# Grandfathering as a Natural Experiment: Can the 2013 Discontinuity Measure the Experience Effect? — Feasibility Memo

**Date:** June 2026
**Status:** EXPLORATORY feasibility research. This memo does **not** re-open FRISA's locked verdict on the 1,500-hour rule (*"sign indeterminate, magnitude small, modest adverse lean"*). It evaluates only whether the August 1, 2013 entry-hours discontinuity is a usable *identification* strategy for the experience-vs-safety question, and whether it should be pre-registered as an arm of the existing FOQA precursor study. A design that *could* measure the effect with protected data access is fully consistent with the verdict being **unmeasured-with-public-data** — it reinforces the "testable but untested" theme. **Symmetric honesty throughout: a data/power failure is not a measured null.**

---

## Executive Summary — Bottom-Line Verdict

**Is the grandfather-RD a VALID, POWERFUL, FEASIBLE identification strategy? Verdict: QUALIFIED on validity and power; NO on feasibility with public data — and the cleanest version of the design is partly undercut by a regulatory fact the premise gets wrong.**

Three findings, in order of importance:

1. **A correction to the premise that changes the design — but does not kill it.** The 2013 rule did **not** broadly "grandfather" incumbent low-hour first officers to keep flying *as they were*. The FAA **explicitly rejected** American Eagle's and American Airlines' requests to grandfather incumbent SICs, and required that **every** Part 121 SIC hold an ATP **or a restricted-ATP (R-ATP)** as of August 1, 2013 ([FAA FOQ Final Rule, 78 FR 42324](https://www.federalregister.gov/documents/2013/07/15/2013-16849/pilot-certification-and-qualification-requirements-for-air-carrier-operations); [AOPA regulatory briefing](https://www.aopa.org/-/media/Files/AOPA/Home/Advocacy/faapilotcertificationandqualificationrequirementsforaircarrieroperations.pdf)). The one genuine "grandfather" in the regulation is **narrow**: 14 CFR §121.436(e) exempts pilots **employed as PIC (captain) on July 31, 2013** from the **1,000-hour air-carrier experience** requirement — it is not a low-total-hours first-officer exemption ([14 CFR §121.436(e)](https://www.law.cornell.edu/cfr/text/14/121.436)). **What the principal is really pointing at exists, but as a *hiring-cohort* fact, not a *grandfather* fact:** first officers who **entered** Part 121 pre-2013 at ~500–700 hours (per GAO/Pilot-Source-Study) were not abolished — they **persisted** in the active population and accumulated tenure. The "abolished counterfactual population" was abolished only at the *entry gate going forward*; the pre-rule entrants remain as a low-*entry*-hours cohort. So the comparison the principal wants — pre-rule-entry vs post-rule-entry cohorts, both observed flying after 2013 — is real. But the rule did not produce a *sharp, clean* "kept flying at 600 hours" discontinuity, because incumbents below threshold still had to convert to an ATP/R-ATP, and a transition path was built specifically to avoid "inadvertently disqualifying incumbent SICs" ([N 8900.225 transition guidance](https://www.faa.gov/documentlibrary/media/notice/n_8900.225.pdf)). The discontinuity is in **entry hours**, not in **who was allowed to fly**.

2. **The strongest design is a cohort comparison keyed to hire date, NOT a regression-discontinuity-in-time at the cutoff — and it is confound-bounded, not confound-clean.** An RD-in-time at August 1, 2013 is *weaker* than it first appears, because the cutoff is also the simultaneous onset of the **whole post-Colgan reform bundle** (Part 117 fatigue, stall/upset training, SMS, PRD) and the assignment variable (calendar hire date) is the same variable every secular trend and hiring-boom rides on. The sharp-cutoff logic that makes RD powerful *does not buy you separation from the bundle here*, because the bundle shares the cutoff. The design that survives is a **hire-cohort comparison at equivalent early-career stage** (entry-hours cohort as the treatment variable, observed in a *common* post-2013 operating environment), which neutralizes the bundle by holding the post-period environment fixed for both cohorts — at the cost of the entry-hours signal **attenuating with tenure** and **survivorship** selecting who from the low-hour cohort is still flying.

3. **Feasibility fails on public data for the same reason the FOQA precursor study fails — and one reason that is *worse*.** The design needs event-level precursor/incident records linked to each pilot's **hire date, entry hours, and tenure**. That linkage is not public. The natural national keystone — the **Pilot Records Database (PRD)** — **explicitly excludes flight-time, duty-time, and rest-time records** ([14 CFR Part 111](https://www.ecfr.gov/current/title-14/chapter-I/subchapter-G/part-111)), so PRD carries hire date and training/qualification history but **not entry hours**, the single most load-bearing variable. The design is therefore **carrier-side or protected-ASIAS-only**, the same blocker FRISA already identified — plus an extra entry-hours-provenance gap. **No NTSB, FAA, or academic pre/post or RD *incident/safety* cohort study of the 1,500-hour rule's entry cohorts exists.** The nearest study — the **Pilot Source Study 2018** — is a pre/post-FOQ cohort comparison on **training performance, not incidents**, and it found the counterintuitive result that post-FOQ hires had **more** hours yet needed **more** training and were **less** likely to complete ([Pilot Source Study 2018](https://www.researchgate.net/publication/340507735)). That is the closest anyone has come, and it is on the wrong endpoint.

**Disposition for pre-registration:** Pre-register the grandfather/hire-cohort contrast as a **secondary, exploratory, confirmatory-ineligible arm** of the existing FOQA precursor study — **not** as a standalone confirmatory instrument. It **adds one thing the within-regional FOQA design genuinely lacks: a temporal entry-hours discontinuity** (a pre-rule low-entry-hours cohort that the post-rule design cannot fabricate), which is a partial answer to the within-regional design's fatal aptitude/selection confound (the pre-rule cohort was selected under a *different, lower* screening regime). But it **imports two confounds the within-regional design does not have** (the reform-bundle/era confound and the tenure-decay confound), and it shares the same data-access blocker. **Net: a useful exploratory triangulation arm, not a confound-dominating upgrade, and not executable on public data.**

---

## 1. The Identification Strategy

### 1.0 First, get the institutional fact right (this reshapes every candidate design)

The principal's framing — *"pilots who entered at ~600 hours were grandfathered and kept flying, so the low-hour population was NOT abolished"* — is **half right in a way that matters for the design**:

- **TRUE:** Low-*entry*-hours first officers who entered Part 121 before August 1, 2013 (regionals routinely hired at ~500–700 hours pre-rule per [GAO](https://flightsafety.org/asw-article/tough-competition/) and the [Pilot Source Study](https://www.researchgate.net/publication/340507735)) **persisted** in the active pilot population, accumulated tenure, and many are still flying. FRISA's own "abolished counterfactual population" phrase is therefore correct **only about the forward entry gate**: after the rule, *new* entrants cannot come in below the R-ATP floor, but the *stock* of pre-rule low-entry-hours pilots was not removed. This is the real, exploitable asymmetry.
- **FALSE (the load-bearing correction):** There was **no broad grandfather letting incumbent SICs keep flying without an ATP.** The FAA **expressly rejected** American Eagle's and American Airlines' grandfather requests and required **all** Part 121 SICs to hold an ATP or restricted-ATP by August 1, 2013 ([AOPA briefing](https://www.aopa.org/-/media/Files/AOPA/Home/Advocacy/faapilotcertificationandqualificationrequirementsforaircarrieroperations.pdf); [FOQ Final Rule](https://www.federalregister.gov/documents/2013/07/15/2013-16849/pilot-certification-and-qualification-requirements-for-air-carrier-operations)). The only true "grandfather" in the regulation is the narrow §121.436(e) exemption of **July-31-2013 incumbent captains** from the **1,000-hour air-carrier-experience** sub-requirement — not a low-total-hours FO exemption ([14 CFR §121.436(e)](https://www.law.cornell.edu/cfr/text/14/121.436)).

**Why this matters for design:** the *treatment variable is entry hours at hire, not "grandfather status,"* and the *running variable is hire date, not "kept flying."* There is no sharp regulatory step in *who may fly* at the cutoff (everyone converted to an R-ATP); the step is in *the entry-hours composition of new hires*. That makes the clean RD weaker and the cohort comparison the workhorse. **[VERIFIED against the Final Rule and §121.436(e); the "incumbent FOs kept flying ungrandfathered-but-converted" mechanism is confirmed across the AOPA briefing and N 8900.225.]**

### 1.1 Candidate (a): Regression-discontinuity-in-time at August 1, 2013

**Design.** Take the just-before hire cohort (entered Part 121 in, say, the 12–24 months pre-cutoff, at lower entry hours) and the just-after cohort (entered post-cutoff under the R-ATP/1,500 regime), and compare their **early-career** incident/precursor rates, both observed flying in the **same post-2013 environment**. Assignment variable: calendar hire date. Discontinuity: entry-hours composition jumps at the cutoff.

**Verdict: WEAKEST of the three for *this* question, despite being the most "rigorous-looking."** The defining virtue of RD — that units just on either side of the cutoff are alike *except for treatment* — **fails here on the exclusion restriction**, because the cutoff is not a clean entry-hours instrument. August 1, 2013 is the simultaneous onset of the **entire reform bundle** (§1, point i below): the just-after cohort differs from the just-before cohort not only in entry hours but in having been *hired into and trained under* Part 117 fatigue rules, mandatory stall/upset training, SMS, and PRD-era screening. The RD's running variable (hire date) is **perfectly collinear with the bundle's onset**. An RD-in-time cannot separate "1,500-hour entry made them safer" from "the bundle made them safer" or from "the secular trend continued" — the very inseparability FRISA's Safety Ledger already flagged as the central attribution problem ([ATP-Rule-Safety-Ledger.md](../../whitepapers/ATP-Rule-Safety-Ledger.md), "Background"). A sharp RD does **not** mitigate this; it *inherits* it, because the discontinuity is shared. (RD-in-time also carries the standard hazards: no units exactly *at* the threshold, bandwidth sensitivity, and autocorrelated time-series error — [RD policy-evaluation primer](https://scienceetbiencommun.pressbooks.pub/pubpolevaluation/chapter/the-regression-discontinuity-design/).)

### 1.2 Candidate (b): Historical cohort comparison at equivalent career stage

**Design.** Compare the pre-rule low-entry-hours cohort's incident/precursor rate **during their own early-career window** (e.g. first 300 line-hours, observed pre-2013) against the post-rule cohort's early-career window (observed post-2013).

**Verdict: confound-DOMINATED — strictly worse than (c).** This is the naive pre-vs-post that the Safety Ledger already condemns: it compares two cohorts observed in **different eras**, so entry-hours is fully confounded with the reform bundle *and* the secular trend *and* equipment/automation vintage *and* reporting-culture maturation (FOQA/ASAP program penetration grew over exactly this period). It answers "were early-career pilots safer before or after 2013," which is the bundle's effect, not the entry-hours effect. Reject.

### 1.3 Candidate (c): The grandfathered/pre-rule cohort as a persisting low-entry-hours counterfactual *within* the post-2013 era

**Design.** Within the **common post-2013 operating environment** (same fatigue rules, same SMS, same airspace, same equipment vintage, same reporting culture — the bundle held fixed for everyone), compare pilots **stratified by entry-hours at hire**: the pre-rule low-entry-hours cohort (now mid-tenure) vs the post-rule high-entry-hours cohort, **with tenure/time-since-hire explicitly modeled** so the contrast is read at comparable accumulated experience or as an entry-hours main effect net of tenure.

**Verdict: STRONGEST of the three — this is the design worth carrying, as a confound-bounded (not confound-clean) instrument.** Its single decisive advantage over (a) and (b): **it holds the reform bundle and the era fixed**, because both cohorts are observed *contemporaneously* in the post-2013 environment. The identifying variation is then **entry hours at hire**, conditioned on tenure — not era. This is exactly the variation the within-regional FOQA precursor design already exploits, **plus** a temporal extension the FOQA design cannot manufacture: a low-*entry*-hours cohort that was screened under the *pre-rule* (lower-bar) regime, which gives partial leverage on the aptitude/selection confound (§2.iii). It does **not** escape tenure-decay (§2.ii) or survivorship (§2.iii); those are its binding limitations, not the bundle.

**Why (c) beats (a) here, stated plainly:** the principal's instinct that a *sharp temporal discontinuity* is more powerful is correct **in general** but **inverted in this specific case**, because at *this* cutoff the discontinuity is shared with the whole reform package. The way to *use* the persisting low-hour cohort is precisely to **stop comparing across the cutoff** (which conflates with the bundle) and instead compare **within the post-period** by entry-hours stratum (which holds the bundle fixed). The grandfather/pre-rule cohort's value is as a **persisting low-entry-hours stratum inside the post-period**, not as one side of a cross-cutoff jump.

---

## 2. The Confounds (rigorous and symmetric)

### (i) ERA / reform-package confound — the decisive one for design (a), neutralized by design (c)

The rule was one of **5+ near-simultaneous post-Colgan reforms** (1,500-hour ATP; Part 117 fatigue; stall/upset training; SMS; PRD), all landing 2013–2014 ([ATP-Rule-Safety-Ledger.md](../../whitepapers/ATP-Rule-Safety-Ledger.md)). A naive **pre-2013-cohort vs post-2013-cohort** comparison (design b, and the cross-cutoff read of design a) conflates entry-hours with the **entire bundle plus the secular safety trend**. The Safety Ledger already establishes (Question 4) that the headline fatality drop was *fully present in 2011–2012, before the rule existed* — i.e., the era is doing work the entry-hours count cannot claim.

- **Does RD-in-time at the sharp cutoff mitigate this? NO** — uniquely for *this* cutoff. RD's exclusion restriction requires that the only thing changing at the threshold is the treatment. Here the threshold is the bundle's onset, so RD inherits the confound rather than defeating it. This is the opposite of the typical RD-in-time case (e.g., a benefit eligibility date) where only one program switches on.
- **Does the within-post-period entry-hours stratification (design c) mitigate this? YES** — by holding the era and bundle fixed for both strata (both observed post-2013), the bundle is differenced out. This is *why* (c) is the workhorse.

### (ii) TENURE / ACCUMULATION decay — the binding limitation of design (c)

The pre-rule ~600-hour entrants have since accumulated **thousands** of hours. The entry-hours "treatment" therefore **attenuates monotonically with time-since-hire**: a pilot's *current* proficiency is dominated by accumulated experience, not entry hours, once tenure is large. Consequences:

- The discontinuity is **sharpest in the first post-hire years** and **vanishes asymptotically**. Any clean read requires either (a) restricting to **early-career windows** (first 300–1,000 line-hours), which for the pre-rule cohort means looking at **pre-2013 records** (re-importing the era confound — a genuine bind), or (b) a **tenure-by-entry-hours interaction** model that tests whether the *entry-hours gap* at matched tenure has closed.
- **Symmetric-honesty note:** if the entry-hours effect has fully attenuated by mid-tenure, a null at mid-tenure is a **"signal decayed below detectability,"** NOT a measured null on the entry-hours effect. The decay is mechanically expected and must not be sold as evidence the entry-hours floor never mattered.

### (iii) SURVIVORSHIP / SELECTION — who among the low-hour entrants stayed

The pre-rule low-entry-hours cohort observed today is a **survivor sample**: low-hour entrants who washed out, were not retained, upgraded out of the FO seat, left for majors, or self-selected away are **not** in it. If attrition was **safety-correlated** (the weakest low-hour entrants left disproportionately), the surviving low-hour cohort is **positively selected**, biasing its precursor rate **down** and making low entry hours look *safer* than the entry population was — a bias **toward** "experience doesn't matter." Conversely, if the strongest low-hour pilots were poached upward fastest, the residual is negatively selected. **The direction is not signable a priori**, so survivorship must be bounded (e.g., with a retention/attrition negative-control and an attrition-weighted sensitivity), not assumed away. Differential attrition between the pre- and post-rule cohorts (different labor-market eras — see iv) compounds this.

### (iv) COMPOSITION — the just-before vs just-after cohorts differ on more than hours

The just-before cohort was hired in a **different labor-market regime** (pre-2022 pay, different hiring-boom timing) than later post-rule cohorts. The pay-tripling natural experiment of 2022 ([ATP-Rule-Safety-Ledger.md](../../whitepapers/ATP-Rule-Safety-Ledger.md), Question 3) changed *who applies and who is retained*. Cohorts also differ on age-at-entry, prior-career mix, and pathway (CFI time-builder vs academy R-ATP). Composition must be matched/covariate-adjusted (carrier, fleet, age-at-entry, pathway, calendar labor-market stratum), and even then **pathway is co-confounded with aptitude/screening** exactly as in the within-regional design ([part141-vs-regional-proxy.md](part141-vs-regional-proxy.md); [PRE-REGISTRATION-precursor-study.md](PRE-REGISTRATION-precursor-study.md) §6). The grandfather arm does **not** escape the aptitude confound; it only adds a *second* screening regime (pre-rule lower bar) that gives partial leverage on it.

**Symmetric scorecard:** design (a) is dominated by (i); design (c) trades (i) away but pays in (ii) and (iii). Neither yields a causal entry-hours-dose verdict cross-sectionally; both can at best yield a **confound-bounded** stratified contrast.

---

## 3. Data Feasibility

### 3.1 What the design needs

Event-level **incident/precursor records** (FOQA exceedances, unstable approaches, NTSB Part 830 reportable events, FAA AIDS, ASAP) **linked per pilot to:** (a) **hire date** (the running/cohort variable), (b) **entry hours at hire** (the treatment variable), and (c) **tenure / time-since-hire** (the decay modifier), plus carrier, fleet, age-at-entry, and pathway covariates. This is a **pilot-level longitudinal linkage of safety events to a hiring/experience record** — strictly more than the within-regional FOQA design needs, because it additionally requires **historical entry-hours provenance** for the pre-rule cohort.

### 3.2 Is it public? NO — and one blocker is worse than the FOQA study's

- **PRD does not carry entry hours.** The obvious national keystone, the FAA **Pilot Records Database**, **explicitly excludes flight-time, duty-time, and rest-time records** from what may be reported ([14 CFR Part 111](https://www.ecfr.gov/current/title-14/chapter-I/subchapter-G/part-111); confirmed in the [PRD Final Rule, 86 FR 31006](https://www.faa.gov/sites/faa.gov/files/2021-08/Pilot_Records_Database_Final_Rule.pdf)). PRD holds **hire date, training/qualification/proficiency outcomes, and separation/disciplinary records** — useful for tenure and for a *training-failure* proxy, but **not the entry-hours treatment variable.** So even the carrier that has PRD access cannot read entry hours out of it; entry hours live in **carrier hiring/application records and pilot logbooks**, which are private. This is a **strictly larger** linkage burden than the FOQA precursor study, which can in principle stratify on a *current/pathway* tier without reconstructing historical entry hours.
- **FOQA/ASAP feed ASIAS de-identified.** As the locked pre-registration documents ([PRE-REGISTRATION-precursor-study.md](PRE-REGISTRATION-precursor-study.md) §3, §12), FOQA/ASAP/ATSAP enter ASIAS **de-identified and aggregated** under MOUs with MITRE/CAASD as trusted third party; pilot-level event-to-hire-record linkage is **exactly what the de-identification architecture exists to prevent.** Linking events to a pilot's hire date *and* entry hours without re-identification is the same **Gate G2** the FOQA study flags as **most likely to resolve NO**.
- **Net:** **carrier-side data-use agreement or a protected-ASIAS AEB-directed study only.** Public data cannot execute this. **This is a data-feasibility failure, not a finding** — the symmetric-honesty rule applies: it tells us the effect is **unmeasured with public data**, never that it is zero.

### 3.3 Has anyone ever done a pre/post or RD cohort study of the 1,500-hour rule's entry cohorts?

**On safety/incident endpoints: NO — confirmed gap.** Searches of NTSB, FAA, GAO, and the transportation-safety/econometrics literature surface **no** regression-discontinuity or pre/post **incident-rate** cohort study of the FOQ rule's entry cohorts. What exists instead, and what it does *instead* of the entry-cohort incident contrast:

- **Pilot Source Study (2010, 2015, 2018)** — the **nearest** work, and it is on **training performance, not incidents.** PSS 2018 collected 9,776 pilot records from five regional airlines and **explicitly compared pre- vs post-FOQ new-hire first officers**. Its headline is directly germane and counterintuitive: post-FOQ hires had **significantly more total flight hours** yet were **more likely to need additional training and less likely to complete training**; the most-successful-in-training profiles included **R-ATP (750h military / 1,000h institutional)** holders and those with **1,500 or fewer** total hours ([Pilot Source Study 2018](https://www.researchgate.net/publication/340507735); [academia.edu mirror](https://www.academia.edu/67556609/)). This is the closest anyone has come to the principal's design, and it (i) uses the *training-completion* endpoint, not safety precursors/incidents, and (ii) finds **more hours ≠ better early performance**, which is itself a caution against assuming the entry-hours sign. **[VERIFIED the study exists and is pre/post-FOQ; its endpoint is training performance, not incident rates.]**
- **FAA FOQ regulatory evaluation (2013)** — found **no quantifiable relationship** between the 1,500-hour count and accidents (already in the Safety Ledger, Side A); it did **not** run an entry-cohort RD.
- **GAO pilot-supply reports (2014, 2024)** — labor-market and supply analyses; **no** safety-cohort RD.
- **NTSB** — no entry-cohort study; the agency never recommended an hour minimum ([Safety Ledger Q1](../../whitepapers/ATP-Rule-Safety-Ledger.md)).

**Conclusion: the safety-endpoint entry-cohort study is a genuine gap**, and the nearest neighbor (PSS) is on the wrong endpoint and points away from a naive "more hours = safer" prior.

---

## 4. Disposition

### 4.1 Is the grandfather-RD valid, powerful, feasible?

| Criterion | Verdict | Reason |
|---|---|---|
| **Valid** (identifies the entry-hours effect?) | **QUALIFIED** | Only design (c) — within-post-period entry-hours stratification — is valid, and only as a **confound-bounded** contrast (tenure-decay + survivorship cap it). The cross-cutoff RD (a) is **invalid** for this question (bundle shares the cutoff). |
| **Powerful** (adds identifying leverage?) | **QUALIFIED-YES** | Adds a **temporal entry-hours discontinuity** the within-regional FOQA design lacks: a pre-rule cohort screened under a *different, lower* bar, giving partial leverage on the aptitude/selection confound. But the leverage **decays with tenure** and is **survivorship-filtered**. |
| **Feasible** (public data?) | **NO** | Needs event-level precursor records linked to hire date **and entry hours**; PRD **excludes flight hours**; ASIAS is de-identified. Carrier-side / protected-ASIAS only — same Gate-G2 blocker, **plus** an entry-hours-provenance gap. |

### 4.2 Versus the within-regional FOQA precursor contrast — does it add identifying power?

**Yes, one specific thing, at a specific cost.**

- **What it ADDS:** a **temporal/era axis of variation**. The within-regional FOQA design ([PRE-REGISTRATION-precursor-study.md](PRE-REGISTRATION-precursor-study.md) §6) is **purely cross-sectional** — it stratifies *current* pilots by pathway/hours and is therefore fully exposed to the **aptitude/screening confound** (the structured-pathway cohort is differently screened; dose-vs-selection is *non-identifiable cross-sectionally*, which that pre-registration concedes as fatal to a causal read). The grandfather arm introduces a cohort that entered under a **historically different (lower) screening regime**, so an entry-hours contrast *across screening regimes* gives **partial, independent leverage** on whether the cross-sectional equivalence is screening-driven. That is a real triangulation gain — the **sharp temporal discontinuity the within-regional design structurally cannot produce**.
- **What it COSTS:** the within-regional design has **no era confound and no tenure-decay problem** (it compares contemporaneous early-career pilots). The grandfather arm **re-introduces both**: the reform-bundle/era confound (if read across the cutoff) and tenure-decay (the pre-rule cohort's entry-hours signal has attenuated). So it is **not a strict upgrade** — it buys a new axis of identification by paying in two confounds the primary design avoided.
- **Net:** **complementary, not dominating.** It belongs as a **triangulation arm** whose agreement-or-disagreement with the within-regional result is informative, not as a replacement primary instrument.

### 4.3 What a pre-registered grandfather-RD arm would require

If carried, it must be locked to the same forking-path discipline as the parent study ([PRE-REGISTRATION-precursor-study.md](PRE-REGISTRATION-precursor-study.md) §11):

1. **Design fixed as (c), not (a):** within-post-period entry-hours stratification with an explicit **tenure (time-since-hire) interaction**; the cross-cutoff RD is **excluded** by pre-commitment (bundle-confounded), carried only as a labeled, non-confirmatory sensitivity.
2. **Treatment = entry hours at hire** (continuous, plus a pre-rule/post-rule cohort indicator), **not** "grandfather status" (which §1.0 shows is the wrong variable). Running/cohort variable = **hire date**.
3. **Tenure-decay pre-committed:** primary read at a **matched early-career window**; a mid-tenure null is pre-labeled **"signal decayed below detectability,"** never a measured null.
4. **Survivorship bounded:** pre-registered **attrition/retention negative-control** and an **attrition-weighted sensitivity**; the survivor-selection sign is declared **a priori unsignable**.
5. **Composition matched:** carrier, fleet, age-at-entry, pathway, and **calendar labor-market stratum** (to absorb the 2022 pay-shock cohort difference). Aptitude/screening confound declared **non-eliminable**, with an **E-value**, exactly as in the parent design.
6. **Confirmatory-ineligible, zero BH-FDR family members** (mirroring the European arm's disposition): it is an **exploratory triangulation arm**, contributes no confirmatory test, and **cannot move the locked net-ledger verdict**.
7. **Data-access branch pre-committed:** absent carrier-side or AEB-directed-study linkage of events to hire-date **and entry hours**, the arm reports **"not executable as specified — entry-hours provenance unavailable in public/PRD data,"** never a relaxed proxy and never a measured null.

---

## Implications for the FOQA Precursor Pre-Registration

**Recommendation: ADOPT as a clearly-labeled, exploratory, confirmatory-ineligible *triangulation arm*; do NOT promote to a primary confirmatory instrument; do NOT let it touch the locked verdict.**

- **Why adopt at all:** it supplies the **one identifying axis the within-regional design provably lacks** — a temporal/cross-screening-regime entry-hours discontinuity giving partial leverage on the aptitude/selection confound the parent pre-registration concedes is *fatal to a causal read cross-sectionally* ([PRE-REGISTRATION-precursor-study.md](PRE-REGISTRATION-precursor-study.md) §1, §6). Agreement between the within-regional contrast and the grandfather-cohort contrast would be a **genuinely stronger triangulation** than either alone.
- **Why keep it secondary:** it re-imports the **reform-bundle/era confound** (design a) and the **tenure-decay** problem (design c) the primary design avoids, and it is **survivorship-filtered**. It is complementary, not dominating.
- **Why it cannot rescue feasibility:** it shares the parent study's **Gate-G2** linkage blocker and adds a **strictly larger** data burden — **PRD excludes flight hours** ([14 CFR Part 111](https://www.ecfr.gov/current/title-14/chapter-I/subchapter-G/part-111)), so entry-hours provenance requires **carrier hiring records or logbooks**, not any public source. The realistic prior is the same **"not executable as specified"** branch.
- **Symmetric-honesty caption (binding):** this arm, if it returns nothing, returns a **data/power failure and/or a tenure-decayed signal — NOT a measured null** on the experience-vs-safety question. It reinforces the locked theme that the rule's marginal safety effect is **testable in principle but untested with available data**, and it does **not** re-open the locked verdict: *"sign indeterminate, magnitude small, modest adverse lean"* — **restated verbatim and untouched.**

---

### Verification notes

- **VERIFIED (high confidence):** Aug 1 2013 effective date; R-ATP thresholds (750 mil / 1,000 four-year / 1,250 two-year); FAA rejection of American Eagle/American Airlines grandfather requests and requirement that all SICs hold ATP/R-ATP; §121.436(e) PIC 1,000-hour-experience grandfather; PRD exclusion of flight/duty/rest-time records; pre-2013 regional entry at ~500–700/800–1,000 hours; Pilot Source Study 2018 is a pre/post-FOQ **training-performance** (not incident) cohort study with the "more hours, more training needed, less likely to complete" result; no NTSB/FAA/GAO/academic **incident-rate** entry-cohort RD or pre/post study exists.
- **[UNVERIFIED / flagged]:** the exact FAA estimate of the *number* of incumbent FOs affected by the conversion (regulatory evaluation figure not retrieved — federalregister.gov and the FAA-hosted final-rule PDF were access-blocked during research; figure should be pulled from 78 FR 42324's regulatory-evaluation section before any citation-grade use); the precise transition timeline/relief mechanics in N 8900.225 (PDF access-blocked; the "avoided inadvertently disqualifying incumbent SICs" language is from secondary FAA-comment reporting and should be confirmed against the notice).
- **EXPLORATORY throughout. Does not re-open the locked verdict.**