METHODOLOGY

Government-Market Intelligence Infrastructure

This document describes the data pipelines, scoring models, and statistical methodologies underlying Signal Congress. It is written for users who want to understand exactly what the platform computes — and where the caveats are.

Sections: Anomaly Scoring · Committee Multipliers · Member Baseline Computation · Corroboration Windows · Alpha Calculation · Cluster Detection · Pipeline Freshness · Caveats

ANOMALY SCORING MODEL

Five Dimensions, One Score

Each congressional trade receives an anomaly score from 0 to 100. The score is computed across five independent dimensions, each scored 0–100, then combined with a weighted average. A committee multiplier is applied to the final score where applicable.

For dimension-level descriptions — what each dimension measures and why it predicts information edge — see the How It Works page.

SCORING FORMULA
raw_score = (sector_deviation × 0.25) + (legislative_proximity × 0.30) + (disclosure_speed × 0.15) + (corroboration_excess × 0.20) + (trade_size × 0.10)
anomaly_score = MIN(100, raw_score × committee_multiplier)
Sector deviation
Trade outside member's established GICS sector pattern
Legislative proximity
Signed days-to-event: hearings, markups, vote windows
Disclosure speed
Filing lag vs member's own historical distribution
Corroboration excess
Number and quality of 8-source corroboration hits
Trade size
Dollar range vs member's own median trade size
COMMITTEE MULTIPLIERS

Access Amplification

Congressional committee assignments give members structural information advantages in specific sectors. The committee multiplier scales the anomaly score upward when a trade occurs in a sector the member's committee directly oversees.

1.6×
Committee Chair
Full oversight authority over the relevant sector
1.4×
Committee Member (direct oversight)
Formal committee assignment covering the sector
1.0×
No committee relevance
Trade in sector outside member's committee scope
MEMBER BASELINE COMPUTATION

24-Month Behavioral Profiles

Signal Congress maintains a behavioral baseline for each active congressional member. Deviations from this baseline drive the anomaly score — a member buying outside their established pattern is more anomalous than a member doing what they always do.

Rolling window24 months
Minimum trades required5 trades (below this threshold, baseline scoring is suppressed)
Update cadenceDaily, following pipeline run at 07:30 UTC
Sector distributionProbability vector across GICS sectors (proportional trade count)
Concentration measureHerfindahl-Hirschman Index (0 = fully diversified, 1 = single sector)
CORROBORATION MATCHING WINDOWS

8 Independent Federal Data Sources

Each corroboration source has a different signal-to-noise profile and is matched using a window and logic appropriate to its data type.

CONTRACT
Government Contracts (USASpending.gov)
60-day window
Member's committee must have oversight of the contract sector.
SOL
SAM.gov Solicitations
60-day window
Layer 1: company name in solicitation title. Layer 2: NAICS code × committee jurisdiction.
LOBBY
Senate LDA Lobbying Registrations
12-month window
Company actively lobbying within the member's committee domain.
FEC
FEC Campaign Finance
Current + prior election cycle
Company PAC contributions to the member or their leadership PAC.
REG
Federal Register (Regulatory)
60-day window
Layer 1: company name in final rule or NPRM. Layer 3: committee jurisdiction × rulemaking agency.
INSIDER
SEC EDGAR Form 4 (Corporate Insider)
60-day window
Same direction only — both buying or both selling. Corporate insider matched to congressional trade ticker.
Executive Convergence (Cabinet Officials)
365-day lookback for cabinet holdings · 60-day match window for congressional trades
Cabinet official OGE 278-T holdings matched against congressional trade ticker.
BIPARTISAN
Bipartisan Cluster Membership
30-day cluster window
Trade is part of a detected bipartisan cluster (2+ members, both parties, same direction). See Cluster Detection below.
ALPHA CALCULATION

Excess Return vs. SPY, Winsorized

Alpha is computed as excess return versus SPY over the measurement horizon (60-day or 180-day window). Returns are winsorized at the 5th and 95th percentile to reduce outlier influence from extreme price moves. This is the figure reported throughout the platform.

Raw alpha (unwinsorized) is available in the backtester for users who want to see the full distribution.

BenchmarkSPY total return (dividends reinvested) over identical holding window
Measurement windows60 days and 180 days from disclosure date (not transaction date)
Winsorization bounds5th and 95th percentile of the return distribution for that strategy
Win rate definitionPercentage of trades where member return exceeded SPY return over the same period
Start dateDisclosure date — reflects real-world signal availability, not insider timing
CLUSTER DETECTION

Bipartisan and Cross-Chamber Patterns

A cluster is defined as 2 or more members trading the same ticker in the same direction within a 30-day rolling window. Clusters are detected across the full trade history and updated daily as new disclosures are filed.

Minimum members2 members
Window30 days (rolling, from earliest trade to latest)
Direction constraintSame direction (all purchases or all sales)
Bipartisan requirementAt least one member from each major party (R and D)
Cross-chamber requirementAt least one House and one Senate member
ETF exclusionTickers classified as PERMANENT_SKIP (broad market ETFs) are excluded
Cluster scoreWeighted by member count, bipartisan status, cross-chamber status, and committee overlap
PIPELINE FRESHNESS

30 Steps · 24 Data Quality Checks

The Signal Congress pipeline runs daily at 07:30 UTC. New trade disclosures appear in the feed within 24 hours of the pipeline run following their filing. Pipeline status and run timestamps are available in the Signal Log.

Run timeDaily at 07:30 UTC
Feed lagNew filings appear within 24 hours of the pipeline run following disclosure
Pipeline steps30 steps (ingestion → normalization → scoring → corroboration → cluster detection → narrative generation)
Data quality checks24 DQ checks — pipeline will not mark a run complete until all pass
Pipeline statusAvailable in the Signal Log (last run timestamp, step count, DQ pass/fail)
CAVEATS & HONEST FRAMING

What the Data Supports — and What It Doesn't

Signal Congress is research infrastructure, not investment advice. The signals we surface are observations drawn from public government records. What you do with those observations — whether you use them for journalism, academic research, policy analysis, or investment research — is your decision to make.

We publish this methodology document because the work of building accurate government-market intelligence requires intellectual honesty about what the data supports and what it doesn't. Backtested alpha is not forward-looking alpha. Anomaly scores identify statistically unusual behavior — they do not confirm the presence of material non-public information, which we cannot observe and do not claim to measure.

If you find an error in our methodology, a false positive pattern we haven't documented, or a statistical claim that doesn't hold up — contact us at research@signalcongress.com.

← Research & BacktestsHow It Works →