Non-Compilation

I. The disturbing observation

Most major public-policy flame wars are decades old. Fertility, taxation, immigration, education, healthcare costs, climate. Each has serious academic literature behind it — parameterised, peer-reviewed, internally rigorous. Almost none of it reaches the room where the public argument actually happens.

The fertility debate in the OECD is the cleanest specimen. Demographers have compiled fertility dynamics since the 1970s. John Bongaarts laid out the proximate-determinants framework in 1978. Wolfgang Lutz at IIASA built multi-state population projections through education and human capital. Peter McDonald wrote the gender-equity theory. Gøsta Esping-Andersen mapped the welfare regimes that shape family formation. The literature is vast and broadly accurate.

Public fertility discourse, meanwhile, runs in a tight loop of single-lever slogans: cash transfers work / cash transfers are wasted / people just don't want children / culture is the problem / housing is the problem / immigration solves it. Each claim is partial. Each could be a candidate input. Almost none of them answer the question the academic literature already knows how to ask: which input is binding for the marginal missing child, in this population, at this time?

The Laffer-curve debate runs through the same surface and a different underlying mechanism. Saez, Slemrod and Giertz published the definitive 2012 review parameterising tax response into elasticity of taxable income, sufficient-statistic conditions, base definition, avoidance channels, time horizons. The Congressional Budget Office and the Joint Committee on Taxation produce estimates and dynamic analyses; the Tax Policy Center and Penn Wharton Budget Model produce public primers and simulations. Public tax discourse continues to oscillate between "tax cuts pay for themselves" and "Laffer is fake" — for reasons very different from fertility's. In fertility the public-usable frame doesn't exist; in tax it exists and doesn't land.

The pattern is real and widespread. The bewilderment is genuine: in a world with this much research capacity, this much funded institutional infrastructure, this many smart people, why do these arguments stay this stupid for this long?

The honest answer is not one mechanism but two. The first failure is non-compilation: the public-usable decision frame has not been assembled. The second is non-adoption: the frame exists but is rejected, captured, or sloganised. These are different failure modes; they affect different domains; they have different repairs. This essay names the first. Its sibling Calculemus names the second.

II. Three objects that get confused

To see either failure clearly, three objects have to be disentangled.

Academic synthesis. Integrated scholarly work that ties together disparate findings within a discipline. Bongaarts on fertility. Esping-Andersen on welfare regimes. Saez and colleagues on tax-response parameters. Pielke on the science-policy interface. These exist. They are dense, technical, written for peer audiences, organised around academic vocabulary, almost entirely illegible to the intelligent generalist. They do not transmit into public reasoning. That is not their job.

Institutional compilation outputs. What dedicated organisations produce when they translate academic synthesis into legible numbers and thresholds. CBO scores. NICE QALY thresholds. IPCC synthesis reports. OECD country chapters. JCT distributional tables. These exist. They are produced precisely to be policy-actionable, and for the engaged staffer class they are genuinely usable.

Public-usable decision frame. A public-facing artifact that explicitly parameterises the decision question into a reusable form. A generalist picks it up, holds it, and uses it to evaluate a new policy claim they have not seen before. The artifact's job is to install a question-shape, not to win an argument.

The critical move: define this third object by its form, not its reception. A public-usable frame exists if a reader can inspect it as an artifact with named variables, causal structure, an update rule, and a policy-evaluation test. Whether the public adopts it is a separate question — and the answer to that question, where the artifact exists, is mostly the territory of Calculemus, not this essay.

With the three objects distinguished, the diagnosis becomes precise:

• For some domains, the public-usable frame doesn't exist at all. The academic synthesis is there; the institutional output may or may not be; the publicly-pickable artifact is missing. This is non-compilation.
• For other domains, the frame exists for the staffer class but doesn't transmit to mass discourse. The TPC Tax Policy Briefing Book is real; the McKinsey GHG abatement cost curve is real; the YIMBY supply-side housing frame is real. These are public-usable in form, but reach only the engaged 5%. This is non-transmission.
• For still other domains, the frame exists publicly and is actively captured, weaponised, or sloganised. The Laffer Curve is a public-usable frame; in mass discourse it degraded into "tax cuts pay for themselves." This is non-adoption / capture.

Non-Compilation focuses on the first case. The fertility specimen lives there. The taxation specimen lives in the second and third cases and serves as a counter-example — the case that forces the distinction.

III. Fertility: the clean non-compilation case

The compiled fertility frame would say something like this:

Fertility is a multi-gate pipeline with candidate inputs — meeting, pair stability, reproductive window, household feasibility, cultural permission, felt future, biological execution. A policy moves births only if it relaxes a gate that is binding for the marginal missing child in the target population.

The gates are not strictly sequential. Pair stability affects household feasibility; household feasibility affects pair stability; cultural permission affects reproductive timing. The point is not that fertility is a machine. The point is that policy discourse currently treats candidate inputs as rival slogans when they are interacting constraints, and a compiled frame installs the disposition to ask which constraint is binding rather than which lever to pull.

I have found no widely propagated public-usable artifact that states the fertility question in this form.

The academic substrate is rich. Bongaarts' proximate-determinants framework decomposes fertility into biological vectors — marriage exposure, contraception, abortion, postpartum infecundability. Lutz models population dynamics through education and human capital. McDonald and Esping-Andersen mapped the social-institutional factors. The pieces are extensively studied.

Specific binding-input findings are also available. Lyman Stone argued in 2019 that increasingly postponed marriage could account for at least half of the recent increase in the US fertility gap and "virtually 100%" of the increase since 2000; his 2023 update revised the marriage-postponement share to roughly three-quarters of the fertility decline for the post-2007 period using updated ACS data. The headline magnitude moved; the binding-constraint identification did not. Within married couples, fertility hasn't changed much. The collapse is in the rate of marriage itself. In compiled terms, that is a binding-constraint identification: the gate that has closed in the US since 2000 is upstream, in pair formation. If pair stability is the binding constraint for the marginal missing child, downstream subsidies — cash transfers, childcare expansion, parental-leave extension — may redistribute income or move some already-paired births, but they will not solve the binding fertility constraint.

The closest existing approximation to a compiled fertility frame is Anna Rotkirch's work for the Finnish government. The 2021 Prime Minister's Office report Recovery of the birth rate and longer life expectancy is state-commissioned, multi-factor, and unusually explicit about first births: it reports that around 75% of Finland's fertility decline in the 2010s was due to fewer first children being born. The 2024 rapporteur work moves further upstream, with explicit "relationship recession" framing and concrete proposals on timing of children, fertility awareness, everyday-life management, couples, and family services.

This is among the clearest state-commissioned approximations I have found. Several other states have produced substantial internal compiled fertility analyses — the South Korean Low Fertility and Aging Society Committee, the Japanese Cabinet Office, the French CNAF, the German BMFSFJ's pre-Elterngeld work — and any honest survey would acknowledge them as further partial-positives. None of them, including Rotkirch, packages as the strict reusable publicly-pickable decision-frame artifact: the multiplicative-pipeline form that a generalist could pick up and apply to evaluate a Korean cash-transfer policy, a Hungarian housing subsidy, or an Italian extended-leave scheme. State-commissioned internal compilations exist; the genre-of-public-artifact does not. That is the precise non-compilation claim.

If the compiled frame existed and had any institutional uptake, the world's fertility-collapse debates — South Korea, Italy, Spain, Japan, China — would be harder to conduct in their current single-lever form. Hungary's massive housing-loan-and-tax-exemption package would be evaluated by asking "which gate does this actually relax, and is that gate binding for the marginal missing Hungarian child?" Instead it gets analysed as either "pronatalist policy that proves cash works" or "pronatalist policy that proves cash doesn't work," depending on the writer's prior. Fertility Requires Margin in this corpus is one such specimen — a seven-input production function with an explicit binding-input diagnostic and a policy-evaluation invariant. Other specimens may exist scattered in places I do not know about. What does not exist is the recognised genre: a form a generalist reader would expect to find, evaluate other proposals against, and demand of any pronatalist intervention. That is the precise non-compilation claim — the form is producible, has been produced in at least one corpus including this one, and has not propagated as a public genre.

IV. Tax: the captured / non-adopted case

Tax is the counter-specimen. It proves that compilation is not enough. It is the case that saves the concept from overreach.

Tax has accumulated public-usable artifacts at varying degrees of compression. The strict parameterised end: the Tax Policy Center's Tax Policy Briefing Book is a publicly-available, generalist-aimed primer that maps the rate / base / avoidance / incidence structure. The Penn Wharton Budget Model provides public tax-policy simulations where users can vary behavioural assumptions. The CBO's How CBO Analyzes the Macroeconomic Effects of Changes in Federal Tax and Spending Policies and Long-Term Budget Outlook are explicit parameterised explainers. The Response Vector in this corpus extends the same parameterised approach cross-domain. The longer-form public-facing end: Ed Kleinbard's We Are Better Than This, Bill Gale's Fiscal Therapy, the public-facing work of Lily Batchelder, Bruce Bartlett, Jason Furman, Greg Mankiw — narrative parameterised translations rather than strict simulator-form artifacts, but they install a question-shape rather than argue a position.

The slogan war persists anyway.

Two things go wrong. First, the artifacts reach the staffer class — the journalists who care, the legislative aides drafting briefing notes, the engaged 5% — but mostly do not transmit into mass discourse. Second, when a frame does reach mass discourse it tends to be captured or hollowed out. The Laffer Curve is the canonical case. Laffer's original insight is real economic content: the revenue–rate relation is not monotonic. The mass debate that followed has, decades later, almost never engaged the parameters that would make the question answerable. It runs instead at the level of disagreement about the very shape of the curve, or — more commonly — at the level of single-parameter isolation arguments. From the left: "headline statutory rate X means there is plenty of headroom." From the right: "we are already past the peak." Each is coherent if the one parameter is held in isolation from the rest. Neither survives composition with the actually-binding reality: effective marginal tax rates that combine income tax, payroll contributions, employer charges, and benefit phase-outs, and which routinely sit twenty to forty points above headline rates; the decomposition of the behavioural response into labour supply, tax avoidance, and compensation bargaining; the elasticity of the specific tax base in question. What matters is not what is true in isolation; what matters is what happens when the parameters compose. The serious literature has answered all of this — the four-channel decomposition in The Response Vector, the elasticity-of-taxable-income parameterisation in Saez–Slemrod–Giertz, the holistic-EMTR work in the optimal-tax tradition — and that work has had effectively zero effect on the public terms of the argument.

This is the territory of Calculemus. When a parameterised frame exists, the question is no longer "what is the question-shape?" but "why does the public refuse to compute the parameters?" The answers there are several: coalitional utility outranks epistemic utility; parameterisation dissolves the moral object factions need to keep intact; mass voters reasonably economise on cognitive load and rely on heuristics, party ID, and elite signalling. The slogan war is a feature of the demand landscape, not a failure of supply.

V. Why public-usable frames are unevenly produced

Why is the form so unevenly produced — abundantly for tax, climate, and housing; barely at all for fertility?

The institutional reward structures point at part of the answer. Academics are mostly rewarded for identifiable causal estimates within tight specifications. Some senior scholars do produce public-usable frames (Kleinbard, Esping-Andersen, McDonald), but the routine path of the discipline rewards the narrow paper, not the synthetic artifact. Ministries are rewarded for administering instruments; they can propose a tax credit, a leave extension, a childcare subsidy, but the institutional bias is toward producing actionable proposals within the ministry's legal toolkit, not toward identifying the cases where the toolkit is the wrong toolkit. Parties are rewarded for coalitional clarity. Journalists are rewarded for conflict and narrative. Moral actors — advocacy organisations, faith institutions, cultural-conservation movements — are rewarded for keeping the moral object intact.

The public-usable-frame function is possible inside these institutions. It is simply not owned by any ordinary institutional role as a routine duty. The function exists, but it is structurally homeless. It depends almost entirely on intellectual entrepreneurs — individuals who take it on outside the standard incentive structures — and on a small set of unusual think tanks (Niskanen, Bruegel, Hamilton Project, Institute for Family Studies) that have explicitly tilted toward translation as part of their mission.

Fertility is harder than tax because the relevant variables are not merely technical. They sit inside sex, pairing, status, gender bargaining, biological time, family obligation, and national continuity. A tax model can ask for elasticity without appearing to inspect the soul of the citizen. A fertility model cannot ask about pair formation without colliding with the most defended parts of private life. The eugenics history makes engineering-style framing of human reproduction visceral. The additive bureaucratic instinct — "apply subsidy to the gate ministries can move" — is institutionally easier than the multiplicative-pipeline framing that says some subsidies will mathematically not move the binding constraint. The cumulative effect is that the same intellectual entrepreneurs who happily produce parameterised frames for tax, housing, and climate do not produce them for fertility.

The amateur reader naturally asks: if institutions don't produce this, why hasn't a single Substack writer, retiree systems-thinker, or motivated graduate student produced the compiled fertility frame? The honest answer is a stack of weak filters that compose into a wall. Most amateurs do not read the source literature. Most who do take a side rather than parameterise — the dominant essay norm is have a thesis, and a compiled frame's thesis ("the question is the wrong shape") violates the essayistic instinct. Most who would parameterise correctly intuit that no audience is waiting for the frame — partisans want ammunition, policymakers want recommendations, journalists want narrative, voters want morality — and write something more recommendation-shaped. Most who do not self-censor write meta-frames for whatever niche they personally care about, and fertility (despite its consequence) sits orphaned because the few writers cognitively equipped to compile it usually choose other topics. Of the compiled frames that do get written, most are filtered out by transmission failure: the artifact has no natural carrier, no genre slot, no compounding citation, and sits where it was written.

VI. Why even good frames do not land

Even where the compiled frame is produced, it usually does not shape mass discourse. Non-Compilation is supply-side: the artifact is missing. Calculemus is demand-side: the artifact exists, but computation is avoided, captured, or masked as values. The demand-side mechanisms are familiar — coalitional utility beats epistemic utility; parameterisation dissolves the moral objects factions need to keep intact; uptake requires institutional teeth (CBO scores are procedurally consequential because bills are argued against them; cost-benefit analysis became standard via OMB Circular A-94; Cochrane-style review became part of evidence-based clinical guidance infrastructure) and where the compiled frame is merely available it competes against slogans and usually loses. The compiled frame existing is necessary but not sufficient. The honest version of the project is dual repair: produce the missing frames where they are absent, and install institutional teeth that make ignoring them costly where they already exist.

VII. How translation-layer genres actually propagate

History gives a precise answer to whether a lone catalyst can install a new genre. Yes for the invention of the form; no for its propagation. The genres that have become standard followed a three-phase pattern.

Phase 1 — epistemic invention. A catalyst or small group invents the question-shape. Archie Cochrane in 1972, in Effectiveness and Efficiency, asked the three questions that defined what became evidence-based medicine: Can it work? Does it work? Is it worth it? The catalyst was identifiable; the substrate around the catalyst (the UK Medical Research Council, post-war public-health institutions) was non-trivial. Cost-benefit analysis was invented at RAND in the 1950s by Charles Hitch, Roland McKean and colleagues as systems analysis adapted to defence procurement.

Phase 2 — boundary infrastructure. The invention needs an organisational broker that standardises the form, trains practitioners, and matchmakes the genre into the hands of decision-makers. Cochrane's vision led two decades later to the Cochrane Centre (1992) and the international Cochrane Collaboration (1993), which codified founding principles, formalised meta-analysis as a methodological standard, and embedded systematic review as part of the infrastructure of evidence-based clinical guidance. J-PAL — founded at MIT in 2003 by Abhijit Banerjee, Esther Duflo, and Sendhil Mullainathan; Banerjee and Duflo, with longtime affiliate Michael Kremer, jointly received the 2019 Nobel — built an evidence-to-policy infrastructure that has helped scale mass deworming programmes to hundreds of millions of children, paired with the MITx MicroMasters program that trains thousands of practitioners. The form propagates only when an organisation owns the standardisation, the training, and the matchmaking.

Phase 3 — institutional capture. The form becomes procedurally consequential for funding, regulation, or policy. RAND's systems analysis propagated across federal government when Robert McNamara, as US Secretary of Defense, imposed it on the Pentagon as the Planning, Programming and Budgeting System; it eventually became standard practice across federal agencies via OMB Circular A-94. Cochrane-style systematic review became part of the infrastructure of evidence-based clinical guidance. The form lasts when failure to use it has consequences.

VIII. The dual repair

The honest position has two pieces.

Where the compiled frame does not exist (fertility being the cleanest case), the repair is to produce it. This is Phase-1 work: invent the question-shape, name the gates, write the artifact, give the form a name, propagate enough examples that the genre becomes legible to other potential producers. A single writer can do this; multiple writers, producing in the same form, make the genre easier to recognise and reproduce. Producing the artifact does not, by itself, pacify the debate it enters — Calculemus's demand-side mechanics are waiting on the other side, and the frame will be metabolised, captured, or moralised away unless institutional teeth are added later. Phase 1 is the prerequisite for the institutional fight, not its conclusion. You cannot demand procedural consequence for an artifact that does not exist; you have to build the artifact so that the fight can be over something concrete.

Where the compiled frame exists but does not shape mass discourse (tax, housing, climate, education being the canonical cases), the repair is not another essay. It is institutional teeth — Phase-2 and Phase-3 work. The compiled frame has to become procedurally consequential for some downstream decision: an editorial standard, a debate-moderator protocol, a regulatory prerequisite, a budgetary gating function. Without that, the frame remains available and the slogan war remains optimal for its actual participants.

These two repairs are different and they apply to different domains. The diagnostic discipline is: for any given flame war, ask whether the public-usable frame exists. If it doesn't, the failure is non-compilation and the repair is to produce it. If it does and isn't adopted, the failure is non-adoption and the repair is institutional. The two are sometimes complementary in a single domain — fertility may eventually need both, the frame produced and institutional teeth installed around it — but they are not the same problem and they do not have the same solution.

Underneath, one structural law keeps surfacing: a system fails where reality enters through a channel the institution does not own. Unowned purposes produce telos gaps. Unowned mechanisms produce moralised causal disputes. Unowned harms produce silent carriers. Unowned decision-frames produce non-compilation. Unowned computation produces Calculemus's slogan wars. Modern democracies have done well enough at owning academic synthesis and funding institutional outputs. They have done badly at owning the translation artifact, and worse at owning the institutional teeth that would make it bind.

This essay is Phase-1 work for one missing-artifact case. It names a missing object, identifies one of its specimens (fertility), and points at the institutional shape — Mechanism Analysis at the pre-enactment stage, the Fourth Branch as the constitutional host — that would have to exist for the missing object to propagate. Fertility Requires Margin is the companion artifact in this corpus: the production-function form instantiated for the fertility case, included as one concrete demonstration of what the missing genre looks like when actually written. The compiled-frame artifacts for the other cases where the publicly-pickable frame is genuinely absent, the institutional teeth for the cases where it exists but does not bind, the bridging organisations that would make the form a genre rather than a sub-cultural curiosity — these remain to be built.

Related:

• Calculemus — the demand-side sibling: why even existing compiled frames do not get computed
• The Response Vector — corpus-internal example of layer-3 work in tax / AI / regulation: the four-channel decomposition (target response / base loss / formal gaming / incidence shifting) the §IV Laffer paragraph references
• The Mechanism Analysis — the pre-enactment artifact form a compiled frame would take
• The Fourth Branch — the constitutional host that would make compiled frames bind
• The Asymmetric Carrier Problem — the structural reason silent-side objects (including missing decision-frames) do not get produced
• The Telos Gap — the upstream condition: no actor owns the purpose
• The Attenuation Layer — the demand-side absorption mechanism that meets the supply-side absence diagnosed here