Eight claim-linked evidence work packages plus one separate research lead. This is a selective supplement, not a blanket September refresh or a new comparative capability assessment.
No capability scores, comparative confidence labels or quantified returns change through this supplement. Separately, the corrections below revise earlier judgments on their existing evidence.
The comparative baseline remains 2026-08-14. Source publication dates, periods observed and the date of this selective review are different facts. The observations below inform specific policy inferences; they do not automatically change a capability rating or establish a quantified return.
Newly added sources appear in the complete evidence library. The research lead below is not counted as admitted evidence.
Corrections to the comparative judgments.
Original August evidence cutoffs and trend horizons remain; September is the correction date. No new source observations are claimed by this record. This is an internal, AI-assisted methodological review, not external peer review or calibrated validation.
Fleet readiness and repair throughput
MU-D3 · MU-T3
Previously
PRC edge / High comparative confidence
Corrected judgment
Insufficient comparative basis / comparison not assessed; High confidence in the documented U.S. maintenance constraint; no axis marker
Why it changed
U.S. maintenance audits do not contain matched PLAN availability or repair-cycle output. Industrial depth is a different construct. Overall maritime remains Contested, score -0.25, because other mission mechanisms remain independently relevant.
Evidence still needed
Class-adjusted available hull-days, repair cycle, accepted deployable returns, and common periods for both fleets.
International information reach
II-D04
Previously
U.S. / partner edge / Low
Corrected judgment
Insufficient comparative basis; historical U.S. audience retained
Why it changed
A method-documented FY2024 U.S. audience and unmatched PRC supply counts cannot identify a present bilateral consumption difference. Relative reach trends are withheld too.
Evidence still needed
Named actor boundaries and matched country/language/audience panels, common periods, unduplicated consumption and post-disruption refresh.
Coalition delivery and durability
DC-D07 · DC-D10
Previously
U.S. / partner edge / Low
Corrected judgment
Insufficient comparative basis; previously published categories remain visible
Why it changed
Absolute security delivery and persistence are retained positive evidence. Removing unmatched Ukraine/NATO cases or holding observability constant leaves no common obligation and stress-exposure denominator for a system-wide edge. Trends no longer imply a relative direction.
Evidence still needed
Matched mission/stakes, obligation-level completion and critical-role retention at 12/24 months, including nonperformance and repair.
Cultural attraction versus effects
II-D06
Previously
Cultural Influence: U.S. / partner edge / Moderate
Education and English choices are opportunity/utility-confounded. Excluding them leaves reputation/arts evidence for attraction, not a causal effect on persistent behavior. No numerical reweighting or causal estimate is claimed.
Evidence still needed
Longitudinal or matched contact studies controlling access, price, visa, earnings, safety and language utility.
Manufacturing scale versus speed and cost
EI-D02
Previously
Manufacturing-scale edge bundled with speed, learning and cost claims
Manufacturing breadth and value added do not directly measure lead time, yields, quality-adjusted costs or qualified surge.
Evidence still needed
Product-level accepted output, lead time, yield/rework, subsidy-inclusive cost and surge timing.
Coercion cases versus population frequencies
IS-P06
Previously
rare_and_objective_dependent; cost more often than policy change
Corrected judgment
cost_not_sufficient_for_compellence
Why it changed
Purposively selected cases identify failures of conversion, not relative frequencies in a population.
Evidence still needed
An independently defined case universe, success/objective definitions and observation windows before estimating prevalence.
Maritime policy priorities
MU-A1
Previously
Highest-return intervention; High feasibility; 1–3 years; faster than equivalent new build
Corrected judgment
Conditional bottleneck hypothesis; feasibility, cost and timing not estimated; explicit alternatives and stop test
Why it changed
Maintenance audits diagnose a constraint but do not compare marginal returns or establish equivalence with other interventions.
Evidence still needed
Named ship class/work package, deadline, marginal resources, recovered usable hull-days, displaced workforce, alternatives and uncertainty.
No role or observability is inferred from the label/confidence. No legacy ID is exempt. Directional records have explicit scoped mechanisms; limited outcome observability lowers selected comparative confidence. For non-directional records lacking separate factual adjudication, factual confidence is explicitly Not independently assessed, not copied from an old comparative rating. These remain ordinal judgments, not probabilities.
Evaluate instrument choice before subsidizing scale
The study finds increased Chinese commercial shipbuilding scale alongside idle capacity and instrument-dependent returns; entry subsidies perform poorly in its model.
Period observed
Firm data, 1998–2014
Release and cutoff relationship
Pre-baseline source newly admitted in this selective supplement.
What changes
Require separate work-package cases for qualification, production and new entry, including utilization and displaced labor.
What remains unknown
The US naval marginal return and strategic option value are not estimated by this study.
Revisit when
Stop or redesign expansion when qualified deliveries fail to rise or scarce labor is displaced from higher-priority contracted work.
Revise redundancy claims when validated value-added accounting reveals shared upstream inputs, then test those inputs against current supplier records.
European aid allocations and continuing reliance on US weapons coexist in the June 2026 tracker update.
Period observed
Allocations through June 2026
Release and cutoff relationship
Published August 13: after the August 12 diplomacy cutoff, before the August 14 capstone cutoff; admitted September 19, not a retroactive refresh of diplomacy.
What changes
Record financier, producer and stockpile owner separately; verify physical delivery before counting operational contribution.
What remains unknown
Replacement production, arrival and operational acceptance remain separate evidence needs.
Revisit when
Revise mutual-dependence assumptions when delivery-level data show a sustained change in production or stockpile reliance.
Maritime renewal: compare alternatives against a common availability window
Qualitative policy inference; not a measured cost-benefit ranking
For the next named ship-class availability requirement, compare accepted usable hull-days by its specified deadline. This supplement does not select a class, deadline or budget without program data.
Repair and qualify existing capacity
May recover availability sooner if the binding task is executable.
Constraint: Requires a scoped work package, materials, qualified workers and acceptance schedule; extra work can displace another repair.
Decision gate: Approve the next increment only with baseline schedule, workforce loading, alternatives and accepted-output reporting.
Add production or yard capacity
May widen longer-term output if qualified demand and complementary inputs exist.
Constraint: Construction, qualification and utilization lag; entry alone need not produce useful capacity.
Decision gate: Require credible demand, phased qualification, utilization and separate accounting for displaced labor.
Use an authorized allied bridge
May cover a specified gap without duplicating every domestic capability.
Constraint: Sovereign permission and independently verified delivery timing remain necessary; Collins audit raises implementation risk, not measured unavailability.
Decision gate: Count only an accepted contribution and keep an alternative until bridge milestones are met.
Common measure: Accepted hull-days for the stated mission and interval; incremental lifecycle resources; schedule uncertainty; work displaced elsewhere.
Decision rule: Prioritize only after comparable program cases exist. Sequence scope/qualification before expansion when those are binding; retain alternatives if bridge timing is uncertain.
Not estimated: Marginal cost per accepted hull-day; Comparative returns across alternatives; Class-specific workforce displacement; Operational availability and combat effect.
Revisit when: Reopen the decision if scope, deadline, workforce availability, partner authorization or independently verified milestones change.
Later result: The completed original study reports 2022 upstream value-added exposure and was completed on 2026-09-23. The specification below records its unexecuted status as of September 19; that historical record is unchanged.
How much PRC value added is embodied in a partner electronics export bundle, compared with US value added?
No estimate has been calculated for this update.
OECD regular ICIO 2025 edition, January 2026 revision, year 2022; use the published 2016–2022 CSV bundle and edition annex.
How to reproduce a bounded estimate
Download the edition-specific CSV and annex from the cited OECD dataset page; record SHA-256, vintage and sector mapping.
Form the intermediate-use matrix Z, gross output x and value-added vector v; reconcile accounts and currency units before calculating.
Set A = Z diag(x)^(-1), excluding or explicitly handling zero-output sectors; compute L = (I - A)^(-1).
Define e as the selected partner-sector gross-export vector, excluding domestic sales and avoiding aggregation overlap. Compute diag(v / x) L e.
Sum origin-sector contributions separately for PRC and US; divide each by the selected export value. Reconcile the all-origin sum against exports and report residuals, revisions and aggregation limits.
No downloaded matrix or computed bilateral estimate is claimed. This specification does not establish current dependence, substitutability, ownership or wartime access.
Post-baseline and post-September 6 synthesis; no criticism of the frozen baseline.
Listing and abstract verified; capital-flow data, underlying methods and vintage have not been independently validated here. No numerical result or capability revision is admitted.
Before admission: Retrieve the full methods and data vintage; reproduce a bounded capital-embodied exposure measure and compare it to conventional TiVA without double counting.
Specify value-added dependency checks before aggregating allied output.
Limit: Lagged sector accounting cannot establish substitution, inventories or wartime availability; no bilateral exposure estimate calculated in this update.