# Broad Research Round — Append-Only Ledger

Charter: Amendment 2 (hash 15d5109c). Every question, dataset, and analysis examined across the three cycles, in order. Nothing dropped.

- **C1** [precommit] (A1) Among credentialed alternative experience pathways to the right seat — R-ATP reductions (military 750 / 4-yr-degree 1,000 / 2-yr-degree 1,250 hours per 14 CFR 61.160), the 2024 FAA Reauthorization Enhanced Qualification Program, and structured-academy / AQP / approved simulator-credit hours — is there an identifiable basis (existing published evaluations, FAA/NTSB/GAO/academic studies, or already-collected program data) to characterize whether ANY such design plausibly dominates the raw 1,500-total-GA-hour instrument on a competence-per-hour or safety-relevant-experience basis? This is a NEW design-comparison estimand, NOT a re-estimate of the locked net-lives sign. | (A2) Can the COMPOSITION of experience (structured/scenario/simulator/AQP hours vs raw total flight time) be distinguished as a candidate mechanism in the available evidence, or is the question a data-power failure (no design-level outcome data linking pathway type to Part 121 safety outcomes)? Report 'not estimable / data-power failure' explicitly if the data cannot separate composition from quantity. | (A3) Is there a registry/credential trail (e.g., Pilot Records Database fields, certificate-action records, ATP-CTP completion) sufficient to even in principle attribute right-seat outcomes to pathway type, or does the linkage not exist in collectible form? Disposition is identifiability, not effect size. | (B1) For the post-2009 Part 121 safety record, do the SEPARATELY-TIMED post-Colgan reforms — the 1,500-hour experience floor (Aug 2013), Part 117 fli​ght-and-duty/fatigue rules (Jan 2014), stall/upset recovery and Part 121 training enhancements, SMS (final rule 2015/2018 phase-in), and the Pilot Records Database (PRD, phased 2021+) — have enough independent timing variation to even partition credit ACROSS reforms, or are they too co-timed/confounded to attribute share? This is a NEW attribution-share estimand over the reform PACKAGE, NOT a recomputation of the experience-floor's net sign. | (B2) What is the denominator problem: with so few post-2009 Part 121 fatal/hull-loss events, does the event count support ANY multi-reform attribution partition at conventional identifiability, or is this a data-power failure to be reported as such (not as a measured null that the floor 'did nothing' or 'did everything')? | (B3) Are there pre-registered, design-based natural-experiment angles (staggered effective dates, cross-operator SMS adoption timing, GA-vs-121 differential exposure to the fatigue rule) that a FUTURE confirmatory round could exploit — i.e., which candidates clear the disposition bar for pre-registration vs which are dead ends?
- **C1** [research] 14 CFR 61.160 R-ATP credit schedule (ALREADY law): military 750 hr; bachelor's aviation major +60 credit hr 1,000 hr; associate's/30+ credit hr 1,250 hr; all with 200 hr cross-country (vs 500 unrestricted) and flight training inside an FAA-approved Part 141 curriculum. So the rule already embeds structured-pathway hour credit; the live question is the SLOPE of that credit schedule. | 2024 FAA Reauthorization Act (Pub. L. 118-63, signed May 16 2024) statutorily created the Enhanced Qualification Program (EQP): an airline/training-center R-ATP pathway with a nationally standardized curriculum (aircraft automation, SOPs, CRM, checklists, simulator scenarios, aircraft-specific instruction) + entry assessment; FAA to issue requirements by Nov 16 2024. | GAO-26-107856 (Feb 2026): as of Feb 2026 FAA had NOT publicly issued EQP requirements (missed the Nov 2024 statutory deadline); GAO recommends FAA set/communicate a public timeline. Report provides NO safety-outcome analysis comparing training pathways.
- **C1** [research] EXPLORATORY-ONLY per Charter Amendment 2 (hash 15d5109c). Locked net-ledger headline restated verbatim and unchanged: 'sign indeterminate, magnitude small, modest adverse lean.' Nothing here re-estimates or re-signs the locked H_A/H_B result. | FAA's own 2013 regulatory evaluation (docket FAA-2010-0100-1925) stated: 'the FAA was unable to find a quantifiable relationship between the 1,500-hour requirement and airplane accidents and hence no benefit from the requirement.' The agency quantified NO marginal accident benefit for the raw hour floor specifically. | The 1,500-hour TOTAL-TIME floor was a CONGRESSIONAL STATUTORY MANDATE (Public Law 111-216 / H.R. 5900: ATP 'shall be at least 1,500 flight hours'), NOT an NTSB recommendation and NOT an FAA cost-benefit-derived number. NTSB's Colgan report (AAR-10/01) recommended training, fatigue, professionalism, records, and stall-recovery fixes — not a total-time floor.
- **C1** [redteam] verdict=needs caveats; 4 paths
- **C2** [precommit] A1 (ICAO MPL natural experiment, rate-denominated). Among jurisdictions that adopted the ICAO Multi-crew Pilot Licence (MPL) competency pathway (e.g., Lufthansa/EASA early adopters plus non-European adopters in Asia-Pacific/Gulf), is there a pathway-stratified accident or serious-incident RATE (events per departure, per flight hour, or per pilot-year) that differs from the traditional CPL/ATPL hours-based pathway? GATE (§A1.5): if no pathway/experience-stratified RATE denominator (departures, hours, or pilot-years by pathway) is obtainable, report 'not estimable / data-power failure' — NOT a measured null and NOT evidence the pathways are equivalent. | A2 (Australia/CASA frozen-ATPL-style and Part 121/135 regime). Does the Australian air-transport accident RATE, denominated by departures or hours and stratified by crew experience/licence pathway, provide an out-of-Europe replication of the hours-vs-competency contrast? GATE: absent an experience-stratified RATE denominator from ATSB/CASA/BITRE movement data, report 'not estimable'; do NOT substitute a GA accident rate for the air-transport estimand (no GA-rate substitution). | A3 (Canada/TCCA pathway). Does Canadian commercial air-transport accident/incident RATE (TSB events over Transport Canada / movement-statistics denominators), stratified by licence pathway or crew experience, replicate or contradict the European hours-threshold signal? GATE: absent a pathway/experience-stratified RATE denominator, report 'not estimable'. | A4 (cross-jurisdiction denominator audit, pre-data). For each of A1–A3, is an experience- or pathway-stratified exposure denominator even PUBLISHED at the granularity the estimand requires (not merely a national fatal-accident count over total departures)? This is a feasibility/identifiability screen answered before any rate is computed; a 'no' here routes the corresponding question to 'not estimable' rather than to estimation. | B1 (fatigue as candidate active ingredient). Is there a published, exposure-denominated association between a fatigue proxy (duty/rest regime, time-on-task, circadian window, FRMS adoption) and air-transport accident/incident rate that is identifiable independently of total flight hours — i.e., does fatigue carry signal that raw hours does not proxy? GATE: if fatigue exposure cannot be denominated against an at-risk base, report 'not estimable'. | B2 (automation dependency / manual-flying skill decay). Is there an identifiable estimand linking manual-flying skill decay or automation dependency (e.g., autoflight-on exposure, recency of manual handling, loss-of-control-in-flight event class) to accident/incident rate, separable from cumulative flight hours? GATE: absent a measure of manual-handling exposure or recency denominated against an at-risk base, report 'not estimable'. | B3 (CRM / multi-crew exposure). Is there an identifiable, exposure-denominated association between CRM training intensity or accumulated multi-crew exposure and accident/incident rate that is distinct from total hours — does multi-crew exposure proxy something hours alone does not? GATE: absent a multi-crew-exposure or CRM-dosage denominator, report 'not estimable'. | B4 (discriminating test, pre-registered direction). Across B1–B3, is there ANY dataset permitting a within-study contrast that holds total flight hours fixed while varying the candidate ingredient (fatigue / manual-skill recency / multi-crew exposure)? If yes, pre-register that the ingredient is a credible 'active ingredient' candidate only when it separates from hours under that contrast; if no such hours-fixed contrast exists, all of B is reported 'not estimable' and NOT as evidence that hours is the right proxy. | M1 (multiplicity / non-repackaging guard). Across all of A1–B4, how many independent looks does this cycle add, and does any favorable finding meet the SAME identifiability and denominator bar as an unfavorable one? Pre-register that no cumulative exploratory finding from C1 or C2 may be recombined into a confirmatory claim — only a fresh future pre-registration can.
- **C2** [research] US Public Law 111-216 / FAA 2013 First Officer Qualification rule requires an ATP (generally 1,500h; R-ATP reductions to 1,000h with 4-yr aviation degree, 1,250h 2-yr, 750h military) to serve as Part 121 SIC; the rest of the world's airline-entry floor is roughly 10x lower. | EASA/ICAO frozen-ATPL pathway seats a new first officer at ~200-250 flight hours; the EASA July 2020 evaluation states the integrated ATPL minimum has been reduced to as low as ~195 flight hours. | ICAO Multi-crew Pilot Licence (MPL, Annex 1, adopted 2006) is competency/simulator-based with very limited real-aircraft and solo time; ~4,018 graduates worldwide by March 2021 (1,358 Lufthansa group), ~5,700 students and ~39 airline ATO programs by end-2018; Germany held 65% of EASA MPLs in 2018; MPL widespread in Asia.
- **C2** [research] Casner et al. (2014, Human Factors 56(8), B747-400 sim, n=16, mean ~17,844 career hours) found a dissociation: psychomotor skills (scan, control inputs) were 'surprisingly resistant to forgetting' but automation-supplanted COGNITIVE skills decayed — with FMS/map off, 44% missed the missed-approach point and 38% flew the wrong missed-approach heading; deficits tracked recency, NOT career-hour totals. This is a within-study, hours-separable contrast. | Ebbatson et al. (2010, Ergonomics) found significant associations between pilots' VERY RECENT flying experience and manual handling/control strategy — recency, not accumulated total hours, indexed handling skill. | AF447 (BEA Final Report, 5 Jul 2012) is canonical ANALOGY not a rate: alternate-law hand-flying at altitude the crew was never trained for, unrecognized deep stall, startle that scripted sim scenarios couldn't reproduce, and a CRM breakdown (dual opposing sidestick inputs, intentions never shared). Both pilots were experienced (PNF ~6,500h; relief captain ~11,000h) — high hours did not supply the missing competency.
- **C2** [redteam] verdict=needs caveats; 6 paths
- **C3** [precommit] A1 (entry-cost barrier magnitude). What is the incremental private cost of building from ~250 hr (commercial-certificate baseline) to the 1,500-hr R-ATP/ATP threshold under each pathway (open-market time-building, CFI hour-building, restricted-ATP degree credits at 1,000/1,250 hr)? Estimand: distribution (median, IQR, upper tail) of out-of-pocket dollars plus forgone/low wages over the time-building stage, expressed as a barrier in dollars and in months-to-threshold. EXPLORATORY: descriptive cost-of-entry only; does NOT touch the locked four cost dials or recompute the net ledger. | A2 (who is screened out — socioeconomic). Is there an identifiable association between the post-2013 entry-cost barrier and the socioeconomic composition of the entrant pipeline (parental income, debt load, need for paid employment during training)? Estimand: descriptive composition shift and an explicitly-labeled association, NOT a causal screen-out estimate. Pre-specify that if the available data cannot separate the rule from concurrent shocks (tuition inflation, 2013-19 hiring cycle, COVID), the finding is reported as 'not estimable / confounded,' never as a measured null or a measured effect. | A3 (who is screened out — demographic/diversity). What is the demographic composition (sex, race/ethnicity, first-generation status) of the certificated-pilot pipeline at the relevant rungs (student -> commercial -> CFI -> R-ATP/ATP), and how has the funnel's conditional retention at the time-building rung moved across the rule transition? Estimand: descriptive funnel composition and stage-conditional retention ratios; diversity implication framed as access-incidence, NOT as a causal diversity effect of the rule. | A4 (unpaid/low-paid stage incidence). How long and how lightly-compensated is the post-commercial, pre-airline time-building stage, and on whom does that unpaid/low-paid duration fall disproportionately? Estimand: descriptive duration and effective-wage distribution of the time-building stage, cross-tabulated by available SES/demographic strata; an equity-incidence estimand (who bears the time/income cost), held to the same identifiability bar regardless of whether the lean is favorable or unfavorable to the rule. | B1 (implied cost per statistical life — ledger-conditioned). Conditioning on the LOCKED net-lives result as a FIXED input (cited, never recomputed), what is the rule's implied cost per statistical life saved, and what does the locked 'sign indeterminate' verdict do to that ratio? Estimand: implied $/VSL computed only as societal-cost numerator over the FIXED locked lives denominator; pre-specify that because the locked sign is indeterminate the cost-effectiveness ratio is itself sign-indeterminate / possibly undefined (division by a near-zero or wrong-signed denominator) and MUST be reported as 'not a well-posed point estimate' rather than a favorable or unfavorable number. | B2 (societal cost numerator scope). What belongs in the societal-cost numerator for a cost-effectiveness framing, separate from the locked private entry-cost (A1)? Estimand: an itemized, bounded cost ledger (training-system deadweight, pilot-supply/wage and regional-route effects, foregone-output of the extended stage) with ranges, NOT a single point; explicitly flagged as a NEW normative-economic estimand that does not re-run, re-weight, or re-label any locked parameter or the displaced-air / road-substitution chain. | B3 (comparator menu — alternative safety investments). What is the published cost-per-statistical-life of a pre-specified menu of alternative aviation-safety investments (e.g., pilot fatigue/rest-rule changes, simulator-based upset/stall recovery training, FOQA/ASAP expansion, ground-proximity and runway-safety tech)? Estimand: a comparator table of externally-sourced $/life figures used ONLY as benchmarks; the rule is positioned against them qualitatively given B1's ill-posedness, with no back-solving of the locked lives number from any comparator. | B4 (QALY/age-weighting sensitivity, exploratory). Under a QALY rather than per-life framing, how sensitive is any cost-effectiveness statement to age-of-victim and discounting assumptions? Estimand: a sensitivity sketch (not a preferred estimate) showing the direction-only movement of the ill-posed ratio under standard QALY/discount conventions; pre-specify it cannot rescue a well-posed ratio while the locked denominator's sign is indeterminate.
- **C3** [research] LOCKED headline restated verbatim: 'sign indeterminate, magnitude small, modest adverse lean.' This memo is EXPLORATORY and does not alter it. | Total cost to airline-qualification commonly cited at $80,000-$100,000+; pure-rental time-building upper tail $150k-$300k (one model: 1,500hr at $129/hr = $193,500). Sources: airlinegeeks, summitflightacademy, usaviationacademy, pilotinstitute. | Modal (CFI) path: near-zero out-of-pocket for the time-building increment because instructors are PAID while building, over ~18-24 months; the cost converts to ~2 years of FORGONE/LOW income. CFI pay ~$22-$30/hr avg and paid only for logged flight hours, not ground/prep. Sources: atpflightschool, airplaneacademy, payscale.
- **C3** [research] DOT/FAA VSL for 2024 analyses = $13.2 million in 2023 dollars; uncertainty range $7.9M (low) to $18.5M (high); income elasticity 0.055; discount at 3% and 7%. Prior anchors: 2022 $12.5M, 2021 $11.8M, 2020 $11.6M. Source: FAA benefit-cost 'Treatment of Values of Life and Injury' values doc, per OST 2021 guidance + annual updates. | FAA Final Rule 78 FR 42324 (Jul 15, 2013) scores the statutory 1,500-hour ATP-certificate requirement at $6,374.4M nominal / $2,213.0M present value (7%, 10yr) and assigns it VERBATIM 'No Identifiable Accident Benefits'; FAA prose: '$6.4 billion ($2.2 billion in present value), almost all of which stems from the 1,500-hour flight time requirement.' | The eligibility/hour-threshold cost line alone = $6,344.5M / $2,181.9M PV. The only safety-justified, cost-beneficial pieces of the 2013 rule were the DISCRETIONARY add-ons (ATP CTP + type rating): cost $312.7M/$138.7M PV vs benefits $576.8M/$251.7M PV - NOT the hour threshold itself.
- **C3** [redteam] verdict=needs caveats; 6 paths

## Red-team verdicts
- C1: needs caveats
- C2: needs caveats
- C3: needs caveats