Regulatory Form and the Cost of Information — The Aviation–Banking Contrast: Frozen Plan and Findings

**The claim under test.** Regulatory FORM — which information channels a regulator admits,demands, or authors — is determined by the domain's information-cost primitives, not by theregulator's identity or preferences. The comparative static: where the mechanism is knownphysics and the validating experiment is commandable, regulation should demand themechanism channel; where the state is socially constituted and the tail scores slowly,regulation should tolerate fitted and authored channels — banking's observed form. **Design.** Four cells with frozen predictions and falsifiers: aviation structuralcertification (C-A), aviation system safety (C-B, the predicted mixed cell), aviationdelegated self-reporting (C-C — the positive control, sharing banking's primitive ofmanufacturer-supplied analysis under commercial incentive), and banking (C-D, imported fromthe parent study). A regulator-identity control (two agencies, one architecture) guardsagainst agency-idiosyncrasy accounts. **Headline result** (all graded cells at the study's top grade; falsifiers F1–F5 alluntriggered): **one admission meta-rule, evaluated at two cost regimes, generates the twoopposite architectures.** The transport-airplane airworthiness standard states thecriterion verbatim — "Structural analysis may be used only if the structure conforms tothat for which experience has shown this method to be reliable" — the same criterion thebanking record contains in its 1999 refusal of portfolio credit models (validationinfeasible within any practical sample). Aviation can command the validating experiment(mandatory flight tests; the regulator may order any test), so even mechanism models aregated and pass; banking's tail scores once per credit cycle, so the same criterion shutsthe model channel and leaves the fitted-and-authored form. Convergence both ways: whereaviation's mechanism knowledge runs out (component failure behaviour), its form convergesto banking's — authored probability thresholds over mechanism-structured failure analyseswith fitted inputs; and where aviation shares banking's delegated self-reporting primitive,the official post-mortems of the 2018–19 accidents document banking's pathology (certifiedassumptions carried over without validation, under documented pressure, disconfirmed byrealized outcomes) and the 2020 certification-reform statute is banking's repair inverification shape (regulator approval of every delegated engineer; regulator validation ofload-bearing assumptions before delegation; mandatory disclosure of safety-criticalchanges). The identity control passed: the EU and US structural-proof rules are materiallyidentical in wording. **Honesty flags preserved and machine-checked by the included ledger:** two nearest-missdisjunctive service-history channels in the airworthiness standard are logged as boundarycases, and their fast-scoring placement — though consistent with the frontier reading —was not frozen ex ante and is carried as an observation only; the current system-safetyguidance is the 2024 revision (cancelling the 1988 circular), with a renumbering caveat;the House committee report is a Majority Staff product and itself names "regulatorycapture" — the capture rival is thereby in the primary record and is bounded, not denied(it predicts the grant and the laxity, not the specific informational failure mode nor theverification-shaped repair); the severity rival predicts stringency, not form; the foreignfinal accident reports, which list crew actions among contributing factors, were notverified here and are flagged; crew performance appears in the verified US post-mortems asthe disconfirmation of the certification assumptions, not as a competing primary cause. **Registration status, stated plainly.** The plan was written before source contact withinthe same working session; no public timestamp predates the study. This deposit creates thepublic record, as the parent deposit did for the banking study. **Relation to the manuscript.** The contrast is reported in the manuscript's onlineappendix as Applications Atlas entry AT.18 ("One admission rule, two architectures") — aclassification register only, with all institutions generic. No proposition of the paperdepends on this study; this deposit is the evidence record behind the entry's factualcontent, and carries the exact citations the entry deliberately omits (14 CFR 25.301/25.303/25.305/25.307/25.571/25.1309 and 21.33/21.35, current text via the eCFR VersionerAPI at title issue date 2026-08-27; FAA AC 25.1309-1B of 2024-08-30; EASA CS-25 Amendment28; the JATR report of 2019-10-11 with finding numbers; NTSB ASR-19-01; the HouseTransportation & Infrastructure Final Committee Report of September 2020; Public Law116-260, Division V).