Long-Term Institutions

Long-term institutions are the machinery by which a decision taken now is supposed to still be in force when everyone who took it is dead. The catalogue is short and familiar: constitutions and the rules for amending them, statutory duties toward future generations, commissioners and ombudsmen with a long-horizon remit, sovereign wealth funds that convert a depleting asset into a permanent one, sunset clauses that are meant to make a law die of its own accord, pension promises that run for two human generations, and repositories designed to stay safe for longer than writing has existed.

The question this brief asks is not whether such things exist. They plainly do. It is whether any of them binds. An institution a successor government can abolish, a fiscal rule a simple majority can rewrite, a duty that carries no consequence for ignoring it and a discount rate that a memorandum can rescind are all commitments in form. This brief asks what the measured record says about which of them hold, and it takes the answer from survival analysis, audit findings and price data rather than from the founding documents.

Where this brief stops. Civilizational Planning owns planning for long futures: foresight units, national five-year plans, the Montreal Protocol, the accuracy of long-range forecasting, and the argument about what social discount rate correctly values a person born in 2200. This brief owns the machinery of durability regardless of what is being made durable. The arbitration rule where the two meet is whether the question concerns the content of a plan or the mechanism holding any plan in place. Three subjects appear in both — Wales's future-generations Act, the ten-thousand-year waste markers, and the 2024 United Nations declaration — and the overlap is deliberate: there they are evidence about the quality of long-horizon planning, here they are evidence about binding force.

One consequence of that cut deserves stating at the outset, because it inverts the way the subject is usually organised. The most durable long-horizon institutions in the record are not the ones designed to be durable. They are pension schemes, sovereign debt and universities — none of which was built as a commitment device, and all of which have outlasted every purpose-built one. The purpose-built devices, taken together, have not yet produced a documented case of changing an outcome.

The second thing worth stating early is what kind of evidence this brief accepts. Long-term institutions are unusually well supplied with founding documents and unusually poorly supplied with outcome data, and the two are easy to confuse. A statute establishing a duty is not evidence that the duty was discharged; a governance charter is not evidence that a fund was left alone; an eternity clause is not evidence that anything was made eternal. So this brief takes its quantities from four places wherever they exist: survival analysis on large samples of constitutions and firms, audit-office findings, market prices and payment records, and the fiscal accounts of the funds themselves. Where only an institution's own description of itself was available — which is the ordinary case for future-generations bodies — the brief says so and treats the absence of outcome data as itself a finding rather than as a gap to be filled with the institution's own account.

Established The base rate of institutional survival is low, and surviving does not make an institution safer. Elkins, Ginsburg and Melton's endurance research — the standard reference — estimates the mean life expectancy of a national constitution at 19 years. A 2024 study covering 209 constitutions in 118 countries between 1974 and 2023, on 4,194 country-year observations, puts the median age at about 12 years, with half replaced by that age. On the corporate side, an analysis of United States publicly traded companies in Compustat back to 1950 finds the typical firm exits after about ten years by merger, acquisition or liquidation — and, crucially, that a firm's mortality rate “had nothing to do with how long it had already been in business or what kinds of products it produced.” An age-independent hazard means an exponential survival curve, which means institutions accumulate no immunity from having lasted. The intuition that an old institution is thereby a durable one is not supported in either class where survival has been modelled properly.

Established The formal entrenchment devices do not do the work attributed to them, and there is now a number for it. A 2025 study coding unamendability rules across more than 180 countries, with the rule coding running back to 1789 and the main analysis covering 1946 to 2018, counts 74 constitutions with eternity clauses, 25 with amendment review and 38 with judicially created unamendability doctrines as at 2018. Then it looks at what happened when the rules were tested. Across 40 documented attempts to remove presidential term limits between 2000 and 2020, courts invoked an available unamendability rule in only 2 cases out of the 31 in which such a rule existed — while 16 attempts were defeated by ordinary political opposition alone. On that record, politics did eight times as much binding as law.

Established Entrenchment is also associated with routing around rather than compliance. The same study finds unamendability rules associated with higher rates of wholesale constitutional replacement — consistent with an executive who cannot amend simply writing a new constitution. Countries with eternity clauses score about 2.6 points lower on Polity and 0.47 standard deviations lower on Voice and Accountability by 2018, and the within-country improvement of 1.1 to 1.4 Polity points after adoption is attributed by event studies to changes occurring before the rule was adopted. The rule follows the democratisation; it does not cause it. And where an eternity clause did stand directly in the way, it was removed: Honduras's constitution made the presidential term limit of Article 374 unamendable, and in 2015 its own Constitutional Court declared the unamendable provision unconstitutional.

Established What does predict constitutional survival is inclusion at the founding, and participation is not the same thing. In the 209-constitution study, processes that included the societal and political groups with the power to defect produced a median durability of 16 years against 10 years for non-inclusive ones, and a 54% lower risk of replacement. Processes that were merely participatory — referendums, public consultation, the thing constitutional-reform campaigns usually demand — produced 14 years against 10 and the relationship was not statistically significant. The proposed mechanism is self-enforcement: an inclusive founding leaves behind constituencies willing to defend particular provisions when they are attacked. That is a political mechanism, not a legal one, and it is the closest thing to a design lever in this literature.

Established In sovereign funds, what is entrenched survives and what is left to statute is spent. Alaska is the controlled comparison, because both regimes run inside one institution. Alaskans amended the state constitution in 1976 to require that at least 25% of mineral royalties be paid into a Principal that cannot be spent. Everything else was left to statute: the Earnings Reserve Account “can be appropriated by the Legislature through a simple majority vote”; the percent-of-market-value draw limit of 5.0% dates only to 2018 and sits in Alaska Statute 37.13.140(b); and the dividend formula — 50% of income available for distribution, under AS 37.13.145(b) — is likewise statutory. The Principal is intact, at $87,898,600,000 as at 31 July 2026. The statutory layer is not: two incompatible draw formulas now operate simultaneously because the 2018 rule was added without repealing the 1980s one, and the state's own Legislative Finance Division puts a 54% probability on the Earnings Reserve being exhausted within the next decade, with overdraws proposed from fiscal year 2027.

Established Alberta is the same experiment without the constitutional layer, and it did not need a crisis to fail. The Heritage Savings Trust Fund was established on 19 May 1976 with a commitment of 30% of annual non-renewable resource revenue. The rate was cut to 15% in 1983, investment income was diverted to budgetary revenue from the same year, and deposits ended altogether in 1987. By 1993 the cumulative diversion of income had reached $15.4 billion, more than double the fund's then value of $15.3 billion. Per-capita value peaked in 1983 at roughly $12,500 per Albertan and stood at about $4,200 at the end of 2021; fair market value at 31 December 2025 was $31.9 billion. This fund was not raided. There was no emergency, no seizure and no default. There was a statute, and a legislature amended it by ordinary majority twice within seven years of founding and then stopped feeding it. A commitment device a simple majority can rewrite is a budget line with a name.

Established The longest-honoured financial commitments in the record were both rewritten by the debtor, at the debtor's convenience. A perpetual bond issued on 15 May 1648 by a Dutch water board, the Hoogheemraadschap Lekdijk Bovendams, has paid interest for 378 years across the Dutch Republic, French occupation, a kingdom, two world wars and several reorganisations of the obligor. Its coupon was also cut unilaterally from 5% to 3.5% and then to 2.5%. The United Kingdom's undated gilts — one tracing to July 1729, another to South Sea stock of 1726 and 1751 — were called in between February and July 2015 because low rates made refinancing cheap. Neither commitment was repudiated and neither was enforced. They lasted because honouring them was cheap and ending them was not worth the trouble, until it was.

Frontier The synthesis this brief offers: nothing legal binds a future sovereign, and four non-legal things appear to. Entrenching a stock rather than a flow; a living, organised constituency with a financial stake; inclusion at the founding, worth a 54% reduction in replacement hazard; and cheapness, which is why an obligation of about eleven euros a year has outlived three constitutions of the country paying it. The instruments most often proposed as commitment devices — duties to consider, advisory commissioners, declarations, sunset clauses and amendment thresholds — are between them supported by no documented case of changing an outcome, one randomised experiment finding they do roughly the opposite of what is claimed, and one large-n finding that courts do not reach for them when the moment arrives.

Frontier The genuinely open question in the constitutional literature is whether the amendment rule matters at all, and two credible sources disagree. Ginsburg and Melton's 2015 analysis — 790 constitutions in the full sample, 194 democratic constitutions, 450 observations on the Comparative Constitutions Project measure — finds the institutional procedural variables statistically insignificant as predictors of amendment rates, and finds instead that “amendment culture is more important than institutional constraints in explaining amendment practices,” where culture is operationalised as the amendment frequency of the country's previous constitution. Their conclusion is blunt: “constitutional designers have little influence over the observed flexibility of their product.” An unrefereed preprint using Bayesian negative binomial models across 57 democratic constitutions, 80 democracies from 1950 to 2013 and 47 to 48 United States states biennially from 1987 to 2009 reaches the opposite conclusion: “a standard-deviation change in constitutional rigidity tends to reliably decrease the number of amendments ratified over time.” This brief does not resolve the disagreement. The defensible middle is narrow and comes from the first paper's own methodology section: seven published indices of amendment difficulty correlate poorly with each other, most below 0.5 and one negatively, which the authors read as “validity problems with extant measures.” Any estimate of the rules' effect inherits that measurement problem, in whichever direction it points.

Frontier Whether a future-generations body has ever changed a decision is unresolved because the record was not kept, which is a different finding from a null result. Hungary's Ombudsman for Future Generations holds the strongest formal instrument in the field anywhere: deriving its mandate from Article P) of the 2011 Fundamental Law, it can examine and comment on national and local legislation, publish non-binding recommendations, initiate investigations, intervene in public administrative court cases on environmental protection, and propose that the Commissioner refer a matter to the Constitutional Court or to the Curia. Its own descriptive account of itself gives no staff size, no annual case or investigation statistics, no Constitutional Court referrals or outcomes, and no instance of a changed decision or law. A peer-reviewed treatment describing the office as having “an outstanding role both at the domestic and international levels” contains none of those figures either. The strongest legal instrument in the field is exercised at an unknown rate with unrecorded effect. Reporting that as evidence of no impact is the same error the advocates make, in the opposite direction.

Frontier Whether constituency rather than design explains fund survival is the live causal question and nobody has tested it. The Alaska Permanent Fund's political durability plausibly rests on roughly six hundred thousand people receiving an annual cheque, which makes any attack on the fund an attack on an organised electorate. That is this brief's inference and it is not a finding; no study testing it was located. It matters because if it is right, the design literature on sovereign funds — governance charters, expert commissions, transparency indices — is optimising the wrong variable, and the operative question for any new fund is who gets paid rather than how the rules are written.

Frontier Kiribati is the anomaly that most embarrasses the design literature, and it has not been explained. The Revenue Equalization Reserve Fund, founded in 1956, is thirty-four years older than Norway's and the oldest continuously operating sovereign fund identified for this brief. It has no constitutional entrenchment and no internationally celebrated governance framework. It took a double hit after 2008 — exposure to failed Icelandic banks compounded by government drawdowns to finance budget shortfalls — falling from A$637 million, or 420% of GDP, in 2007 to A$570.5 million, or 350% of GDP, in 2009. It then recovered to A$1,153.4 million in 2019, $1.352 billion in 2021 and $1.523 billion in 2024. A fund that was both raided and caught in a banking collapse is, seventy years on, at its largest ever, while several funds surrounded by commissions and constitutional amendments have worse records. No source consulted explains why.

Frontier What written commitment actually changes is a measured but strange result. A 2025 study of constitutional compliance covering 171 countries from 1951 to 2020, on 8,743 country-year observations and roughly 2,000 amendments and over 300 new constitutions, scores de jure against de facto practice across 14 protections. New constitutions in democracies raise compliance by about 24% of a standard deviation, sustained across a ten-year window and concentrated in political rights. New constitutions in non-democracies raise it by about 8% of a standard deviation for a year or less; amendments there manage about 6% for two to three years. Written commitment improves behaviour where behaviour was already broadly lawful and barely at all where it was not, which is close to the inverse of what a commitment device is supposed to do.

Handwave Any claim that a mechanism exists which could bind an institution over ten thousand years. The marker literature that is usually invoked here is about whether a message survives its senders, not whether an institution can be held to a promise; the two problems have been conflated for decades. The best-documented commitments in the entire record ran for roughly three centuries and then had their terms changed by the obligor. There is no commitment analogue at the marker horizon, no literature proposing one, and nothing in the survival data that would make one plausible.

Established The first constraint is that a rule is only as strong as the instrument carrying it, and almost every long-horizon rule is carried by ordinary statute. Alberta's deposit rule was amended by ordinary majority in 1983 and abandoned in 1987. Alaska's dividend formula and its draw limit both sit in statute and both have been overridden or duplicated. This is the binding constraint because it is upstream of every other: no amount of good design in a statute survives a legislature that can amend the statute in an afternoon, and legislatures can.

Established The second is that constitutional entrenchment protects only what it names, and it usually names the wrong thing. Alaska's constitution protects the Principal, which nobody wanted to spend, and says nothing about the Earnings Reserve, which is what everybody wants to spend. The entrenchment worked perfectly and was almost irrelevant. Designing entrenchment onto the contested layer rather than the uncontested one is the obvious response and no jurisdiction in this brief has done it.

Established The third is that courts do not use the entrenchment instruments they have. The most direct measurement available: across 40 attempts to remove presidential term limits between 2000 and 2020, courts invoked an unamendability rule in 2 of the 31 cases where one was available. Sixteen attempts failed on political opposition instead. Whatever theory of judicial guardianship underwrites eternity clauses, the observed behaviour of the guardians does not match it, and Honduras's court struck down its own eternity clause rather than enforcing it.

Frontier The fourth is that the bodies charged with representing the future hold no enforcement power almost anywhere. A duty to be considered, a right to opine, a power to delay. The strongest instrument located in this field is a proposal that a superior commissioner refer a matter to a constitutional court, held by a deputy inside a general rights office — and the office publishes no count of how often it has been used or with what result.

Established The fifth is that institutional memory degrades faster than materials do, and somebody has quantified it. Carried forward from the published brief: two independent expert panels designing markers to deter intrusion into a waste repository over ten thousand years put the probability of a marker physically persisting at between 0.40 and 0.85 depending on construction, and the probability that an intruder would correctly interpret the message at between 0.01 and 0.60. The gap between those two rows is the whole finding: the medium outlives the message by an order of magnitude in reliability. One panel assumed scholarship capable of translation would survive somewhere; the other assumed political change and concluded that passive controls “can never be assumed to eliminate the chance of inadvertent and intermittent human intrusion”.

Established The sixth is that the near term binds harder than the far term, which is the most counter-intuitive constraint in the subject. Also carried forward: a deep geological repository designed for hundred-thousand-year isolation submitted its operating licence application at the end of 2021, saw the regulator's substantive review begin in mid-2022, went through two formal deadline extensions with the regulator citing deficiencies and inconsistencies in the submitted material, and received a favourable safety statement only in August 2026. Four and a half years for a regulatory opinion, on a facility intended to outlast every institution reviewing it, with the licence still not granted at the time of writing.

Frontier The seventh, and the one nobody designs against, is cost. The commitments that lasted longest lasted partly because ending them was not worth the administrative trouble — a perpetual bond now paying on the order of eleven euros a year, serviced across 378 years. When ending them did become worthwhile, they ended: the United Kingdom called in 286-year-old debt in 2015 because low interest rates made refinancing cheap. A commitment whose continuation depends on remaining beneath the threshold of attention is not robust; it is unnoticed.

Established Nothing here waits on a research result. Every constraint in this brief is statutory, fiscal, judicial or administrative, and every one of them could be relieved tomorrow by a body that wanted to relieve it. That is unusual in this corpus and it is worth stating plainly rather than burying: the reason long-horizon institutions do not bind is not that we do not know how to build them.

Established What they do wait on is a set of four conditions, each of which has a measured counterexample behind it. An instrument a successor government cannot rewrite by ordinary majority — which none of the abolished commissions had, and which Alberta's deposit rule conspicuously lacked. An enforcement mechanism rather than a duty to consider, since the audit finding is precisely that consideration without consequence produces prominence without change. Entrenchment placed on the contested layer rather than the uncontested one, which no jurisdiction has yet attempted. And a constituency that will personally notice a default, which is what distinguishes pension obligations — carrying an audited $1.32 trillion gap on a 74% funded ratio across United States state plans in 2023 and still being paid — from every purpose-built device in this brief.

Established What waits on this brief is everything downstream of a promise. Nuclear Waste Solutions depends on a custodial institution surviving longer than any institution in the historical record has survived. Civilizational Planning depends on the machinery described here for every plan it evaluates: a foresight finding that changes nothing is a durability failure, not a forecasting one. Global Cooperation Models inherits the same commitment problem between states rather than between generations, and the term-limit data suggests the answer there will also be that politics binds and instruments do not.

Frontier And one dependency runs the wrong way round, which is worth flagging because it is easy to miss. The quality of the evidence in this brief depends on institutions choosing to be audited. The single most-cited fact in the whole subject is an audit finding, and it exists because one national audit office decided that a statutory duty was within its remit. Nothing obliges any other audit office to make the same decision, nothing obliges a future-generations body to publish its own caseload, and nothing obliges a sovereign fund to report whether its rule was followed rather than merely what it is worth. The field's evidence base is therefore itself a long-horizon institution with no commitment device behind it, and on this brief's own argument that makes it exactly as durable as the current willingness to fund it.

Established The cheapest useful experiment is to audit these bodies the way spending is audited, repeatedly and independently. It has been done properly once, ten years in, in one jurisdiction, and it produced the clearest finding in the field. Running the same exercise at five-year intervals across every jurisdiction that now has such a body would generate the comparative evidence that does not currently exist — and would, incidentally, establish whether the Welsh finding is about the Act or about Wales.

Established The most valuable single measurement is also the cheapest: publish the counts. Hungary's ombudsman office could state, in a paragraph, how many legislative opinions it issued, how many Constitutional Court referrals it proposed, how many were made, and how many succeeded. Every future-generations body in the world could do the same. Until one of them does, the entire field's impact question is unanswerable in principle rather than in practice, and the absence of the counts is the most informative fact about the institutions producing them.

Frontier The natural experiment already running is Alaska's earnings reserve. The state's own analysts put the probability of exhausting the spendable account within a decade at 54%, with overdraws proposed from fiscal year 2027. If the account is exhausted while the constitutionally protected Principal remains untouched, that is about as direct a confirmation of the stock-versus-flow proposition as any institution will ever supply, and it resolves within ten years without anyone building anything for the purpose. If instead the legislature reconciles the two draw formulas before the account runs dry, the counter-reading — that statutory rules can be repaired by ordinary politics before they fail — gains its first real data point.

Frontier The decisive test of the entrenchment literature is a term-limit case in which a court uses the rule. The measured base rate is 2 uses in 31 opportunities. Every future attempt to remove a term limit in a country with an eternity clause is an observation on that ratio, they occur at a rate of roughly two a year on the 2000 to 2020 record, and nobody needs to design the experiment because autocrats keep running it.

Frontier The unrun experiment that would settle the constituency hypothesis is a fund with an entrenched dividend. Alaska entrenched a corpus and left the payment statutory. A jurisdiction that constitutionalised the payment instead — the flow rather than the stock — would test directly whether it is the legal instrument or the electorate that does the binding. No jurisdiction has done it, and the fact that none has is itself evidence about which constituencies get their preferences written into constitutions.

Established A negative result worth recording as an experiment in its own right. Sunset clauses were tested experimentally on 1,639 United States adults randomised across topic, political valence and sunset type, including a conditional sunset tied to an effectiveness evaluation. The finding: “sunsets did not increase overall support for laws, contrary to prior suggestions of an overall compromise effect,” with an asymmetry in which sunsets increased liberals' support for conservative legislation but not the reverse. A device widely recommended as a bipartisan compromise mechanism, tested directly, did not produce the compromise and exposed one side to exploitation.

Frontier And the measurement nobody has attempted: present candidate marker designs to people with no context and test interpretation. The 0.01 to 0.60 comprehension range is expert elicitation, not observation, and it is the widest uncertainty band anywhere in this brief. It is also trivially testable on living subjects at negligible cost.

Established The one design lever with a clean measured effect is which part of an institution you entrench. Every sovereign fund in this brief separates a stock — the corpus, principal or capital — from a flow, being the income, draw or dividend. The record is uniform: whichever layer sits in a constitution survives, and whichever sits in statute is rewritten. Alaska demonstrates both halves inside one institution and one legislature.

Established The mechanics of Alaska's failure mode are worth stating precisely, because they are a general failure mode of layered rules. The 1980s statutory formula draws 21% of a five-year average of income. The 2018 percent-of-market-value rule draws 5% of a trailing five-year average of assets, being the Earnings Reserve and corpus combined. The legislature enacted the second without repealing the first, so two incompatible formulas now govern the same account. Two further effects compound it: the fund's shift toward illiquid long-horizon investments raises returns while reducing the realised income that flows into the spendable account, and annual inflation-proofing transfers to the corpus continue even though the market-value rule arguably makes them redundant. None of this is a failure of investment management. It is a failure of statutory housekeeping, and it is what produces a 54% chance of running the account dry.

Established The counterfactual arithmetic on Alberta is instructive and it belongs to an interested party. The Canadian Centre for Policy Alternatives, an advocacy think tank whose model supplies the estimate, calculates that maintaining the original 30% deposit rule would have produced a fund of about $688 billion by 2021, that 15% would have produced about $500 billion, and that even with no contributions at all after 1987, simply retaining investment income rather than diverting it would have produced about $237 billion. The direction and order of magnitude are robust against the fund's actual trajectory; the precise figures are the advocate's and this brief does not adopt them. The third scenario is the load-bearing one, because it isolates a single decision — whether the fund keeps its own earnings — and Alberta did not restore that until 2025.

Established Sunset clauses have an architecture and it does not point where the name suggests. A sunset provision sets an expiry date on a statute, on the theory that the burden of proof shifts to whoever wants it continued. In the empirical record the mechanism runs the other way. Roughly 13% of bills carrying a sunset provision passed the United States Congress against 5% of permanent legislation — so a sunset roughly doubles a bill's chance of enactment. Expiry is rare: on one international survey, “the expiry of anti-terrorism legislation is extremely rare,” and the sixteen sections of the USA PATRIOT Act originally due to lapse on 31 December 2005 were “reauthorised several times in the following years after only very limited evaluation.” The scrutiny at renewal is documented rather than assumed: one 2003 House of Lords renewal debate had four peers speak, one of them the minister, and a 2011 debate ran 32 minutes.

Established Marker engineering is the one place where redundancy is the whole design, and it is carried forward from the published brief unchanged. The expert panels' recommendations are almost entirely about layering: earthen berms with deliberately hostile geometry, inscribed monoliths, buried message chambers with stone-slab records, capsules in ceramics and sintered alumina, a world map of other waste sites, periodic tables and astronomical calendars, and preservation of part of the facility for future archaeology. Multilingual text across many languages, and minimal reliance on pictographs — the opposite of the popular assumption that symbols travel better than words. The operative conclusion was that no single approach suffices. Whether any of it has been built is not established: markers are a post-closure activity and no repository has closed.

The nearest neighbour is Civilizational Planning, and the two briefs are best read as a pair with a clean division: that one asks whether anyone can plan across generations, this one asks whether anything holds a plan in place once made. They share three subjects deliberately — the Welsh Act, the ten-thousand-year markers and the 2024 declaration — and each takes a different half. Read together they make an argument neither makes alone: that the planning literature's failure to find effects and this brief's failure to find binding instruments are the same finding approached from opposite ends.

Nuclear Waste Solutions owns the disposal engineering from which this brief borrows its extreme case, and the relationship is one-directional: the geology is settled and the custodianship is not. Future Democracies shares the mortality finding directly, since the constitutional survival data in this brief is data about democracies dying. Global Cooperation Models faces the same commitment problem across space rather than across time, and the term-limit evidence here — instruments unused, politics decisive — is the most transferable result this brief has to offer it.

Outside the map, four fields carry most of the quantitative content and none of them describes itself as being about long-term institutions. Comparative constitutional law supplies the survival analysis, and it is the only body of work in this subject with large samples, coded covariates and hazard models. Organisational ecology supplies the firm-mortality result, which is where the age-independent hazard finding comes from and which no writing on institutional design appears to have absorbed. Public pension actuarial practice supplies the largest long-horizon commitment anyone actually honours, together with the discount-rate choice that quietly decides how much of it falls on the unborn. And financial history supplies the only three-century commitments in the record, along with the uncomfortable observation that both of them were renegotiated by the debtor rather than being enforced by the creditor.

Two adjacencies are worth naming for what they are not. Semiotics and linguistics appear in this subject only through the marker literature, and they belong to the communication problem rather than the commitment problem; treating them as central has made the field look more exotic and less tractable than it is. And behavioural work on temporal discounting in individuals — the psychology of impatience — is frequently invoked as an explanation for why institutions fail to plan. Nothing in this brief's evidence requires it: the failures here are legislatures amending statutes they were entitled to amend, and courts declining to use instruments they were entitled to use.

Established The institution that does the binding, in every case examined, is the one with a constituency — and none of them was built for the purpose. United States state pension plans carry a reported funding gap of $1.32 trillion as at 2023 on an overall funded ratio of 74%, unchanged from the previous year, with employer costs ranging from 6.2% of payroll in South Dakota to 62.6% in Connecticut. These are enforceable multi-decade promises — sixty to seventy years from a hire date to the last survivor benefit — honoured through recessions, and in several states protected against impairment more strongly than any future-generations statute anywhere. Nobody calls them long-term institutions. They are the most successful ones in the record, and they work because the beneficiaries are alive, organised and voting.

Established The bodies actually charged with the long horizon are, by contrast, structurally weak in a specific and consistent way. The standalone parliamentary commissioner in one jurisdiction was downgraded to a deputy inside a general human-rights office under a new constitution, retaining a proposal route to the constitutional court which is the strongest formal instrument any body in this field holds. Another national commission created by statute in 2001 with a genuine delaying power over legislation was abolished in 2006 on the stated ground of unjustified budgetary burden, with no successor appointed for four years and the enabling law repealed only after a supreme court ruled the non-appointment unlawful. At international level the institution was declined outright: a 2024 declaration “takes note of” a proposed special envoy rather than establishing one, and in June 2025 the member-state budget committee rejected the post and it was removed from the following year's proposed budget.

Established The abolition reasons are worth reading carefully, because the stated reason and the operative reason differ and the operative one is not frivolous. The official ground in the clearest case was budgetary. The contemporaneous account attributes it to the commissioner's independence and discretionary power over bills, and to a legislator's objection that the commission failed to distinguish between supplying facts and determining what is morally good. That objection is the real one, and it is not obviously wrong — a body speaking for people who do not exist yet is making a moral claim, and there is no procedure for adjudicating it. An institution whose authority rests on a contested moral claim and whose budget rests on an annual vote is not a durable institution; it is a persuasive one, subject to renewal.

Frontier The comparative evidence base is the field's weak spot and it is weak by omission rather than by difficulty. The literature on future-generations bodies is largely descriptive institutional design published by the bodies themselves and by advocacy organisations, with almost no outcome measurement: no staffing figures, no counts of opinions issued against opinions acted on, no evaluation. Every one of those numbers is already in the offices' own files. The honest position is that we know these bodies exist, we know several were abolished or downgraded, and we do not know what any of them achieved.

Frontier The institution that does not exist is an auditor of long-horizon commitments. One national audit office has examined one such Act once. No body anywhere audits whether a sovereign fund's rule has been followed, whether a sunset was scrutinised at renewal, whether a future-generations opinion changed anything, or whether an entrenched provision was honoured. Where such auditing has happened it has been decisive — an audit finding is the single most-cited fact in this whole subject — and the reason it is decisive is that almost nothing else in the field is measured at all.

The evidence base in this field is dominated by parties describing themselves, and the pattern is consistent enough to be a rule. Established Every future-generations body examined for this brief publishes a descriptive account of its own powers and none publishes a count of its own interventions or their outcomes. Hungary's office lists seven categories of power, including a route to the Constitutional Court, and supplies no staff size, no annual caseload, no referral count and no result. Sovereign fund managers publish market values and governance charters. Advocacy think tanks supply the counterfactuals. This brief marks interested parties in its reading list and takes quantitative claims from audit offices, peer-reviewed survival analysis and price data wherever both are available.

Established The standing pattern is nonetheless unusually favourable and it is worth saying so. Unlike most governance subjects, this one has genuine audit-institution evidence — a national audit office on the flagship Act, an international financial institution's surveillance on the flagship fund, a nuclear regulator on the repository — and it has a real quantitative literature in comparative constitutional survival analysis. Those sources are consistently less flattering than the bodies' own accounts, which is the ordinary result and the reason for preferring them.

Established The discount rate remains the place where intergenerational ethics is actually decided, and it is decided by parameter in two separate systems. Carried forward: a pure-time-preference term of half a per cent and an elasticity of one are ethical positions expressed as arithmetic, and moving the growth assumption from 2.0% to 1.5% and the elasticity from 1.0 to 1.25 shifts the standard rate by more than a percentage point, which changes which projects are worth doing for people not yet born. That it is settled by technical guidance rather than by legislation is the substantive point, and it is why a memorandum could reverse it. The second system is less discussed and larger: United States state pension plans assume returns of 5.8% to 7.6%, most clustering at 6.8% to 7.2%, and the assumption is a discount rate. A plan assuming 7.2% rather than 5.8% books a smaller liability today and shifts contributions onto people not yet hired. The same ethical choice is made twice, once in a public appraisal document and once inside an actuarial valuation nobody debates.

Frontier The deeper problem is the one the abolished commission ran into and it has not been answered. Representing future generations requires someone to say what those generations will want, and nobody can know. A legislator's objection to the best-known such commission — that it failed to distinguish between supplying facts and determining what is morally good — is not obviously wrong. The strongest available answer remains procedural: preserve option value, avoid irreversibility, publish the undiscounted profile and let the fairness argument happen in the open. That is notably weaker than the claim these institutions are usually founded on.

Established A disclosure obligation this brief takes on itself. Four of its sources are advocacy organisations or interested parties whose framing is not relied on: an advocacy think tank supplies the Alberta counterfactual, a democracy-assistance foundation supplies the sunset-clause survey, a fund manager supplies its own fund's structure, and an ombudsman's office supplies its own powers. Three sources are tertiary encyclopaedia entries, used for chronology and marked as such. And thirteen sources are carried forward from the published brief without being re-fetched in this round, because the research tool available could only retrieve URLs surfaced by a bounded search budget; every figure attributed to them is unchanged and no new claim rests on any of them.

Established The civilisational finding is that no legal device in the record binds a future sovereign, and the evidence for it is now numerical rather than rhetorical. Courts invoked available unamendability rules in 2 of 31 term-limit cases while ordinary politics stopped 16. A constitution's estimated mean life expectancy is 19 years and the median age of the post-1974 cohort is about 12. A firm's mortality rate is independent of its age. A statutory deposit rule was halved seven years after founding and abandoned eleven years after. A dividend formula written into statute has been overridden for a decade. A 286-year-old debt was redeemed when redeeming it got cheap. Every mechanism examined for this brief turned out to be exactly as durable as the ongoing political consensus that tolerated it, and no more.

Frontier What appears to bind is not legal at all, and this is the constructive half. Four things recur in the cases that lasted. Entrenchment of a stock rather than a flow, which is why Alaska still has its Principal. A living, organised constituency with a financial stake, which is why pension promises — carrying a $1.32 trillion audited gap on a 74% funded ratio across United States state plans and still being paid — are the most durable long-horizon commitment any government keeps despite never having been designed as one. Inclusion at the founding, worth a 54% reduction in replacement hazard where mere public participation is worth nothing measurable. And cheapness: an obligation of about eleven euros a year survived 378 years partly because ending it was never worth anyone's afternoon.

Speculative The general principle this case illustrates, stated for the corpus. Durability is not a property of instruments; it is a property of coalitions, and instruments matter only insofar as they create or protect one. A rule with a constituency outlives a rule with a court behind it. This has an uncomfortable implication for the whole future-generations project, which is founded on representing a constituency that by definition cannot vote, cannot organise and cannot notice a default — and which, on the evidence assembled here, has never been shown to have changed a decision anywhere.

Handwave Claims that any existing institution meaningfully constrains what a future government may do are not supported by anything in this record. The strongest counterexample anyone can offer — a three-century bond — had its terms cut twice and was eventually called.

These horizons track statute, budget cycles, litigation and regulatory process rather than research results:

• 10 yr: Established Alaska's Earnings Reserve is exhausted or it is not, on the state's own 54% estimate, with overdraws proposed from fiscal year 2027 — the clearest near-term test of the stock-versus-flow proposition. Established The first repository operating licence is granted or refused, which will be the only real-world datum on hundred-thousand-year institutional design. Frontier At least one further future-generations body is abolished, downgraded or merged, on the base rate so far. Frontier Roughly twenty more attempts to remove presidential term limits occur, on the 2000 to 2020 rate, testing the 2-in-31 judicial-invocation ratio.
• 25 yr: Speculative The plausible split is that fiscally constituted long-horizon institutions persist while morally constituted ones do not, because a number in a rule with a constituency behind it is harder to repeal than a duty to consider. Frontier Whether any future-generations body publishes a count of its own interventions and their outcomes is the single variable most worth watching, and it costs nothing. Frontier Whether any jurisdiction entrenches a flow rather than a stock would be the first genuine design innovation in this field since 1976.
• 50 yr: Speculative Repository closure and marker construction begin somewhere, converting the message-preservation literature from design study into practice for the first time. Speculative Kiribati's fund passes its centenary, which would make it the longest-lived sovereign fund by a margin no purpose-built institution has approached. Handwave And on the constitutional base rate — a mean life expectancy of 19 years and a post-1974 median age of about 12 — the expected number of constitutional replacements worldwide over fifty years runs into the dozens, which is a statement about the reference class rather than a forecast about any country.
• 100 / 250+ yr: Handwave Beyond useful forecasting. The two longest-running commitments in the record — a 1648 perpetual bond and an eighteenth-century consol — both had their terms rewritten by the obligor, so even the observed three-century cases do not license a forecast. Handwave And the panels said it themselves: the evolution of existing cultures and the creation of new ones over ten thousand years cannot be known, and passive controls can never be assumed to eliminate intrusion.

• Depends on Nothing on this map. No claim in this brief waits on a result produced by another; the constraints are statutory, fiscal and judicial, and they are recorded below. This is one of the few slots in the corpus where the honest answer to “what is blocking progress” is that nothing is blocking it except the absence of anyone wanting to be bound.
• Requires (not on this map) An instrument a successor legislature cannot rewrite by ordinary majority — Alberta's deposit rule was cut in 1983 and abandoned in 1987 without a crisis. An enforcement mechanism rather than a duty to consider, since the audit finding is exactly that consideration without consequence produces prominence without change. Entrenchment placed on the layer people actually fight over: Alaska constitutionalised a principal nobody wanted to spend and left the spendable account to a simple majority, with a 54% chance of exhausting it within a decade. Published counts of interventions and outcomes, absent from every future-generations body examined, which is why the field's central question remains unanswerable in principle. And a constituency with a financial stake, which is what makes pension promises the most durable long-horizon commitment any government actually keeps despite never having been designed as one. All five are institutional or fiscal; none is a research result.
• Enables Decisions taken now that remain defensible to people who cannot object to them. No typed enabling edge is claimed: the briefs that assume long-horizon capacity assume it generically rather than depending on any particular instrument described here, and on this evidence they should assume less of it.
• Adjacent Civilizational Planning, the sibling slot, which owns the plans while this owns the machinery; Nuclear Waste Solutions, which supplies the extreme case; Future Democracies, which shares the finding that these bodies die of politics rather than of failure; Global Cooperation Models, where the same commitment problem runs between states; and outside the map, comparative constitutional law, public pension actuarial practice, and financial history.

“The standard endurance research shows that harder-to-amend constitutions last longer.” Established It does not, and its own authors have since said the relationship cannot be reliably measured. The famous 19 years is an estimate of mean life expectancy, not the observed average age of living constitutions, and it is quoted as the latter constantly. The follow-on work finds seven published indices of amendment difficulty correlating below 0.5 with one another, one negatively, calls this a validity problem with the measures, finds the procedural variables statistically insignificant, and concludes that “constitutional designers have little influence over the observed flexibility of their product.” What does have a clean measured relationship with survival is process inclusion: a 54% lower replacement hazard, and a median durability of 16 years against 10. Public participation — the thing reform campaigns actually demand — shows no significant relationship. This brief also records that an unrefereed preprint using different methods finds rigidity does reliably reduce amendment rates, and does not resolve the disagreement.

“Eternity clauses put certain principles permanently beyond politics.” Established They are common and they are not the constraint. 42% of post-war constitutions adopted to 2011, or 82 of 192, contain some type of eternity clause, and 32% of constitutions of all time, or 172 of 537. When they were tested, they were mostly not used: in 40 documented attempts to remove presidential term limits from 2000 to 2020, courts invoked an available unamendability rule in 2 cases out of 31, while 16 attempts were stopped by ordinary political opposition. Honduras is the clean refutation: its constitution made the term limit of Article 374 unamendable, and in 2015 its own Constitutional Court declared the unamendable provision unconstitutional. And the rules are associated with more wholesale constitutional replacement, not less — an executive who cannot amend writes a new constitution instead.

“Norway's fund is the world's exemplary long-horizon commitment device, and Alaska is the other one.” Established Both are important and neither is the oldest. Kiribati's Revenue Equalization Reserve Fund was founded in 1956, thirty-four years before Norway's, with no constitutional entrenchment and no celebrated governance framework. It was drawn on to cover budget shortfalls after 2008 and caught in the failure of Icelandic banks, falling from 420% to 350% of GDP between 2007 and 2009, and it stood at $1.523 billion in 2024, its largest ever. It is almost never in the comparison set, and the design literature has not explained it.

“Alberta's Heritage Fund was raided.” Established It was not, and the truth is worse for the framing. There was no emergency appropriation, no seizure and no default. A legislature cut a statutory deposit rate from 30% to 15% in 1983, diverted the fund's investment income to budgetary revenue from the same year, and ended deposits entirely in 1987 — each step by ordinary majority, in fair weather. By 1993 the diverted income of $15.4 billion exceeded twice the fund's then value of $15.3 billion. A device that fails without anyone attacking it is a more damaging finding than one that fails under assault, because there is no adversary to design against.

“Alaska proves constitutional protection works.” Established It proves constitutional protection protects exactly what it names. The Principal, protected since 1976 and fed by a constitutional minimum of 25% of mineral royalties, is intact at $87,898,600,000. Everything left to statute — the 5.0% draw limit adopted in 2018, the 50%-of-distributable-income dividend formula — has been overridden or duplicated, and the state's own analysts put a 54% probability on exhausting the spendable account within a decade. The entrenchment did not fail. It was placed on the layer nobody was fighting over.

“Sunset clauses make bad laws expire.” Established On the record they make laws pass. Approximately 13% of bills carrying a sunset provision passed the United States Congress against 5% of permanent legislation — the sunset roughly doubles the chance of enactment. Expiry is rare: on one international survey, the expiry of anti-terrorism legislation is “extremely rare,” and the 16 sections of the USA PATRIOT Act originally set to lapse on 31 December 2005 were “reauthorised several times in the following years after only very limited evaluation.” Renewal scrutiny is thin by measurement rather than by reputation: four peers spoke in one 2003 House of Lords renewal debate, one of them the minister, and a 2011 debate lasted 32 minutes. And when the compromise theory was tested directly on 1,639 randomised subjects, sunsets did not increase overall support for laws at all.

“Sunset laws have been shown to deliver large economic gains.” Handwave A 2025 peer-reviewed analysis of United States states from 1963 to 2022 reports agency-based sunsets raising real GDP per capita by $14,000, a 59.8% proportional gain. This brief reports the estimate and does not endorse it: a fifty-nine per cent increase in state output from a review calendar is not a credible effect size, and the figure should be read as evidence that the identification is fragile. The usable finding from the same paper is its description of the mechanism — that agencies “seem to be renewed from the political inertia that they were meant to overcome,” with gains front-loaded and attenuating as agencies adapt to review.

“Centuries-old bonds show that some obligations really are permanent.” Established They show the opposite once you read the terms. The 1648 Dutch water-board bond has paid for 378 years, and its coupon was cut by the obligor from 5% to 3.5% and then to 2.5%, without the creditor's consent. The United Kingdom's undated gilts, including a 2.5% Consolidated Loan issued in July 1729 with £162.1 million outstanding when it was redeemed on 5 July 2015, and 2.5% Annuities tracing to South Sea stock of 1726 and 1751, were all called in between February and July 2015 because low rates made refinancing cheap. Neither obligation was repudiated and neither was enforced. Both were rewritten at the debtor's convenience, which is the general case.

“Old institutions are durable institutions.” Established Not in either class where survival has been modelled. A firm's mortality rate “had nothing to do with how long it had already been in business or what kinds of products it produced,” which is an age-independent hazard and therefore no accumulated immunity. The most spectacular counterexample available: Kongō Gumi, a temple-construction firm founded in 578, survived roughly 1,428 years and was absorbed in January 2006 into a wholly owned subsidiary of another construction group, the original entity retaining only real estate and being renamed. It fell to debt and a takeover, not to catastrophe. Fourteen centuries bought it nothing at the end.

“The ten-thousand-year marker problem is the extreme case of the commitment problem.” Established It is the extreme case of the communication problem, and running the two together has muddled the subject for decades. Every question the marker literature asks is whether a message survives its senders. None of it addresses whether an institution can be held to a promise, and there is no commitment analogue at that horizon in any literature consulted. Within the marker problem itself, the persistence and comprehension figures are also routinely inverted: physical survival is 0.40 to 0.85 and correct interpretation runs as low as 0.01, so treating “will the marker last” as the hard question gets the finding backwards.

“The evidence shows future-generations commissioners have no impact.” Frontier The evidence shows the impact was never measured, and the distinction matters. Hungary's ombudsman — holding the strongest formal instrument in the field, a route to the Constitutional Court via the senior Commissioner — publishes no staff size, no case counts, no referral counts and no outcomes in its own account of itself, and the peer-reviewed literature describing it supplies none either. No documented case in which any such body changed a decision was located anywhere for this brief. That is a missing denominator, not a measured zero, and reading it as the latter is the mirror image of the advocacy claim.

“The future-generations Act stopped a motorway.” Handwave Carried forward, and still the clearest causal over-claim in the field: the inquiry inspector recommended approval, and the scheme died after a change of First Minister and a fiscal judgement about cost. The Act shaped the vocabulary and the option set, which is a real contribution and a different claim. Established And the Act's own auditor says it is not driving system-wide change, ten years in, with some public bodies giving it little or no explicit consideration.