Moskowitz GD: a high composite, a low ML head, and the question they don't both answer

Issue 0, 2026-05-20.

This week’s headline correlation scores 85.0 / 100, high-severity tier. Member: Jared Moskowitz. Ticker: GD (General Dynamics Corporation). Read on for what the engine actually saw.

Headline correlation

Jared Moskowitz (D-Florida), GD (General Dynamics Corporation)

Composite score: 85.0 / 100 (high tier)

Transaction	purchase, $1,001, $15,000, 2026-03-23
Disclosure	2026-04-30 (38 day lag)
Sector	Aerospace and Defense, Aerospace and Defense
Committees	House Committee on Foreign Affairs
ML annotation	ML heads diverge: distillation 100.0, forward-CAR 0.0 (sidecar values; both are model scores, not realized returns)

Engine narrative withheld (correction, 2026-08-27). The machine-generated summary previously shown here described the jurisdiction factor as committee overlap and cited lobbying filings as supporting evidence. The factor is a sector match, not a committee match, and lobbying is not a scoring input. The factor contributions remain in the data sidecar; see the methodology page.

Correction, 2026-08-28. This brief was written on 2026-05-20 from an engine 1.1 scoring pass (composite 84.7 with the price-movement factor still live, a single ML head at 31.9, and a SHAP row citing price_movement_raw, a factor removed in v1.2). The replication sidecar under this page was re-emitted from the engine’s 2026-06-21 pass (rollup rollup__quiver_2026-03-23_M001217_GD, scored 2026-06-21T02:07Z) and disagreed with the body on every headline number. The composite, the supporting-signal scores, the ML annotation, and the SHAP row above now match that sidecar; the argument below was written against the earlier figures and still holds on the corrected ones (the two heads sit 100 points apart).

Supporting signals.

• Contract (score 85.0, 2026-03-25), to THE BOEING COMPANY, $135,778
• Contract (score 82.0, 2026-03-27), to THE BOEING COMPANY, $262,452
• Contract (score 65.4, 2026-04-22), to HOWMET AEROSPACE INC., $70,951

SHAP top features (distillation head, sidecar). jurisdiction_raw (+11.20), timing_raw (+7.58), disclosure_lag_raw (+5.78)

Abnormal returns vs SPY. n/a (7d) / n/a (30d) / n/a (90d)

Why the divergence is the story

The composite-distillation head says: “this looks suspicious by the same pattern composite flags.” The forward-CAR head says: “the trade did not actually pay off relative to SPY over the next 30 days.” When the two heads disagree by more than fifty points, the editorial read is not “ignore this” and not “this is the smoking gun.” The read is: the public-record correlation is real, the market return signal is not. That is itself useful context for a journalist or oversight body deciding whether to chase the lead.

Runner-ups this week

• Thomas H. Kean Jr (R), LLYVK (Communication Services), score 87.3, traded 2025-12-16
• Thomas H. Kean Jr (R), LLYVA (Communication Services), score 86.8, traded 2025-12-16
• Jared Moskowitz (D), GE (Industrials), score 84.2, traded 2026-03-23
• Jonathan Jackson (D), TMUS (Communication Services), score 83.6, traded 2026-01-13

Newcomers we’re watching

Members with under 5 total trade disclosures get listed but not scored. We want enough history to compare against before reading signal into their patterns. This week’s newcomers we’re watching:

• Cliff Bentz (R), 1 lifetime trade(s); most recent: INTC on 2026-04-09
• James E. Banks (R), 1 lifetime trade(s); most recent: SBUX on 2026-04-15
• Sheri Biggs (R), 1 lifetime trade(s); most recent: IBIT on 2026-03-04
• Warren Davidson (R), 1 lifetime trade(s); most recent: GEHC on 2026-03-23
• Jennifer Mcclellan (D), 1 lifetime trade(s); most recent: VZ on 2026-04-07
• Greg Stanton (D), 1 lifetime trade(s); most recent: TCNNF on 2026-05-06
• Laura Friedman (D), 2 lifetime trade(s); most recent: CVR on 2026-04-08
• Terri A. Sewell (D), 2 lifetime trade(s); most recent: TPR on 2026-03-25
• Greg Steube (R), 2 lifetime trade(s); most recent: None on 2026-04-08
• Debbie Dingell (D), 2 lifetime trade(s); most recent: PTA on 2026-04-21

This week in the dataset

• 1,854 new trade disclosures ingested
• 7,644 new bill records
• 21,066 new lobbying filings
• 8,516 new correlations scored
• Top composite score this week: 87.3

Methodology footer

Capitol Signals correlates congressional financial disclosures against federal contract awards, legislative activity, and lobbying filings. Composite scores combine six factors. The ML overlay (distillation head + forward-CAR head) is editorial annotation, not a primary signal. See the full methodology for the formula and the open data for replication inputs.

All data is sourced from public records: QuiverQuant for trade disclosures, USAspending for contracts, Congress.gov for legislation, and lda.gov for lobbying filings. Published under CC0.

Capitol Signals is open source. This brief is not investment advice. Read the headline correlation as a starting point for journalism, oversight, or research, not as a buy or sell signal.