River Intelligence: Trade Signal Correlation (G3 Pro)

{
  "code": "## Trade Signal Correlation Hypothesis Matrix\n\n### Methodology\nCross-reference 13 data sources in the oneworld_trade dataset for temporal and categorical correlation signals. Hypotheses ranked by testability and expected signal strength.\n\n---\n\n### HIGH CONFIDENCE Correlations (Strong Causal Chain)\n\n**H1: CSMS Messages → Federal Register Rules (Lead Time: 30-90 days)**\n- CSMS \"Informed Compliance\" messages from CBP often precede formal Federal Register rulemakings\n- Mechanism: CBP publishes guidance → collects comments → publishes final rule\n- Test: Match CSMS message keywords to Federal Register entries within 90-day window\n- Expected r²: 0.6-0.8 for antidumping/countervailing duty CSMS→FR correlation\n\n**H2: FDA Import Alerts → HTS Reclassification (Lead Time: 60-180 days)**\n- FDA Import Alert 99-XX (automatic detention) on specific product origins leads to:\n  - Importers switching to similar products from different origins\n  - CBP re-examining classification of \"substitute\" products\n- Test: FDA alerts on Chapter 20/21 (prepared foods) → reclassification volume spike in those chapters\n- Expected r²: 0.4-0.6\n\n**H3: CSMS AD/CVD Messages → Tariff Rate Changes (Lead Time: 0-30 days)**\n- CSMS messages announcing new antidumping/countervailing duty orders directly cause tariff changes\n- This is causal, not correlation — CSMS IS the notification channel\n- Test: Parse CSMS for \"antidumping\" → match to hts_tariffs rate changes\n- Expected r²: 0.9+ (near-perfect for AD/CVD-specific messages)\n\n---\n\n### MEDIUM CONFIDENCE Correlations (Plausible Mechanism)\n\n**H4: BIS Entity List Updates → Trade Volume Shifts (Lead Time: 0-60 days)**\n- Entity List additions restrict exports → import patterns shift as supply chains reroute\n- Test: BIS entity additions for country X → census_trade_data import volume changes from X\n- Expected r²: 0.3-0.5 (dampened by time lag and substitution effects)\n\n**H5: UN Comtrade Flow Anomalies → CBP Enforcement (Lead Time: 90-365 days)**\n- Trade flow anomalies (sudden volume spikes from transshipment countries) predict CBP enforcement actions\n- Test: Comtrade flow outliers (bilateral) → CSMS enforcement messages or Federal Register special duties\n- Expected r²: 0.2-0.4\n\n**H6: Informed Compliance Publications → Oracle KB Accuracy Improvement**\n- New ICPs add domain knowledge → Oracle KB entries citing ICPs have higher classification confidence\n- Test: ICP publication date → confidence score improvement in affected HTS chapters within 30 days\n- Expected r²: 0.5-0.7 (internal system metric)\n\n---\n\n### SPECULATIVE Correlations (Worth Testing)\n\n**H7: Polymarket Trade Signals → Actual Tariff Changes**\n- Prediction market probabilities for trade policy events (tariff announcements, trade deals)\n- Lead time would be market pricing efficiency — hours to weeks\n- Test: Polymarket contract resolution → Federal Register tariff amendments\n- Expected r²: Unknown — novel hypothesis, no academic precedent for trade-specific prediction markets\n\n**H8: FDA Recalls → Cross-Border Seizure Patterns**\n- FDA recalls for specific product types → CBP seizure data for same product type at ports\n- Test: FDA recall announcements → CBP document/newsroom entries mentioning same product\n- Expected r²: 0.3-0.5\n\n**H9: WTO Dispute Filings → Retaliatory Tariff Actions**\n- WTO disputes have long timelines (years) but intermediate actions (preliminary findings) correlate with tariff adjustments\n- Test: WTO dispute milestones → Federal Register tariff modifications on disputed goods\n- Expected r²: 0.3-0.6 (long lag reduces practical value)\n\n**H10: Census NAICS Reclassification → HTS Chapter Restructuring**\n- When NAICS codes get restructured, affected product categories may see HTS harmonization\n- Test: NAICS revision dates → HTS chapter update frequency\n- Expected r²: 0.2-0.3 (different governance bodies, weak coupling)\n\n---\n\n### Correlation Matrix (13 Sources)\n\n| Source → | CSMS | Fed Reg | FDA | BIS | BTS | Census | Comtrade | WTO | USITC | ICP | CBP News | HTS Tariffs | Polymarket |\n|----------|------|---------|-----|-----|-----|--------|----------|-----|-------|-----|----------|-------------|------------|\n| CSMS     | —    | **H1**  | .   | .   | .   | .      | .        | .   | .     | .   | .        | **H3**      | .          |\n| Fed Reg  | .    | —       | .   | .   | .   | .      | .        | **H9** | .  | .   | .        | .           | **H7**     |\n| FDA      | .    | .       | —   | .   | .   | .      | .        | .   | .     | .   | **H8**   | **H2**      | .          |\n| BIS      | .    | .       | .   | —   | .   | **H4** | .        | .   | .     | .   | .        | .           | .          |\n| Comtrade | .    | .       | .   | .   | .   | .      | —        | .   | .     | .   | **H5**   | .           | .          |\n| ICP      | .    | .       | .   | .   | .   | .      | .        | .   | .     | —   | .        | **H6**      | .          |\n\n### Implementation in River Pipeline\n```python\n# New river stage: stage_correlate\ndef stage_correlate(config, dry_run=False):\n    \"\"\"Test hypotheses H1-H10 using BQ temporal joins.\"\"\"\n    # Query pairs of data sources with time-windowed joins\n    # Compute Pearson r for event co-occurrence within window\n    # Store results in river_intelligence_findings (finding_type='correlation')\n```\n\n### Recommended Priority\n1. **H3** (CSMS→tariff) — near-certain, validate first as baseline\n2. **H1** (CSMS→FR) — high confidence, 30-90 day lead = actionable\n3. **H2** (FDA→HTS) — medium, but high commercial value if confirmed\n4. **H7** (Polymarket→tariff) — novel, IP potential if signal found",
  "content": "## Trade Signal Correlation Hypothesis Matrix\n\n### Methodology\nCross-reference 13 data sources in the oneworld_trade dataset for temporal and categorical correlation signals. Hypotheses ranked by testability and expected signal strength.\n\n---\n\n### HIGH CONFIDENCE Correlations (Strong Causal Chain)\n\n**H1: CSMS Messages → Federal Register Rules (Lead Time: 30-90 days)**\n- CSMS \"Informed Compliance\" messages from CBP often precede formal Federal Register rulemakings\n- Mechanism: CBP publishes guidance → collects comments → publishes final rule\n- Test: Match CSMS message keywords to Federal Register entries within 90-day window\n- Expected r²: 0.6-0.8 for antidumping/countervailing duty CSMS→FR correlation\n\n**H2: FDA Import Alerts → HTS Reclassification (Lead Time: 60-180 days)**\n- FDA Import Alert 99-XX (automatic detention) on specific product origins leads to:\n  - Importers switching to similar products from different origins\n  - CBP re-examining classification of \"substitute\" products\n- Test: FDA alerts on Chapter 20/21 (prepared foods) → reclassification volume spike in those chapters\n- Expected r²: 0.4-0.6\n\n**H3: CSMS AD/CVD Messages → Tariff Rate Changes (Lead Time: 0-30 days)**\n- CSMS messages announcing new antidumping/countervailing duty orders directly cause tariff changes\n- This is causal, not correlation — CSMS IS the notification channel\n- Test: Parse CSMS for \"antidumping\" → match to hts_tariffs rate changes\n- Expected r²: 0.9+ (near-perfect for AD/CVD-specific messages)\n\n---\n\n### MEDIUM CONFIDENCE Correlations (Plausible Mechanism)\n\n**H4: BIS Entity List Updates → Trade Volume Shifts (Lead Time: 0-60 days)**\n- Entity List additions restrict exports → import patterns shift as supply chains reroute\n- Test: BIS entity additions for country X → census_trade_data import volume changes from X\n- Expected r²: 0.3-0.5 (dampened by time lag and substitution effects)\n\n**H5: UN Comtrade Flow Anomalies → CBP Enforcement (Lead Time: 90-365 days)**\n- Trade flow anomalies (sudden volume spikes from transshipment countries) predict CBP enforcement actions\n- Test: Comtrade flow outliers (bilateral) → CSMS enforcement messages or Federal Register special duties\n- Expected r²: 0.2-0.4\n\n**H6: Informed Compliance Publications → Oracle KB Accuracy Improvement**\n- New ICPs add domain knowledge → Oracle KB entries citing ICPs have higher classification confidence\n- Test: ICP publication date → confidence score improvement in affected HTS chapters within 30 days\n- Expected r²: 0.5-0.7 (internal system metric)\n\n---\n\n### SPECULATIVE Correlations (Worth Testing)\n\n**H7: Polymarket Trade Signals → Actual Tariff Changes**\n- Prediction market probabilities for trade policy events (tariff announcements, trade deals)\n- Lead time would be market pricing efficiency — hours to weeks\n- Test: Polymarket contract resolution → Federal Register tariff amendments\n- Expected r²: Unknown — novel hypothesis, no academic precedent for trade-specific prediction markets\n\n**H8: FDA Recalls → Cross-Border Seizure Patterns**\n- FDA recalls for specific product types → CBP seizure data for same product type at ports\n- Test: FDA recall announcements → CBP document/newsroom entries mentioning same product\n- Expected r²: 0.3-0.5\n\n**H9: WTO Dispute Filings → Retaliatory Tariff Actions**\n- WTO disputes have long timelines (years) but intermediate actions (preliminary findings) correlate with tariff adjustments\n- Test: WTO dispute milestones → Federal Register tariff modifications on disputed goods\n- Expected r²: 0.3-0.6 (long lag reduces practical value)\n\n**H10: Census NAICS Reclassification → HTS Chapter Restructuring**\n- When NAICS codes get restructured, affected product categories may see HTS harmonization\n- Test: NAICS revision dates → HTS chapter update frequency\n- Expected r²: 0.2-0.3 (different governance bodies, weak coupling)\n\n---\n\n### Correlation Matrix (13 Sources)\n\n| Source → | CSMS | Fed Reg | FDA | BIS | BTS | Census | Comtrade | WTO | USITC | ICP | CBP News | HTS Tariffs | Polymarket |\n|----------|------|---------|-----|-----|-----|--------|----------|-----|-------|-----|----------|-------------|------------|\n| CSMS     | —    | **H1**  | .   | .   | .   | .      | .        | .   | .     | .   | .        | **H3**      | .          |\n| Fed Reg  | .    | —       | .   | .   | .   | .      | .        | **H9** | .  | .   | .        | .           | **H7**     |\n| FDA      | .    | .       | —   | .   | .   | .      | .        | .   | .     | .   | **H8**   | **H2**      | .          |\n| BIS      | .    | .       | .   | —   | .   | **H4** | .        | .   | .     | .   | .        | .           | .          |\n| Comtrade | .    | .       | .   | .   | .   | .      | —        | .   | .     | .   | **H5**   | .           | .          |\n| ICP      | .    | .       | .   | .   | .   | .      | .        | .   | .     | —   | .        | **H6**      | .          |\n\n### Implementation in River Pipeline\n```python\n# New river stage: stage_correlate\ndef stage_correlate(config, dry_run=False):\n    \"\"\"Test hypotheses H1-H10 using BQ temporal joins.\"\"\"\n    # Query pairs of data sources with time-windowed joins\n    # Compute Pearson r for event co-occurrence within window\n    # Store results in river_intelligence_findings (finding_type='correlation')\n```\n\n### Recommended Priority\n1. **H3** (CSMS→tariff) — near-certain, validate first as baseline\n2. **H1** (CSMS→FR) — high confidence, 30-90 day lead = actionable\n3. **H2** (FDA→HTS) — medium, but high commercial value if confirmed\n4. **H7** (Polymarket→tariff) — novel, IP potential if signal found"
}