the mechanism designer.

The arrival of frontier AI systems as plausible participants in mechanism-design settings is, in our reading, a substantive new technical question. There are at least three angles from which it is being studied.

The first is the design of mechanisms by AI systems. Dütting, Feng, Narasimhan, Parkes, and Ravindranath’s 2019 paper “Optimal Auctions through Deep Learning” showed that neural networks can be trained to discover near-optimal multi-item auctions in settings where closed-form solutions do not exist, with the network learning both the allocation rule and the payment rule jointly under incentive-compatibility constraints.³² The result generalises to broader mechanism-design settings and suggests that the design space of mechanisms is substantially richer than the analytical literature has been able to characterise.

The second is the participation of AI systems in mechanisms designed by humans. The classical literature on combinatorial winner determination — exemplified by Sandholm’s 2002 paper in Artificial Intelligence — already addressed the case in which solving the allocation problem is itself computationally hard.³³ What is new is the case in which the bidders are computationally heterogeneous, with some bidders able to evaluate their preferences over much larger combinatorial bundles than others. The incentive-compatibility analysis under that asymmetry is incomplete.

The third is the case in which a learned coordinator is itself one of the parties — for example, an AI system that aggregates and routes resources on behalf of a population. The questions of which classes of allocation mechanisms remain incentive-compatible, individually rational, and resistant to manipulation when one party is a learned coordinator with much greater inference capacity than the others is, to our knowledge, an open research problem. We treat it as such in the Apik manifesto, where it appears explicitly in the open-problems list, and in the Economic Orchestration research pillar, where it is one of the central technical workstreams.

We do not have a clean answer here. The point we want to make in this essay, where the focus is on the broader mechanism question, is that the design of distribution protocols at planetary scale is happening at a moment when one of the standard assumptions of mechanism design — computational symmetry — is breaking, and that the implications of that break are not yet understood. The taxonomy in the previous section is, in that sense, the easy part. The hard part is figuring out which of those mechanisms remain well-behaved when the participants include systems we have only just begun to characterise.

Funding feasibility at planetary scale

The taxonomy does not, by itself, settle the feasibility question. The feasibility question is whether the productive surplus that the mechanisms route is, in fact, large enough to fund a per-person dividend at a level that is economically meaningful. We can do the back-of-envelope arithmetic in public.

Global GDP in 2024 was approximately $110 trillion at market exchange rates, according to World Bank data.³⁴ The world’s population is approximately 8 billion. A per-capita dividend has the form (annual per-person amount) × (population) = (total annual dividend cost), and the question is what fraction of global GDP that cost represents.

A $5,000 per person per year dividend to 8 billion people would cost $40 trillion per year — approximately 36 percent of global GDP. This is, on any reasonable accounting, infeasible from the current productive surplus. The arithmetic is dispositive: a global gross output of $110 trillion cannot fund a transfer of $40 trillion without compressing every other claim on that output (consumption, investment, public goods, the maintenance of capital stock) to a degree that no political economy has ever sustained.

A $1,000 per person per year dividend to 8 billion people would cost $8 trillion per year — approximately 7 percent of global GDP. This is in a meaningfully different regime. For comparison: global military spending in 2024 was approximately $2.4 trillion, according to SIPRI data, or about 2 percent of global GDP.³⁵ Global fossil-fuel subsidies, on the IMF’s most-cited 2023 estimate by Black, Liu, Parry, and Vernon in “IMF Fossil Fuel Subsidies Data: 2023 Update,” reached $7 trillion globally in 2022 when both explicit and implicit subsidies are counted — approximately 7 percent of global GDP.³⁶ A $1,000-per-person dividend is in the same order of magnitude as global fossil-fuel subsidies and roughly three times global military spending. It is large. It is not obviously impossible.

A $500 per person per year dividend would cost $4 trillion per year — approximately 3.6 percent of global GDP. This is in the range of what might be funded from cross-border tax mechanisms of the kind the OECD/G20 Pillar Two precedent suggests are achievable. The OECD’s global minimum tax is projected, on the OECD’s own estimates, to raise approximately $200 billion annually globally — an order of magnitude smaller than $4 trillion, but a baseline against which more comprehensive cross-border mechanisms could be compared.³⁷

The pattern is, we think, clear. The ceiling on a planetary dividend is set by what the productive surplus can credibly fund, and the productive surplus has to grow before the dividend can. The path is not “from nothing to $5,000-per-person.” The path is closer to: today’s surplus, today’s mechanisms, today’s coordination capacity, supports a dividend at hundreds of dollars per year per person. Growth and mechanism maturation — the latter possibly faster than the former — extend it. The dividend sits within the productive economy, not above it.

The arithmetic also clarifies what the feasibility question is not about. It is not about whether $40 trillion per year exists somewhere — it does, in aggregate. It is about whether $40 trillion per year is the kind of claim that can be routed through any tax-and-transfer system that has ever existed without producing economic effects that would themselves erase the surplus. We do not believe it can. The same arithmetic holds, with appropriate scaling, for jurisdictional sub-aggregates: the U.S. could fund a $5,000-per-person dividend more easily than could the global economy, because U.S. per-capita output is higher; a sub-Saharan African jurisdiction could fund only a much smaller per-capita amount from its own surplus. The implication is that any planet-scale dividend is, in the medium term, going to be a non-uniform one — and that is itself a mechanism-design question of considerable complexity, which we mostly defer to the floor essay and the transition essay.

What the arithmetic suggests, for the purposes of this essay, is that the binding constraint on a credible distribution mechanism is not the absence of mechanisms but the size of the surplus they have to work with. The mechanisms in the taxonomy will become more powerful as the productive base grows. The growth, in turn, depends on the coordination layer that the rest of the Apik agenda is concerned with. The mechanism question and the production question are, in the long run, complements rather than substitutes.

Inflation, asset-price effects, and the supply side

The most persistent objection to a meaningful per-person dividend is the inflation objection: if you give everyone $1,000, prices will rise by a corresponding amount, and the real value of the transfer will be eroded. The objection is sometimes correct and sometimes not, and it is worth being precise about when each.

The clearest empirical evidence on the question, in a rural economy, is the Egger et al. (2022) GiveDirectly result discussed earlier. In a partially-tradeable local economy, a recurring cash transfer produced a fiscal multiplier of approximately 2.5 with limited and category-specific local price effects. The categories in which prices did rise modestly were locally non-tradeable goods — fresh produce, transport — where local supply could not adjust quickly. The categories in which prices did not rise were tradeable goods, where the village’s demand was absorbed without strain by the broader market. The multiplier exceeded one because the recipients’ spending generated income for non-recipients, who then spent the income themselves; the local economy was operating below capacity, and the transfer expanded real activity rather than nominal prices.

The richer treatment of inflation incidence across the income distribution — what economists call inflation inequality — is in Xavier Jaravel’s 2019 paper “The Unequal Gains from Product Innovations” in the American Economic Review.³⁸ Jaravel showed, using detailed scanner data, that the basket of goods consumed by lower-income households experienced systematically higher inflation than the basket of goods consumed by higher-income households over the period 2004–2015, primarily because product-innovation activity (and the deflationary effects associated with it) was concentrated in higher-income product categories. The implication is that a dividend whose nominal value is constant in time has its real value eroded faster for lower-income recipients than for higher-income ones, even before any general inflationary effect of the transfer itself. This is a complication for fixed-nominal designs and an argument in favour of indexation rules that track the consumption baskets of the recipient population.

The strongest inflation argument against a cash-only dividend, however, is the housing-supply argument. The empirical literature here is unusually clear. Albert Saiz’s 2010 paper “The Geographic Determinants of Housing Supply” in the Quarterly Journal of Economics constructed a measure of housing-supply elasticity across U.S. metropolitan areas and showed that the elasticity is sharply heterogeneous: some markets respond to demand shocks with quantity adjustments, and others respond almost entirely with price adjustments.³⁹ Glaeser and Gyourko’s 2018 paper “The Economic Implications of Housing Supply” in the Journal of Economic Perspectives reviewed the subsequent literature and laid out the policy implications: in supply-constrained markets, demand-side interventions are partially or substantially captured by landlords through higher rents.⁴⁰ A per-capita cash transfer in a supply-constrained housing market is, in the limit, a transfer to incumbent landlords.

This is, we think, the strongest empirical critique of UBI-only approaches and the strongest case for combining UBI with UBS — specifically, with policies that expand housing supply, either directly through public construction (the Vienna model) or indirectly through zoning and permitting reform. The argument is not that the cash transfer is bad; it is that the cash transfer’s real-value depends on whether the markets where the transfer is spent can accommodate the demand without price adjustment. Where they cannot — pre-eminently, in housing in supply-constrained urban areas — the cash transfer needs to be paired with a supply-side intervention or be substantially captured.

The asset-price version of the same argument applies to other supply-constrained categories: healthcare in markets with provider scarcity, education in markets with credentialing bottlenecks, transit in markets where rights-of-way are saturated. The honest design implication is that the inflation objection is correct in those markets specifically, and that the response is to address the supply constraint rather than to abandon the transfer. The mechanism-design question is whether the supply-side intervention can be sequenced with the demand-side transfer in a way that prevents the leakage. That question is not yet settled in the empirical literature, and it is one of the more interesting policy-design problems for which there is room to do meaningful work.

Political economy and incidence

A final question, on which this essay will be brief, is the political-economy question: who pays for the dividend, and who receives it. The mechanisms in the taxonomy are silent on this question only superficially. In practice, a sovereign-wealth-style fund, a UBI, a UBS, an automation tax, a NIT/EITC, and a token-based dividend all imply different distributional incidences on both the funding and the receiving side, and the differences matter politically.

The framing we find most useful comes from Saez and Zucman’s The Triumph of Injustice (2019) and their broader research programme on the distribution of taxation in advanced economies.⁴¹ Their central empirical claim is that the effective tax rate on capital income in the U.S. has declined over the past four decades while the effective tax rate on labour income has been roughly stable or rising, with the result that the overall U.S. tax system has become substantially less progressive than is typically assumed. The implication for a dividend funded from labour-income taxation is that the dividend leaves the capital-versus-labour share of national income unchanged, while a dividend funded from capital or automation taxation shifts the share. The two designs, on the funding side, have substantially different distributional incidences even if the per-capita disbursement is identical.

Piketty’s Capital in the Twenty-First Century (2014) and the subsequent Piketty, Saez, and Zucman 2018 paper “Distributional National Accounts: Methods and Estimates for the United States” in the Quarterly Journal of Economics provide the long-run empirical context for these arguments.⁴²⁴³ The historical pattern Piketty documents — capital income growing faster than national income over multi-decade periods, with the result that the share of income flowing to capital owners rises — is the empirical claim that motivates much of the contemporary interest in capital-side and automation-side funding mechanisms.

The operational point we want to make is straightforward. The funding side of mechanism design is as important as the distribution side. A dividend funded from a tax on labour income leaves the capital-versus-labour share of income unchanged; a dividend funded from capital or automation taxation shifts it. The two designs, even if they produce identical per-capita disbursements, have substantially different distributional consequences, and the choice between them is a political-economy choice as much as it is a technical one. We do not pretend to settle that choice in this essay. We do think it is important to be precise that it is a choice, and that the debate over distribution mechanisms is incomplete if the funding-side question is treated as a residual.

A reference architecture

We want to close with a sketch of one concrete proposal — without endorsing it as the answer, and without claiming it is the only credible composition. The point is to show that the mechanisms are not exotic, that they compose from existing institutional components, and that the engineering question is one of integration rather than invention.

The proposal has four components.

First, a sovereign-wealth-style fund seeded with revenues from automation taxes, AI-compute fees, and broader capital-income reforms in the spirit of the Acemoglu/Manera/Restrepo agenda. The fund operates with the legal and trustee structure of the Alaska Permanent Fund, with explicit constitutional protection against political raiding. The Norwegian Fund’s transparency standards and its ethical-investment guidelines provide an operational template for governance.⁴⁴ The fund’s investment universe is broad-based and indexed; its return-target is real returns above inflation plus a population-growth adjustment.

Second, an annual per-capita dividend to every verified resident, modelled on the Alaska PFD design but calibrated to the fund’s sustainable real-return capacity. The verification mechanism is the binding constraint on the dividend: it has to be robust to fraud (sufficient to defeat duplicate registrations) without being so invasive that it imposes unacceptable privacy or autonomy costs on recipients. The Worldcoin / proof-of-personhood literature is one input here; existing national identity systems and biometric voter rolls are another. The dividend is paid in fiat currency, indexed to the consumption basket of the bottom income decile.

Third, universal basic services in the housing, healthcare, transit, broadband, and childcare categories, alongside the dividend. The provisioning is local and federated; the standards are national. The argument for including UBS is the supply-elasticity argument from the previous section: in markets where supply is inelastic, the dividend alone leaks to incumbent suppliers, and the supply-side intervention is what makes the dividend’s real value robust. The UBS component preserves recipient discretion in the cash dividend by removing the categories in which discretion is most likely to be defeated by market structure.

Fourth, an EITC-style top-up that preserves the work-incentive properties for the share of the population that chooses to work. The top-up phases in from zero earnings, plateaus at a calibrated income level, and phases out at a higher one. The phase-out rate is calibrated to be small enough that the implicit marginal tax on additional earnings is below the plain-vanilla labour-income tax rate, so that the transfer system does not punish the choice to participate in the labour market. The argument for retaining this component is that the existing EITC infrastructure works at scale and that there is no obvious reason to discard it merely because the floor has been lifted by the dividend and UBS components.

The four components together provide redundancy. If the fund underperforms in a given year, the dividend is lower but the UBS floor is unaffected. If a particular UBS category is undersupplied — housing in a particular city, say — the dividend provides cash to seek alternative provisioning. If the labour market shifts in ways that compress earnings for participating workers, the EITC top-up rises to absorb part of the shock. No single component bears the full weight; each is a partial provision of the same thing.

The proposal is unfinished. It does not specify the funding levels with precision; it does not specify the international coordination required to prevent capital flight; it does not specify the verification mechanism in operational detail; it does not specify the governance structure of the fund’s trustees in ways that would resist political capture over multi-decade periods. Each of those is a substantial engineering and policy problem in its own right. The point of the sketch is to demonstrate that the mechanisms, in composition, are not exotic. They have been built before. The novelty is in the composition, not in the components.

Where to read further

The two essays that follow this one in the /abundance section take the next two steps. The Floor treats the question of what minimum guarantee, in composition, the components above should yield, and at what level. Precedents is a longer historical treatment of the dividend programmes referenced in the taxonomy — Alaska, Norway, Iran, Macao, Eastern Band of Cherokee Indians casino payments — at the level of detail this essay deferred. Transition is the institutional-path essay: how an economy gets from where it is to a credible reference architecture. Post-labour treats the harder long-run question of what work, identity, and meaning look like in a regime where the economic compulsion to labour has been substantially relaxed. The technical substrate of mechanism design under AI participation is treated in the Economic Orchestration research pillar and the systems-level treatment. The broader argument for why coordination is the load-bearing primitive is in the manifesto, and the operational safety stance that governs this work is in the safety principles.