Module 6 — Global Economic Growth and Overseas Markets

Normative specification. Module 6 is deterministic (no LLM). It scores whether economies and markets outside the United States are a tailwind or headwind for US equities, on a −2…+2 scale, and maps that onto US sectors/industries/sub-industries and foreign-exposed companies. It aggregates regional growth (Eurozone, UK, Japan, emerging-markets-ex-China, other developed markets), pulls China in once as a finished handoff from Module 7, adds a global-activity leg (trade volume + industrial production), overlays an economic-surprise adjustment, and confirms with markets using both a plain industry ETF and an ETF/SPY relative-strength ratio. It emits company geographic-exposure alerts and foreign-company→industry spillovers. It does not decide the final risk/currency/yield regime, select a stock, set entry/stop/target, or place an order — those belong to the specialist modules, Output 2, and the trading bot. The US is excluded from the scored regional aggregate.

Implementation target (not yet built): routes/module6.py → snapshots/module6.json (+ a stateful store .data/module6_events.db for spillover overlays), surfaced in a "Module 6" navbar view. It is the overseas-growth counterpart to the sibling deterministic modules and depends on Module 7 (China) upstream.

Implementation note (v1 ceilings). The Part 4 scoring is pure arithmetic; its numeric inputs (GDP/PMI/IP/trade) are data-feed lookups, and only the foreign-company spillover classification (Calc 13) needs NLP. The intended split is the same as Module 5: an LLM extraction pass (agents/skills/module6_extract.md) only for the spillover/company-event front-end, then a deterministic scorer for everything else, with zero model calls on the numeric hot path. Regional-macro ingestion (Eurostat, ONS, Japan Cabinet Office/METI, S&P Global PMI, CPB World Trade Monitor, IMF weights, a licensed consensus/surprise feed) is not integrated today — we have FRED + Polygon/Massive, which cover the market-confirmation leg but not the official regional releases; v1 caps DATA_COVERAGE low until those feeds exist. Module 7 does not exist yet, so v1 runs PARTIAL_EXCLUDING_CHINA (confidence capped below HIGH per Part 11.3). Two numeric decisions are proposed, not sourced (Part 15.3 ±2% ratio threshold, Part 15.4 confidence weights). Cross-module handoffs (Part 10) and Output 2 are placeholders until those modules exist.

---

Section One — Deterministic specification

Part 1 — Master calculation dependency (14 calcs)

1. Data Validation → DATA_STATUS, MISSING/STALE/CONFLICT lists, REGIONAL_WEIGHT_COVERAGE_PCT, PARTIAL_OUTPUT_STATUS, FAILURE_REASON.
2. Regional Universe and Weights → per-region PPP-GDP weight %, VALID coverage % (uses China weight + Module 7 output).
3. Regional Economic Growth Engine → GDP/PMI/IP scores → REGION_RAW_GROWTH_SCORE → REGION_FINAL_GROWTH_POINT, REGION_GROWTH_STATUS.
4. Overseas Production and Trade Engine → overseas IP point + trade-volume point → GLOBAL_ACTIVITY_POINT, DESCRIPTIVE_ONLY_FLAG.
5. Base Module 6 Economic Score → weighted regional growth + China + global activity → BASE_MODULE_6_DIRECTION_POINT.
6. Economic-Surprise Overlay → standardised surprise → SURPRISE_ADJUSTED_DIRECTION_POINT.
7. Overseas Equity and Relative-Strength Engine → regional + industry absolute/relative status, LOCAL_CURRENCY_USD_CONFLICT_FLAG.
8. Overseas Bond and Currency Conditions → per-region bond/currency condition, funding-stress & recession-concern region counts.
9. Overseas Risk Confirmation → 6-vote total → OVERSEAS_RISK_STATUS, HARD_STRESS_CONFIRMATION.
10. Market Confirmation Adjustment → FINAL_MODULE_6_DIRECTION_POINT.
11. Sector/Industry/Sub-Industry Impact → final impact lists + REGIONAL_CONTRIBUTION + MIXED_IMPACT flags.
12. Company Geographic-Exposure Alerts.
13. Overseas Company-to-Industry Spillover.
14. Confidence and Final Output → MODULE_6_CONFIDENCE/RISK/PERMISSION, OUTPUT_1.

Part 2 — Direct input parameters

• 2.1 Run-control: MODULE_NUMBER 6; UTC+ET timestamps; RUN_TYPE (SCHEDULED|EVENT_TRIGGERED|REVISION|MANUAL); DATA_STATUS_IN; NEW_ENTRY_STATUS_IN; ACTIVE_RESTRICTION_LIST_IN.
• 2.2 Provider assumption: all required feeds assumed available (provider selection is implementation config, not a calc gap) — but freshness, provider-consistency and missing-data rules still enforced; no silent estimation/substitution.
• 2.3 Configurable parameters (all thresholds applied unrounded):
• Region blend: GDP 30% · Composite PMI 40% · IP 30%.
• Base blend: Regional growth 70% · Global activity 30%; global activity = IP 50% · Trade 50%.
• Lookbacks: GDP latest 2 quarters, median prev 20 quarters; PMI trend 3 releases; IP/trade current 3 mo vs prior 3 mo; equity/ratio/currency EMA 20 sessions, return 5 sessions; yield-change 5 sessions; surprise history 3 years; surprise overlay 1 US session.
• Coverage: min regional 80% · min EM-ex-China 80%.
• Exposure: minor 10% · major 20% · company alert 20%.
• Spillover: ≥3 peers · rel move ≥1.00% · RVOL ≥1.50 · window first 60 min · structural capacity ≥5% within 24 mo.
• Risk votes: RISK_ON ≥+4 · RISK_OFF ≤−4 · max 1 opposite vote.
• 2.4 Scored regions: Eurozone, UK, Japan, China (from Module 7), EM-ex-China, other DM; US excluded from the scored aggregate.
• 2.5 Other DM: ≥ Canada, Australia, Switzerland + approved DM not already in EZ/UK/Japan; each country in exactly one region.
• 2.6 China pass-through: consume Module 7's final point/status/equity/confidence/impact lists; never recalculate Chinese GDP/PMI/IP/retail/property/credit/stimulus/yuan/equities.
• 2.7 Per-region economic inputs: real GDP QoQ (+prior/history), composite/mfg/services PMI, IP index, release timestamp, observation period, initial/revised, seasonal-adj, annualisation, official source, pre-release consensus median.
• 2.8 Market inputs: local-currency index, USD-adjusted index/ETF, 5-session return, 20-session EMA, region ETF/SPY ratio (+its 5-session return & 20-session EMA), 2y/10y yields (+5-session changes), currency vs USD, EM sovereign spread, market-open/holiday status.
• 2.9 Industry map fields: INDUSTRY_ID, primary + backup ETF/index, absolute instrument, relative numerator/denominator (normally SPY), ratio, fixed peer group, benchmark, effective/review dates, validity status.
• 2.10 Prescribed industry map (starting config, validate each before deploy): Semis SMH/SOXX; Software IGV/XSW; Biotech XBI/IBB; Regional banks KRE/IAT; Homebuilders XHB/ITB; Airlines JETS/peers; Transports IYT/XTN; Aero-defence ITA/XAR; Gold miners GDX/GDXJ; Oil E&P XOP/IEO; Oil services OIH/IEZ; Pipelines AMLP/ENFR; Steel SLX/peers; Aluminium peers (no ETF if none representative); Copper COPX/ICOP; Ag-inputs peers + MOO/SPY; Shipping BOAT/peers; Solar TAN/RAYS; Wind FAN/peers; Crypto-equities WGMI/broad; China internet KWEB/PGJ; Pharma XPH/PPH; Retail importers XRT + high-import peers; Autos CARZ/peers; Machinery peers + XLI/SPY; Hotels/travel PEJ; Luxury peers. Every ratio is PRIMARY/SPY.
• 2.11 Map review: at deploy, first trading day each quarter, on ETF closure/merger/mandate/ benchmark change, on liquidity drop, or when the ETF stops representing the industry. Revisions change future calcs only — never rewrite history.

Part 3 — Data definitions and validation

• 3.1 Controlling sources: Eurostat (EZ GDP/IP); UK ONS (UK GDP+revisions); Japan Cabinet Office (GDP), METI (IP); S&P Global (common PMI method, 50=neutral, value-added composite); CPB World Trade Monitor (global trade + IP); IMF (classification + PPP weights); Module 1's live provider (equity/ETF/index/yield/FX); licensed consensus provider (medians + surprise history); geographic-exposure provider + audited filings (regional exposure). News = alerts only.
• 3.2 Source priority: latest controlling official release → latest official revision → licensed reproduction → approved backup (only when main invalid) → news (alert only). Never average official vs unofficial.
• 3.3 Max data age: live market 2 min (else last official close); PMI 40 d; IP 70 d; GDP 120 d; CPB trade/production 100 d; surprise index 2 business days; consensus captured ≤24 h pre-release; company geo-exposure 120 d; IMF weights = latest WEO; ETF map reviewed within current quarter.
• 3.4 Min coverage: ≥80% of overseas regional weight and ≥80% of EM-ex-China weight; valid weights renormalised to 100% only when coverage passes; China stays separately identified.
• 3.5 Missing-data: never estimate/repeat-prior/substitute-neighbour/substitute-unrelated-ETF. Non-critical missing + coverage ≥80% → partial result. Critical coverage <80% → total economic output INVALID, 0 points. Module 7 missing → PARTIAL_EXCLUDING_CHINA. Missing market confirmation does not invalidate valid economic data. Missing ETF → backup → peer group → MARKET_CONFIRMATION_UNAVAILABLE (adds no confirmation).
• 3.6 Provider consistency: one PMI methodology across regions; one consensus provider per surprise history; one geo-exposure definition; one ETF/ratio map per period; keep local vs USD returns both separate; never mix annualised/non-annualised GDP.
• 3.7 Initial vs revised: store both; release-day surprise uses initial, continuing conditions use latest revised; a revision never rewrites the original surprise; recalc after every revision.
• 3.8 Sessions: 5-session calcs use completed sessions, skip holidays, use official close on early-close, mark CLOSED_CURRENT_DATA while a market is closed, never duplicate a close; US ETF ratios use the US calendar, local indexes the local calendar.
• 3.9 Rounding: thresholds on unrounded values; display GDP/production/trade 2 dp, PMI 1 dp, yields 3 dp, bp changes 1 dp, exposure 2 dp, raw scores 2 dp.

Part 4 — Deterministic calculations

• 4.1 GDP (real, SA, QoQ, non-annualised; average latest 2 quarters vs 5-yr median): ≥+0.50% & above median → +2; >+0.10% → +1; −0.10%…+0.10% → 0; <−0.10% & >−0.50% → −1; ≤−0.50% → −2.
• 4.2 PMI (each of composite/mfg/services): ≥53.0 → +2; 50.5–<53.0 → +1; 49.5–<50.5 → 0; 47.0–<49.5 → −1; <47.0 → −2. Trend (vs 3 releases ago): ≥+1.5 IMPROVING, ≤−1.5 DETERIORATING, else STABLE. Composite drives the regional PMI score; mfg/services are detail only.
• 4.3 Industrial production (3-mo avg vs prior 3-mo avg, % change): ≥+1.50% → +2; +0.25%…<+1.50% → +1; −0.25%…+0.25% → 0; <−0.25% & >−1.50% → −1; ≤−1.50% → −2.
• 4.4 Regional raw growth = composite PMI×40% + GDP×30% + IP×30%. Raw→point boundaries (reused everywhere): ≥+1.25 → +2; +0.35–<+1.25 → +1; −0.35…+0.35 (exclusive) → 0; −1.25…−0.35 → −1; ≤−1.25 → −2.
• 4.5 EM-ex-China: exclude China, weight valid countries by IMF PPP, require ≥80% coverage, convert via 4.4 boundaries; <80% → INVALID, 0.
• 4.6 Global trade volume (3-mo vs prior 3-mo): same thresholds as 4.3. If the only series bundles US+China inseparably → display, DESCRIPTIVE_ONLY_FLAG=TRUE, 0 trade points, no double-count.
• 4.7 Overseas IP: same formula/thresholds as 4.3, excluding US and (when Module 7 supplies it) China.
• 4.8 Global activity = overseas IP×50% + trade×50%, convert via 4.4 boundaries.
• 4.9 Weighted regional growth = Σ(regional point × valid IMF PPP weight) + China Module-7 point × China weight, ÷ total valid weight. China appears as a separate contribution and contributes exactly once.
• 4.10 Base Module 6 = weighted regional growth×70% + global activity×30%, convert via 4.4.
• 4.11 Economic surprise: raw = actual − consensus median; standardised = raw ÷ SD of comparable 3-yr forecast errors (require ≥24 releases). ≥+1.0 POSITIVE, −1.0…+1.0 NEUTRAL, ≤−1.0 NEGATIVE. Positive surprise on base 0/+1/+2 → one level more positive (cap +2); negative on 0/−1/−2 → one level more negative (cap −2); a surprise never reverses the base sign; overlay expires after one completed US session.
• 4.12 Plain ETF absolute direction: STRONG_POSITIVE = above 20-EMA & 5-session return ≥+3.00%; POSITIVE = above EMA & 0<ret<+3.00%; NEUTRAL = price/return disagree or ret=0; NEGATIVE = below EMA & −3.00%<ret<0; STRONG_NEGATIVE = below EMA & ret ≤−3.00%.
• 4.13 ETF/SPY ratio (= ETF÷SPY): STRONG_RS = ratio above 20-EMA & 5-session ratio ret ≥+2.00%; RS = above EMA & 0<ret<+2.00%; NEUTRAL = disagree or 0; RELATIVE_WEAKNESS = below EMA & −2.00%<ret<0; STRONG_RW = below EMA & ret ≤−2.00%.
• 4.14 Combined industry confirmation (first match): (1) plain ≥POSITIVE & ratio ≥RS → CONFIRMED_BULLISH; (2) plain ≤NEGATIVE & ratio ≤RW → CONFIRMED_BEARISH; (3) plain up + ratio weak → ABSOLUTE_UP_RELATIVE_LAG; (4) plain down + ratio strong → ABSOLUTE_DOWN_RELATIVE_ RESILIENCE; (5) either neutral → UNCONFIRMED; (6) ETF/ratio missing but peers valid → peer confirmation; (7) all missing → MARKET_CONFIRMATION_UNAVAILABLE. No macro point is added from confirmation — it only changes confidence/allocation reliability.
• 4.15 Why both: plain ETF answers "is the price rising?"; ratio answers "is it beating SPY?". Not bullish merely for falling less than SPY; not strong merely for rising while lagging.
• 4.16 Regional equity direction: compute local-currency + USD-adjusted + regional ETF/SPY ratio. Confirmation needs local & USD signs to agree, ratio supporting when available; a conflict → MIXED_CURRENCY_TRANSLATION. (MSCI ACWI ex-USA = approved broad ex-US benchmark.)
• 4.17 Currency (foreign vs USD): STRONG = 5-session ret ≥+1.00% & above 20-EMA; WEAK = ≤−1.00% & below EMA; else NEUTRAL. Confirmation only.
• 4.18 Bond yield (10y, 5-session change): GROWTH_CONFIRMING = equities+ , currency not weak, +5…<+20 bp; EASING_SUPPORTIVE = equities+ , −5…<−20 bp; INFLATION/FUNDING_STRESS = equities−, currency weak, ≥+20 bp; RECESSION/DEFLATION = equities−, ≤−20 bp; else MIXED. No direct growth points.
• 4.19 Overseas risk vote (6 votes: broad, Europe, UK, Japan, EM-ex-China equity, EM currency): +1 positive / 0 neutral-mixed / −1 negative. ≥+4 with ≤1 negative → RISK_ON; ≤−4 with ≤1 positive → RISK_OFF; else MIXED. ≥2 funding-stress regions → RISK_ON prohibited, use MIXED_WITH_FUNDING_STRESS unless risk-off passes.
• 4.20 Market confirmation adjustment: econ+ & risk-on → CONFIRMED_POSITIVE; econ− & risk-off → CONFIRMED_NEGATIVE; opposite pairings → CONTRADICTED; else MIXED. Contradicted moves one level toward zero (+2→+1, +1→0, 0→0, −1→0, −2→−1). Markets can't reverse the official direction.
• 4.21 Sector/industry geo-exposure: REGION_CONTRIBUTION = REGION_FINAL_POINT × REGION_EXPOSURE_PCT ÷ 100. Keep revenue/customer/production/export/supply-chain exposure separate. Caps: <10% → 0; 10–<20% → max magnitude 1; ≥20% → max magnitude 2.
• 4.22 Direction confirmation: CONFIRMED_POSITIVE_INDUSTRY only when econ point +, plain ETF +, ratio shows RS (mirror for negative). On disagreement: preserve the economic point, mark MARKET_NOT_CONFIRMING, reduce confidence, add/reverse nothing.
• 4.23 Mixed regional exposure: if + and − regional contributions coexist → MIXED_IMPACT_FLAG TRUE, preserve each, never call it neutral just because the net is 0, display largest +/− causes.
• 4.24 Company geo-exposure alert: create when one-region exposure ≥20% and that region's final point is +2/−2 and data valid. +2 → OPPORTUNITY_ALERT, −2 → RISK_ALERT; state name, ticker, region, exposure %, type, activity, direction, confidence, duration. No trade.
• 4.25 Foreign company→industry spillover (all required): event official/exchange-filed/ company-verified; changes capacity/competition/pricing/technology/market-access; effect extends beyond the triggering company; ≥3 global peers or 1 approved industry ETF react as expected; relative move ≥1.00%; RVOL ≥1.50 in first 60 min. Alternative structural path: verified capacity added/removed ≥5% of global industry capacity within 24 mo. Output identifies trigger, country, event, industries helped/harmed, plain-ETF & ratio reactions, other causes, temp vs structural, duration, confidence.
• 4.26 Spillover point: temp market spillover exposure 10–<20% → max 1; ≥20% → max 2; verified structural ≥5% → max 2; unverified/company-only → 0 industry points (company stays visible).
• 4.27 Confidence:
• DATA_COVERAGE_SCORE = valid regional weight coverage (0–100).
• INDICATOR_AGREEMENT = 100 if GDP/composite-PMI/IP share one non-zero sign; 75 if two agree & third 0; 50 if one+/one−/one 0; 25 if two conflict with the third; 0 if critical invalid.
• GLOBAL_ACTIVITY_AGREEMENT = 100 same sign as weighted regional growth; 50 if either 0; 0 if conflict.
• MARKET_CONFIRMATION = 100 econ+absolute+ratio all confirm; 75 econ+absolute confirm, ratios neutral; 50 mixed; 25 absolute vs ratio conflict; 0 markets contradict.
• CONFIDENCE_PCT = coverage×35% + indicator×30% + global-activity×20% + market×15%.
• Bands: ≥80% HIGH; 60–<80% MEDIUM; 40–<60% LOW; <40% INVALID.

Part 5 — Descriptive bands

• 5.1 Direction: +2 STRONG_OVERSEAS_TAILWIND | +1 OVERSEAS_TAILWIND | 0 MIXED_OR_NEUTRAL | −1 OVERSEAS_HEADWIND | −2 STRONG_OVERSEAS_HEADWIND.
• 5.2 Regional growth: STRONG_EXPANSION | EXPANSION | STABLE_OR_MIXED | CONTRACTION | STRONG_CONTRACTION | INVALID.
• 5.3 Industry confirmation: CONFIRMED_BULLISH | CONFIRMED_BEARISH | ABSOLUTE_UP_RELATIVE_LAG | ABSOLUTE_DOWN_RELATIVE_RESILIENCE | UNCONFIRMED | UNAVAILABLE.
• 5.4 Overseas risk: RISK_ON | RISK_OFF | MIXED | MIXED_WITH_FUNDING_STRESS | INVALID.
• 5.5 Impact scope: REGION | COMPANY | SUB_INDUSTRY | INDUSTRY | SECTOR_WIDE | GLOBAL_INDUSTRY_SPILLOVER | MIXED | UNMAPPED.
• 5.6 Permission: UNCHANGED | REDUCED_CONFIDENCE | INVALID_DATA | INHERITED_PAUSE.

Part 6 — Helper conditions

VALID_CHINA_HANDOFF; VALID_REGIONAL_COVERAGE (≥80%); PMI_EXPANDING (≥50.5) / PMI_CONTRACTING (<49.5); ABSOLUTE_INDUSTRY_UP/DOWN; INDUSTRY_RELATIVE_STRENGTH/WEAKNESS; CONFIRMED_INDUSTRY_ BULLISH/BEARISH; MARKET_CONTRADICTS_ECONOMY; MEASURABLE_REGIONAL_EXPOSURE (≥10%); COMPANY_ALERT_EXPOSURE (≥20%); VERIFIED_COMPANY_SPILLOVER (all 4.25 pass); MIXED_REGIONAL_EFFECT; EXISTING_PAUSE_ACTIVE; ETF_MAP_VALID; MARKET_CONFIRMATION_UNAVAILABLE.

Part 7 — Regime decision order (stop at first match)

1 critical data invalid → 2 coverage <80% → 3 strong negative growth + risk-off → 4 strong positive growth + risk-on → 5 moderate negative → 6 moderate positive → 7 econ vs markets contradict → 8 +/− regional effects coexist → 9 neutral. Final: STRONG_OVERSEAS_TAILWIND | OVERSEAS_TAILWIND | MIXED_OVERSEAS_CONDITIONS | OVERSEAS_HEADWIND | STRONG_OVERSEAS_HEADWIND | INVALID.

Part 8 — Short-term / event overlays

• Economic-surprise: starts at release, lasts 1 completed US session, max ±1, never rewrites base, never crosses zero.
• Overseas market: recalcs while markets open, uses absolute + ratio separately, changes confidence, may move the point one level toward zero on direct contradiction, adds no new point.
• Foreign-company spillover: starts only after 4.25 passes; temporary price component expires after 5 completed relevant sessions; structural capacity effect persists until capacity/pricing/ technology/market-share evidence changes; triggering-company info stays separately visible.

Part 9 — Score/impact rules (summary)

Region = 40% composite PMI + 30% GDP + 30% IP. Base = 70% weighted regional growth + 30% global activity. China contributes exactly once via Module 7. Surprise may adjust ±1; contradiction may move ±1 toward zero. Industry economic exposure is separate from market confirmation; plain ETF and ratio never substitute for each other; no confirmation point is added to the macro score; all final points capped −2…+2.

Part 10 — Handoffs and double-counting controls

• Module 7 (China): owns detailed China; M6 consumes completed results, contributes once, never recalculates.
• Module 5 (policy/trade): owns the original US policy; M6 doesn't re-score it, but may score a separately verified overseas/global-industry spillover.
• Shipping/freight: pass trade-volume, regional import/export weakness, supply-chain exposure, shipping-sensitive effects; M6 keeps only the broad growth effect.
• Yield/financial-conditions: pass regional yields, changes, sovereign spreads, funding stress; M6 makes no final yield-direction score.
• Currency: pass regional currency direction, 5-session change, EMA status, conflicts; M6 uses currency as confirmation only.
• Sector-rotation / relative-strength: pass plain-ETF + ratio direction + confirmed status; M6 uses them only to confirm; the RS module owns final ranking.
• Output 2: pass final direction, regional contributions, separate China contribution, sector/industry/sub-industry impacts, plain-ETF + ratio status, company alerts, spillovers, confidence/risk. Output 2 must not count the same overseas-growth effect twice.

Part 11 — Hard overrides

1. Existing pause — preserve; M6 may not cancel/weaken.
2. Critical data failure — affected point 0, confidence INVALID, valid unaffected outputs remain, no trade approved.
3. Missing China handoff — China contribution 0, PARTIAL_EXCLUDING_CHINA, confidence not HIGH, other regions continue.
4. ≥2 funding-stress regions — risk-on prohibited, market confirmation can't raise confidence, no automatic bearish reversal.
5. Invalid industry map — primary→backup→peer group→UNAVAILABLE (economic exposure stays visible, no confirmation adjustment).

Part 12 — Execution order

Read run-control → preserve pauses → validate economic/forecast/market/IMF/ETF/exposure/Module-7 data → build scored universe → GDP → composite PMI → mfg/services → IP → regional raw growth → EM-ex-China → overseas IP → trade → global activity → weighted regional growth (China once) → base score → surprise overlay → regional absolute/relative equity → industry plain-ETF → ratio → combined confirmation → bond → currency → overseas risk → market confirmation/contradiction → sector/industry geo-exposure → mark mixed → company alerts → spillovers → confidence → hard overrides → handoffs → Output 1 → trade approved FALSE → order generated FALSE.

Part 13 — Non-calculative constraints

M6 must not: recalculate China; re-score a Module-5 policy; produce the final global-risk/currency/ yield regime or final sector-RS ranking; treat a rising ratio as proof the ETF is up (or vice versa); select a stock; generate a long/short trade; set entry/stop/target; place/close/reverse a position; change a price-based stop; cancel another module's pause; make the final sector selection.

Part 14 — Required Output 1

Structured output: identification (module no/name, version, UTC+ET, run type, provider list); reliability (DATA_STATUS, valid coverage %, China-handoff status, industry-map status, missing/stale/conflict lists, partial status, failure reason, next recalc); per-region results (name, weight, GDP value/score, composite/mfg/services PMI value+condition, IP change/score, raw growth, final point, status, initial/revised); China contribution (final point, weight, weighted contribution, growth/equity status, confidence, data status, sector/industry handoff); global activity (overseas IP change/point, trade change/point, raw score, point, descriptive-only flag); surprise (actual, consensus, raw/standardised, status, overlay point/expiry); overseas market confirmation (per-region absolute/relative status, bond & currency condition lists, risk vote total, risk status, market-confirmation status); per-industry confirmation (id, primary/backup, plain-ETF price/EMA/return/status, ratio value/EMA/return/ status, combined confirmation, peer status, map validity); final outcome (base raw/point, surprise-adjusted, final point/status, confidence %/level, risk level, permission); sector/industry/company effects (helped/harmed, final impact lists, regional contribution list, mixed flags, company geo-exposure alerts, spillover alerts); handoffs (per destination: raw/calculated fields, status, fields-not-scored, double-counting control); trading-bot status (final decision NOT_PRODUCED_BY_MODULE_6, trade approved FALSE, order generated FALSE, stock selected FALSE, final sector selected FALSE); audit (formula/threshold/regional-weight lists, industry-map version + revision history, provider list, excluded data, revision + overlay history, next recalc).

Part 15 — Items not fully defined in the source (named ceilings)

1. Provider names — not selected. Treatment: use field definitions + source hierarchy + freshness + consistency rules. (IMPLEMENTATION CONFIG — NOT A CALC GAP.)
2. ETF/peer maintenance — future closures/mandate changes unknowable. Treatment: quarterly + event-triggered map review; primary → backup → peers. (SOURCE-MAINTENANCE RULE.)
3. Relative-strength ratio thresholds — not in source. Proposed: ±2.00% over 5 sessions + 20-session ratio EMA for strong RS/RW. (PROPOSED DETERMINISTIC RESOLUTION.)
4. Confidence weights — not in source. Proposed: 35% coverage, 30% indicator agreement, 20% global-activity agreement, 15% market confirmation. (PROPOSED DETERMINISTIC RESOLUTION.)

---

Section Two — Implementation options and concerns

(Beyond Markets engineering analysis, not part of the normative spec.)

How it relates to our system

Module 6 is the overseas-growth member of the same deterministic auto-trading family as Modules 2–5 — a paradigm distinct from the current LLM report pipeline. It is the quantitative sibling of what our macro_outlook / market_brief reports and the dashboard's macro/liquidity snapshot already do qualitatively.

Module 6 concept	Beyond Markets today	Fit
Method: deterministic rules + fixed thresholds	11 LLM agents, prose reports	opposite paradigm
Output: −2…+2 + handoff packets	human-read markdown	no scoring layer
Regional GDP/PMI/IP scoring	FRED partial; no Eurostat/ONS/METI/S&P-PMI ingestion	partial data gap
Global trade/IP (CPB World Trade Monitor)	not integrated	data gap
Consensus + economic-surprise index	not integrated (licensed)	data gap
Industry ETF + ETF/SPY ratio confirmation	Polygon/Massive aggregates present	buildable today
Company regional revenue exposure	SEC filings + FMP (coarse)	partial
China handoff (Module 7)	doesn't exist	hard dependency
Consumer: auto-trader + Output 2	swing pipeline → Alpaca; no aggregator	adjacent, not wired

Deterministic vs LLM — the decision

Same split as Module 5, but M6 is more deterministic: almost every input is already a number (GDP/PMI/IP/trade/prices), so the LLM's only real job is the foreign-company spillover front-end (Calc 13: "is this event official? do ≥3 peers react?"). Everything else is arithmetic and must stay deterministic — reproducible, auditable (Part 14 FORMULA/ THRESHOLD lists + revision history), and cheap on the 2-minute market-hours hot path. A model never emits the final score. Intended boundary: agents/skills/module6_extract.md for spillover extraction only → deterministic scorer for Parts 4/9/11.

Better data fit than Module 5, but not zero-gap

M5 needed ~9 government-legal feeds we have none of. M6's hardest inputs are macro data, and we already own the market-confirmation half (Polygon/Massive) plus some macro (FRED). Still missing: the official regional releases (Eurostat, UK ONS, Japan Cabinet Office/METI), S&P Global PMI, CPB World Trade Monitor, IMF PPP weights, and a licensed consensus/surprise feed. Part 2.2/15.1 deliberately treat provider selection as implementation config — so the spec is complete without vendor names, but real regional points can't flow until those feeds exist.

Concerns

• Hard Module 7 dependency. No China module → spec's own fallback is PARTIAL_EXCLUDING_CHINA, confidence capped below HIGH. M6 can't be fully correct until M7 is built.
• Presumes the sibling ecosystem. Handoffs (Part 10), inherited pauses (Part 11.1) and Output 2 assume modules and an auto-trader we don't have — build as placeholders.
• Two open numeric decisions (Part 15.3 ±2% ratio threshold, Part 15.4 confidence weights) are proposed, not sourced — same class as M5's "NOT DEFINED" ceilings; lock them before real scoring.
• The prescribed ETF map rots (Part 2.11) — a dead/merged/mandate-changed ETF silently mis-scores an industry without the quarterly/event-driven review.
• Deploy: server-side (routes/module6.py → snapshots/, .data/); ships via the normal rsync; snapshots/ and .data/ are already deploy-excluded (server-owned, generated at runtime).

Two build paths

• Path A (light, recommended first): build only the industry ETF + ETF/SPY ratio confirmation engine (Calcs 12–14) over existing Polygon/Massive data using the Part 2.10 map, scored by a small deterministic Python function. Isolated, testable, needs no new vendors, and also feeds a future sector-rotation view — independent of both the regional-macro ingestion and Module 7.
• Path B (faithful): the full deterministic engine + regional-macro ingestion (Eurostat/ ONS/METI/CPB/IMF/consensus) + the Module 7 China dependency + Output 2, matching the sibling modules. Large, and it presumes the auto-trading half we don't yet have.

Recommendation: LLM extraction (spillover only) → deterministic scoring. Start with Path A (the ETF/ratio slice), lock the Part 15.3/15.4 numbers, and keep every regional-macro feed and the Module 7 handoff as named ceilings until they exist.