W1 forensics 1: the five transport mechanisms (registration 4951218279)

[MaxGhenis]

W1 forensics 1 — the five transport mechanisms

Diagnostic under the FROZEN registration #42 comment 4951218279 (the registration wins). Reported, not gated (reported_not_gated: true); one run; publishes regardless. Measures the five initialization/support/scope mechanisms candidate 1 isolated (grading 4951216895) before candidate 2 designs.

Standard — the gate-2b forensics rounds: instrumentation bit-identity is EXACT (0.0) for every re-simulation of the committed transport_deployment_v1 machinery; the additive decompositions telescope at machine epsilon (all_identity_reconciliations_at_machine_epsilon: true, bar 64·float64_eps). Artifact runs/gate_w1_forensics1_v1.json written via write_new(..., sidecar=True) — the first artifact to adopt the #152 env sidecar (.env.json, environment + gates.yaml contract ref).

Q5 — the single most consequential number: C1's non-reversal is ROBUST

Does a realistic earnings tail move PPI past NRA (reverse the C1 fingerprint)? No — robustly, in the conservative direction.

The committed transport draws cell.quantile(u_year) at every career age — always positive, no p0 zero-mass — an upper read on the tail (37.7% of taxable payroll above the cap). The upper read is the heaviest plausible tail, the case most favourable to the PPI↔NRA swap, and even there PPI savings (0.0169) sit far below NRA (0.2023). Applying the certified p0 (the zero/low-earning years the participation dynamics imply, coupling to Q1) lightens the tail (0.377 → 0.370) and moves PPI savings down to 0.0137 — still below NRA (0.2022), the gap widening from −0.185 to −0.189. Transported AIME does not lift PPI past NRA; a realistic tail only pushes it further away. (C2's committed elimination↔+2pp swap holds under both tails — both exceed the ~16.1% break-even.)

The five findings

Q	Mechanism	Finding (numbers)
Q1 marital	never-married-at-18 entry + certified hazards	Extending exposure barely helps (mean |b|≈0.028; b turns negative past peak marriage as dissolution offsets marriage); the hazard-level residual dominates (mean |c|≈0.095, in 9/10 cells). Observed-init O == rate_a (the frame's own A_MARITL) is the prohibited regenerated_surface identity → initialization (entry-state seeding), not exposure, is the necessary lever. Refines the pre-registered "initialization dominant" into a sharper mechanism.
Q2 participation boundary	gate-1 fits 25–59, no sex covariate	Nearest-bin extrapolation clears 0/6 boundary cells; the train-fitted boundary extension clears 4/6 (62–69 participation + the 18–24 profile); the frame's own ages clear all (the identity). 18–24 participation resists — PSID heads/spouses over-state it (0.887) vs the CPS all-person frame (0.642): a coverage delta atop the support gap.
Q3 household scope	generator composes adults only; rate_a all-person	Scope + composition telescope exactly. Scope is NOT the whole miss. Scope dominates the large-size cells (children collapse them); a large composition residual dominates size-1 (deployed over-produces lone adults even on the adult universe — the same coresidence under-generation as Q1). Contract-consistent fix needs child attachment (certified fertility machinery) and coresidence repair.
Q4 DI level bridge	M4 work-disability prevalence vs SSA beneficiary stock	GATE-DESIGN finding. The concept delta (duration-accumulated insured-beneficiary stock vs work-disability point-prevalence) dominates (>0.5 share; the 60-FRA gap is +21.3pp duration-concept vs +2.6pp M4 shape). No contract-consistent lever closes it — the concept bridge needs a DI-entry hazard + duration model + an insured denominator the gate never defines and that the DI ASR evidence base doesn't even archive. For the ceremony record, not a candidate lever.
Q5 tail upper read	see above	C1 non-reversal robust (the consequential number).

c2 design implications

• Q1 → seed the marital chain's entry state from an initial-state model (not the terminal A_MARITL); the never-married-at-18 entry is the lever, not the certified hazards.
• Q2 → extend the gate-1 fitted support to the boundary ages and add a sex covariate.
• Q3 → attach children via the certified fertility machinery + repair adult coresidence; scope alone can't make a candidate scoreable on the locked all-person rate_a.
• Q4 → do not spend a candidate lever on the family-B DI bands (unreachable without a gate-design concept bridge).
• Q5 → C1's non-reversal is robust, so C2 (not C1) is the fingerprint the transport moves; don't chase a PPI-over-NRA reversal the conservative-direction argument rules out.

Suite

• tests/test_gate_w1_forensics1.py — 17 always-runnable (artifact tier): reconciliations at machine epsilon, decomposition telescoping, the C1-robustness and gate-design invariants.
• tests/test_gate_w1_forensics1_reproduction.py — 5 PSID reproduction pins (integration_psid): the marital cube bit-identity, the career-panel bit-identity, observed-init == A_MARITL, the PSID boundary support.
• Tier manifest refreshed (artifact +17, integration_psid +5), CI-equivalent delta.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
social-security-model	Ready	Preview, Comment	Jul 12, 2026 1:47pm