electoral-domain-methodology.md

May 24, 2026 · View on GitHub

Hack23 Logo

📕 Electoral & Domain Methodology

📊 Family D — Lens-Specific Analytical Depth
🎯 Current-cycle election · Next-cycle election · Voter Segmentation · Coalition Mathematics · Historical Parallels · Media Framing · Implementation Feasibility · Forward Indicators

Owner Version Effective Date Classification

📋 Document Owner: CEO | 📄 Version: 1.4 | 📅 Last Updated: 2026-05-01 (UTC) 🔄 Review Cycle: Quarterly | ⏰ Next Review: 2026-07-21 🏢 Owner: Hack23 AB (Org.nr 5595347807) | 🏷️ Classification: Public

🎯 AI-FIRST Methodology Card

🚦 Read this card before writing a single paragraph. It names the artifact this methodology owns, the gate check it satisfies, the evidence-density target it must hit, and the Pass-1 / Pass-2 discipline required by .github/copilot-instructions.md §5 (AI-FIRST Quality Principle).

FieldValue
PurposeFamily D — step-by-step production of the 7 electoral & domain-lens artifacts (election-2026, voter-segmentation, coalition-mathematics, historical-parallels, media-framing, implementation-feasibility, forward-indicators).
InputsFamily A synthesis; Riksdagen voting records; SCB segment cuts; Sainte-Laguë seat math; Statskontoret reports
OutputsAll 7 Family D artifacts
Owning artifact(s)All 7 Family D artifacts
Owning gate checkCheck 8 (Family D structure: ≥ 10 dated forward indicators, seat-count table in coalition-mathematics, Statskontoret citation in implementation-feasibility when an agency is named)
Citation density target≥ 1 evidence anchor per coalition row; ≥ 1 dated source per forward indicator; ≥ 1 SCB segment cut per voter-segmentation cohort
Banned phrasesEnforced via political-style-guide.md §Machine-readable banned-phrase list
Threshold sourcereference-quality-thresholds.json → thresholds[articleType][artifact] (fallback defaults.coreArtifactFloor)

✅ Pass-1 checklist (creation — minimal viable artifact)

  • coalition-mathematics.md has a seat-count table with Ja/Nej/Avstår or party-to-seats mapping
  • forward-indicators.md has ≥ 10 dated indicators across the 4 horizon sections
  • implementation-feasibility.md cites statskontoret.se (or none found) when a recognised agency is named
  • Produce every required sub-section listed in the owning template
  • Add ≥ 1 evidence anchor (dok_id, vote id, named MP, or primary-source URL) per analytical claim
  • Apply the correct WEP confidence band for the run's horizon (72h / week / month / quarter / year / cycle)
  • Include ≥ 1 themed Mermaid diagram with style … or themeVariables config (where structurally meaningful)
  • Cross-link the relevant template under analysis/templates/ and the gate check it satisfies

🔁 Pass-2 checklist (read-back & improve — AI-FIRST mandatory)

  • Re-run Sainte-Laguë for any new poll vintage; show working in coalition-mathematics
  • Add ≥ 1 historical parallel with date + outcome to historical-parallels.md
  • Verify media-framing reflects ≥ 3 outlets across the political spectrum
  • Re-read the file end-to-end; flag every claim that lacks an evidence anchor and add one
  • Replace every banned phrase listed in political-style-guide.md §Machine-readable banned-phrase list with an evidence-anchored alternative
  • Tighten WEP language: never above likely without ≥ 3 cycle-aged sources for year/cycle horizons
  • Strengthen Mermaid (color-coded style … directives, themeVariables, ≥ 5 nodes where the structure admits it)
  • Add ≥ 1 second-order effect, cui-bono note, or counterfactual where the artifact admits one
  • Verify citation density meets the per-file target below and the gate's evidence-density rules

🟢 Exemplar (good — pattern-match this)

(forward indicator) "T+30 — H902FiU1 budget final vote (2026-05-22): lead indicator = Ja votes from L bench; expected baseline 16/16. Disconfirming evidence: ≥ 2 L abstentions → coalition stress. Source: kammarens kalender (riksdagen.se)."

🔴 Anti-exemplar (failure mode — never ship this)

(failure mode) "Future events may shift the political landscape. The coalition's stability remains uncertain." — no dates, no indicators, no dok_id, banned phrases.

🔄 Tradecraft Anchors

ElementValueReference
F3EAD StageANALYZE → DISSEMINATEThis methodology applies domain lenses (electoral, historical, media, implementation, forward-watch) to enrich intelligence products
PIRs Servedelection-2026-analysis.md (current-cycle) serves PIR-6 (Election Integrity); coalition-mathematics.md serves PIR-1 (Coalition Stability); implementation-feasibility.md serves PIR-5 (Fiscal Trajectory), PIR-7 (Democratic Norms)See political-style-guide.md §PIR/EEI Catalog
Admiralty FloorPolling data requires ≥[B2] (named pollster, date, sample); historical parallels require ≥[A1] (official records from Riksdag archive)See political-style-guide.md §Admiralty Code
WEP Requirementelection-2026-analysis.md (alias: election-cycle-analysis.md) seat projections with WEP probability; forward-indicators.md triggers with WEP likelihood; scenario probabilities must sum to 100%See political-style-guide.md §WEP + ODNI
ICD 203 GateStandard 5 (customer relevance — forward indicators), 6 (logical argumentation — coalition mathematics), 9 (visual information — all files)See political-style-guide.md §ICD 203
SAT(s)Morphological (election-2026, coalition-mathematics); Outside-In Thinking (historical-parallels, voter-segmentation, media-framing); Premortem Analysis (implementation-feasibility); Indicators and Signposts (forward-indicators)See political-style-guide.md §SATs

🔀 Multi-cycle electoral lens

This methodology is cycle-parameterised. Two separate fields govern cycle-anchor semantics — they must not be conflated:

  • electionCycleAnchor (analysis/article-types.json → types[*].electionCycleAnchor) — the article-type registry field; one of current | next | both | none. Set per article type and consumed by workflows to decide which anchor(s) to generate.
  • cycleAnchor (scripts/horizon-context.ts → horizonContext(typeId, articleDate).cycleAnchor) — the date-derived active cycle; only current | next. Resolved at run time from analysis/article-types.json → electionCycles by activeCycleAnchor(articleDate) — returns next when articleDate falls inside the next cycle's start..end window, current otherwise.

Anchor enum

FieldValueMeaning
electionCycleAnchor (article-type registry)currentArticle type is anchored to the in-progress mandate cycle
electionCycleAnchor (article-type registry)nextArticle type is anchored to the upcoming mandate cycle
electionCycleAnchor (article-type registry)bothDual-anchor run — produces artifacts for both cycles simultaneously; used during the ±30-day rollover window
electionCycleAnchor (article-type registry)noneArticle type does not use an election-cycle lens (e.g., realtime-monitor on a non-electoral event)
cycleAnchor (HorizonContext, date-derived)currentarticleDate falls before the next election anchor; in-progress cycle is authoritative
cycleAnchor (HorizonContext, date-derived)nextarticleDate falls on or after the next election anchor; upcoming cycle is authoritative

Swedish election-cycle anchor dates

Cycle IDLabelStartEnd
tido-2022Tidö Mandate (2022–2026)2022-09-112026-09-13
post-2026Post-2026 Mandate (2026–2030)2026-09-132030-09-08
(future — not yet in registry)Post-2030 Mandate (2030–2034)2030-09-08(TBD)

The authoritative cycle registry lives in analysis/article-types.json → electionCycles. These dates are consumed by scripts/horizon-context.ts → activeCycleAnchor() and must not be hardcoded in any analysis template.

Rollover trigger

The cycle rolls over when ARTICLE_DATE is within ±30 days of the next election anchor. Activation is determined by the ext module's own shell predicate (comparing $ARTICLE_DATE against analysis/article-types.json → electionCycles.next.start), not by horizonContext() — workflows do not currently call tsx scripts/horizon-context.ts. The pseudo-code below illustrates the logic; the authoritative implementation is in .github/prompts/ext/cycle-rollover.md §1:

# Ext-module shell predicate (simplified)
DAYS_DELTA=$(( ($(date -d "$ARTICLE_DATE" +%s) - $(date -d "$ELECTION_DATE" +%s)) / 86400 ))
cycleRolloverActive = abs(DAYS_DELTA) <= 30

When active, the ext module applies the phased dual-anchor behaviour defined in §Cycle-rollover playbook below. It does not mutate analysis/article-types.json; electionCycleAnchor in the registry stays whatever the operator has set and is flipped separately (T+45 to T+60 via a CEO-approved Normal change).

📋 Cycle-rollover playbook (date-triggered)

Full procedure: .github/prompts/ext/cycle-rollover.md. The table below is an inline summary; always defer to the ext module for the authoritative runbook.

RuleDetail
Activation windowARTICLE_DATE within ±30 days of analysis/article-types.json → electionCycles.next.start
T-30 → T-1Both anchors generated; next/ artifacts marked [provisional pre-election]
T+0 (election day)Emergency single-run via news-realtime-monitor; no cycle flip until results are locked
T+1 → T+30Both anchors generated; current/ artifacts gain # 📜 Mandate retrospective H1; next/ upgraded from provisional
T+31 → T+45Only next/ regenerated; current/ frozen as historical record
Filename rename rulePost-rollover, workflows write election-cycle-analysis.md (the cycle-agnostic name); election-2026-analysis.md remains a valid alias indefinitely via FILENAME_ALIASES in scripts/render-lib/aggregator/order.ts. When both files exist in the same folder the aggregator de-duplicates by AGGREGATION_ORDER — election-2026-analysis.md appears first in that list, so the legacy filename is preferred at render time; write only one name per run to avoid the ambiguity
Carry-forward ruleMandate-fulfilment scorecard, KU reprimands ledger, coalition cohesion trajectory archived to analysis/cycles/<cycle-range>/; PIRs with inheritsCycle: true carry forward verbatim
PIR archival ruleCycle-scoped PIRs receive status: "archived", archivedReason: "cycle-rollover-<election-date>", and a successor where applicable; see ext module §4
Operator flipRegistry flip (electionCycleAnchor in analysis/article-types.json) is a CEO-approved Normal change under Change_Management.md; typically T+45 to T+60 post-election once government formation is complete
Audit trailEvery activation appends a row to analysis/cycles/rollover-log.md. Note: analysis/cycles/ does not exist in the repo yet — operators must mkdir -p analysis/cycles/ as part of the manual runbook (see ext module §5); automated creation is deferred to the planned scripts/cycle-rollover.ts

📈 Cycle-trajectory artefact

The 24th artifact for the news-election-cycle workflow is cycle-trajectory.md. It provides a multi-year trend assessment of Swedish political-economic trajectory across the full 4-year mandate.

  • Template: analysis/templates/cycle-trajectory.md
  • Produced by: news-election-cycle only — all other workflows must not write this file
  • Save path: analysis/daily/${ARTICLE_DATE}/election-cycle/${CYCLE_ANCHOR}/cycle-trajectory.md
  • Blocking gate: required by .github/prompts/05-analysis-gate.md § long-horizon checks
  • Content: ICD 203 BLUF + WEP per year + horizon bands T+1y / T+2y / T+5y, combining SCB national-accounts trajectory, IMF WEO multi-vintage projections, and Riksdag throughput metrics (vote counts, committee productivity, KU reprimands)
  • Cycle-rollover rewind: on registry flip, this file's content rewinds to a T+0 baseline (empty trajectory for the opening cycle); subsequent runs accrete evidence organically — see ext module §3.3

🎯 Purpose

Family D supplies issue-specific analytical lenses that sharpen every workflow across electoral cycles, voter segments, coalition arithmetic, historical echoes, the media environment, delivery questions, and forward watch-lists.

These products are core — every run produces all 7. The output set is stable across morning, evening, realtime, weekly, and monthly workflows. What adapts per run is item-level depth (tier L1 → L3) and the substance inside each file — not whether the file exists.

Core — every run produces all 7

TemplateBehaviour on a light-event dayBehaviour on a P0-dense day
election-2026-analysis.md (alias: election-cycle-analysis.md) — Note: the alias is filename-only; the underlying template analysis/templates/election-2026-analysis.md remains 2026-scoped (pre-election window 2025-10 → 2026-09) until a cycle-parameterised template lands. Post-rollover, workflows write election-cycle-analysis.md but the content is generated from the same template with the cycle context injected at runtime via electionCycleAnchor.Seat-projection delta vs. last poll + which parties crossed the 4 % threshold; for the current cycle it tracks the campaign; for the next cycle it tracks post-election government-formation contextFull seat projection + coalition viability + campaign-phase alignment of every P0 to the active cycle; electionCycleAnchor (article-type registry) determines which cycle lens is applied
voter-segmentation.mdBaseline segment positions (5 axes: age, geography, education, income, incumbency)Per-document segment impact table with ≥5 cohorts and quantified swing estimates
coalition-mathematics.mdCurrent seat map + pivotal-vote reference + confidence-vote arithmeticScenario branching: each contested vote run through Sainte-Laguë, pivotal-actor detection, SD-Tidöbloc-opposition matrix
historical-parallels.md (variant: historical-baseline.md)Closest baseline precedent ≤ 40 years with similarity score, or explicit "no-precedent" findingFull parallel with outcome, stakeholder behaviour, and Bayesian base-rate update
media-framing-analysis.mdPer-party + per-press-quadrant framing of the day's lead story, plus longitudinal-frame deltaPer-document frame audit with quoted examples and platform-by-platform amplification scores
implementation-feasibility.mdAudit of in-flight delivery backlog (IT, budget, regulatory, workforce) when no new bill landsFull RACI + milestone + risk-register for each new delivery obligation
forward-indicators.md≥10 indicators across 4 horizons (72 h / week / month / election) refreshed every runIndicators + thresholds + trigger-to-hypothesis mapping for every P0/P1
flowchart LR
    classDef core fill:#1565C0,stroke:#0D47A1,color:#FFFFFF
    classDef electoral fill:#E3F2FD,stroke:#1565C0,color:#0D47A1
    classDef historical fill:#FFF8E1,stroke:#FFC107,color:#3E2723
    classDef media fill:#F3E5F5,stroke:#7B1FA2,color:#311B92
    classDef impl fill:#E8F5E9,stroke:#4CAF50,color:#1B5E20
    classDef fwd fill:#FFF3E0,stroke:#FF9800,color:#BF360C

    T[Family A/B complete<br/>→ Family D core run]:::core

    D1[election-2026-analysis.md<br/>alias: election-cycle-analysis.md]:::electoral
    D2[voter-segmentation.md]:::electoral
    D3[coalition-mathematics.md]:::electoral
    D4[historical-parallels.md]:::historical
    D5[media-framing-analysis.md]:::media
    D6[implementation-feasibility.md]:::impl
    D7[forward-indicators.md]:::fwd

    T --> D1
    T --> D2
    T --> D3
    T --> D4
    T --> D5
    T --> D6
    T --> D7

🗳️ Part 1 — Current-cycle Election Analysis (election-2026-analysis.md)

Filename variants: election-2026-implications.md — identical structure. The cycle-agnostic alias election-cycle-analysis.md maps to the same canonical artifact via FILENAME_ALIASES in scripts/render-lib/aggregator/order.ts; when both filenames exist in a folder the aggregator de-duplicates and emits only the first-encountered name. Post-2026-rollover, workflows may write election-cycle-analysis.md as the generalized name; election-2026-analysis.md remains a permanently valid alias.

Purpose

Translate today's policy activity into electoral consequences for the active-cycle Riksdag election: seat trajectories, bloc viability, mandate strength, and pre-election narrative positioning. The active cycle is determined by cycleAnchor from scripts/horizon-context.ts → horizonContext(). During the ±30-day rollover window, both current-cycle and next-cycle sections are produced (see §Cycle-rollover playbook).

Input

  • synthesis-summary.md (current events)
  • Opinion polling (SIFO, Novus, Demoskop, Sentio) — use only published, dated figures
  • 2018 + 2022 election results (for baseline trajectory); 2026 results once available for next-cycle analysis
  • Seat-allocation rules (Sainte-Laguë modified + 4 % threshold + 12 % regional)

Output — required structure

  1. Bloc standing table — for Tidöavtalet (M, KD, L, SD) vs Opposition (S, V, MP, C) vs cross-bloc:
    • Current poll average · 30-day change · 90-day change · Last election
  2. Seat-projection table — per party, showing polling-average seat translation with 4 % threshold risk flag
  3. Threshold-risk watch — any party within ±1.0 pp of 4 %
  4. Narrative positioning — how today's event helps / hurts each party's election narrative
  5. Mobilisation implications — turnout implications (young-voter, first-time, non-voter)
  6. Seat-projection Mermaid — color-coded bar by party
  7. Bloc-standing Mermaid — color-coded stacked bar

Required Mermaid — bloc standing

graph LR
    classDef tido fill:#1565C0,stroke:#0D47A1,color:#FFFFFF
    classDef opp fill:#D32F2F,stroke:#B71C1C,color:#FFFFFF
    classDef cross fill:#FFC107,stroke:#F57F17,color:#3E2723

    M[M<br/>19.1 %<br/>🟦 Tidö]:::tido
    SD[SD<br/>18.4 %<br/>🟦 Tidö]:::tido
    KD[KD<br/>4.8 %<br/>🟦 Tidö]:::tido
    L[L<br/>3.9 %<br/>⚠️ threshold]:::tido
    S[S<br/>31.2 %<br/>🟥 opposition]:::opp
    V[V<br/>7.1 %<br/>🟥 opposition]:::opp
    MP[MP<br/>4.6 %<br/>🟥 opposition]:::opp
    C[C<br/>7.3 %<br/>🟧 cross-bloc]:::cross

Quality gate

  • All 8 parties covered
  • Threshold risk explicit for any party <5 %
  • Polling averages cite ≥2 pollsters with dates
  • Seat math uses Sainte-Laguë modified
  • Narrative-positioning section neutral (equal depth across parties)

👥 Part 2 — Voter Segmentation (voter-segmentation.md)

Purpose

Identify which voter segments are materially affected by today's events and how each segment's pre-existing political alignment interacts with the new signal.

Input

  • SCB demographic microdata (age × region × income × education × sector)
  • Polling cross-tabs (published segment-level data only)
  • Prior election turnout + vote-by-segment data (SCB Statistiska Meddelanden)

Output — required structure

  1. Segment panel — for each of: Young-urban (18–34, storstad), Mid-career family (35–54), Retired (65+), Rural / glesbygd, High-income (top quintile), Low-income (bottom quintile), Secondary-sector workers, Public-sector workers:
    • Current political alignment · Policy exposure to today's events · Likely response
  2. Segment-issue Mermaid — color-coded quadrant (alignment × exposure)
  3. Mobilisation index — which segments are most "move-able" given the current evidence
  4. Disaggregation notes — where SCB data permits further cuts (e.g. foreign-born, women 35–44, etc.)

Required Mermaid — segment quadrant

quadrantChart
    title Segment alignment vs policy exposure
    x-axis "Low policy exposure" --> "High policy exposure"
    y-axis "Left-bloc leaning" --> "Right-bloc leaning"
    quadrant-1 "High-exposure right"
    quadrant-2 "High-exposure left"
    quadrant-3 "Low-exposure left"
    quadrant-4 "Low-exposure right"
    "Young-urban 18-34": [0.72, 0.28]
    "Retired 65+": [0.65, 0.71]
    "Rural / glesbygd": [0.41, 0.68]
    "Public-sector": [0.78, 0.35]

Quality gate

  • ≥8 segments covered
  • Every segment has SCB-cited demographic magnitude
  • Segment-issue quadrant includes all covered segments
  • Mobilisation index explains movability with evidence
  • Privacy rule: no segment drawn to <1000 persons to avoid re-identification

Segment privacy threshold — rationale (binding)

The <1000 persons threshold is not a stylistic preference; it is a derived control rooted in three converging concerns:

  1. GDPR Article 4(1) re-identification risk. SCB-published microdata is anonymised at the aggregate level. A segment drawn to fewer than ~1000 persons in a country of 10.5M can become indirectly re-identifying when crossed with a second variable (e.g., "rural drivers in Norrbotten under 35 with K-bil sport") — the cross-tabulation can collapse to a handful of households. The Article 29 Working Party Opinion 05/2014 on Anonymisation Techniques names singling-out, linkability, and inference as the three risks; sub-1000 micro-cells defeat all three. SCB's own micro-data publication policy follows a related minimum-cell-size principle.
  2. Statistical confidence. For a binary outcome (e.g., "voted SD") the 95 % CI half-width at p = 0.5, n = 1000 is ±3.1 pp; at n = 500 it widens to ±4.4 pp; at n = 100 to ±9.8 pp. Below n = 1000 the segment delta is generally smaller than the confidence interval, meaning the analysis cannot distinguish signal from sampling noise. Reporting such a segment as a quantitative finding is innumerate.
  3. Editorial avoidance of small-group stigmatisation. Segments drawn small enough to be re-identifiable are also drawn small enough to invite the editorial sin of using a single named individual as a stand-in for the whole group. The 1000-person floor enforces narrative restraint — talk about "rural Norrbotten residents under 35" not "Karin from Boden, 28."

Implementation rule: when SCB cross-tabs collapse below 1000, collapse one axis (e.g., merge two adjacent age bands or two neighbouring counties) until the cell count clears 1000. Document the collapse in the segment definition; never report the sub-1000 cell. If the collapsed segment is still analytically interesting at n ≥ 1000, use it; if not, drop it from the analysis.

🧮 Part 3 — Coalition Mathematics (coalition-mathematics.md)

Purpose

Compute the arithmetic of government formation, confidence, and policy coalition — which combinations clear 175 seats, which survive a misstatement of confidence, and which emerge as issue-specific coalitions for today's vote.

Input

  • Latest Riksdag seat distribution (349 seats)
  • Current government composition
  • Party-platform compatibility matrix (based on voting history from search_voteringar)
  • Today's vote results when applicable

Output — required structure

  1. Seat-count table — per party, current seats
  2. Majority viability matrix — all plausible 3–5 party coalitions with seat totals, tolerance to loss of 1 party, ideological coherence score
  3. Confidence arithmetic — who can block a confidence vote, who can trigger it
  4. Issue-specific coalitions — for today's top 3 items, which coalition actually materialised (from voting data)
  5. Cohesion indicators — party vote-split percentages on today's items
  6. Coalition-viability Mermaid — color-coded by viability status

Required Mermaid — coalition viability

graph TB
    classDef maj fill:#4CAF50,stroke:#1B5E20,color:#FFFFFF
    classDef min fill:#FFC107,stroke:#F57F17,color:#3E2723
    classDef blocked fill:#D32F2F,stroke:#B71C1C,color:#FFFFFF
    classDef hypo fill:#9E9E9E,stroke:#424242,color:#FFFFFF

    C1[M+SD+KD+L<br/>175 seats · current]:::maj
    C2[S+V+MP+C<br/>158 seats]:::min
    C3[S+C+MP<br/>132 seats]:::min
    C4[M+S grand coalition<br/>186 seats · hypothetical]:::hypo
    C5[M+SD only<br/>140 seats · blocked by L/KD exit]:::blocked

Quality gate

  • All coalitions summed to exact seat counts
  • Tolerance-to-defection column filled per coalition
  • Issue-specific coalitions cite dok_id and vote counts from search_voteringar

Worked example — Sainte-Laguë modified seat allocation

Sweden uses the modified Sainte-Laguë method with a 4 % national threshold and a 12 % regional threshold (a party scoring under 4 % nationally still gets seats from a constituency where it exceeds 12 %). The first divisor for any party is 1.4 (the modification), then 3, 5, 7, 9, … The seat goes to the party with the highest quotient at each round.

Worked example using a hypothetical 11-seat constituency (Stockholm county, simplified to illustrate the algorithm):

PartyVotes÷1.4 (seat 1)÷3 (after 1)÷5 (after 2)÷7÷9÷11Final seats
S90 00064 28630 00018 00012 85710 0008 1824
M75 00053 57125 00015 00010 714——3
SD55 00039 28618 33311 000———2
V30 00021 42910 000————1
MP12 0008 571—————1
L9 0006 429—————0 — under 4 %, no regional override

Algorithm walk-through: Round 1: highest quotient (S ÷ 1.4 = 64 286) → S gets seat 1. Round 2: re-divide S by 3 (= 30 000); highest of all current quotients is M ÷ 1.4 = 53 571 → M gets seat 2. Round 3: re-divide M by 3; SD ÷ 1.4 = 39 286 wins → SD gets seat 3. Continue until all 11 seats are awarded.

Key gotchas the AI must respect:

  1. The 1.4 modifier replaces the first division for every party — not 1.0, not 0.7. Using straight Sainte-Laguë (3-5-7-…) inflates small-party seat counts and is wrong.
  2. Threshold check happens before allocation: parties under 4 % nationally are excluded from the divisor table unless they cleared 12 % in this specific constituency.
  3. Tied quotients are broken by total vote count (higher wins); document the tie-break rule used in the analysis.
  4. Constituency seats vs. levelling seats (utjämningsmandat) are separate phases. Sweden has 310 fixed constituency seats + 39 levelling seats. The first phase distributes 310 by constituency-level Sainte-Laguë; the second phase reassigns 39 to make the national distribution proportional. Always state which phase is being modelled.
  5. Sources: seat counts from search_voteringar and Valmyndigheten election archives; vote totals from SCB / Valmyndigheten primary CSVs (never approximated).

A coalition-mathematics file that asserts a seat outcome without showing the Sainte-Laguë computation fails the gate; copy this table form (or link to a calculator script's output) and adapt the numbers.

Quality gate (continued)

  • Cohesion indicators include numeric vote-split percentages
  • Hypothetical coalitions clearly labelled

📜 Part 4 — Historical Parallels (historical-parallels.md)

Filename variant: historical-baseline.md — identical structure.

Purpose

Locate today's event within a named prior episode (≤40 years back) and learn from what happened next last time.

Input

  • Riksdag archive (search_dokument with historical rm parameters)
  • SOU (Statens offentliga utredningar) archive
  • Statskontoret reports when historical implementation, public-management or agency-capacity lessons are relevant
  • Academic political-history references where applicable

Output — required structure

  1. Prior-episode cards — ≥3 named parallels; per card:
    • Name + date · Actors · Policy domain · Outcome · Duration · Source
    • Similarity score (1–10) with justification
    • Difference callouts — what is not analogous
  2. Timeline Mermaid — color-coded timeline of prior episodes
  3. Lesson-extraction table — which lessons transfer, which do not
  4. Base-rate note — based on prior parallels, prior probability of each scenario outcome

Required Mermaid — timeline

timeline
    title Historical parallels for today's event
    section 1990s
        1994 : Maastricht pre-accession debate
        1999 : EMU referendum prep
    section 2000s
        2003 : Euro referendum
        2008 : Financial-crisis emergency budget
    section 2010s
        2013 : RUT-expansion opposition coordination
        2018 : Post-election 134-day formation crisis
    section 2020s
        2022 : Tidöavtalet ratification
        2024 : NATO ratification package

Quality gate

  • ≥3 named parallels with dates and sources
  • Similarity score present with justification
  • Difference callouts explicit (not every lesson transfers)
  • Base-rate note quantifies prior-probability observation
  • Timeline Mermaid covers full spread

📢 Part 5 — Media Framing & Influence-Operations Analysis (media-framing-analysis.md)

Purpose

Document the frames being used by each actor and major media outlet, and the manipulation / influence-operations layer riding on top of them, so readers can separate substantive content from strategic communication and from coordinated inauthentic behaviour. v2.1 (2026-05-01) extends v2.0 with a founding "no-neutral-media" doctrine: every cited outlet must appear in an Outlet Bias Audit (ownership · funding mix · board-appointment authority · documented editorial lean · Reuters Institute Trust score · PO/PON complaints · foreign-actor link), every algorithmic claim must name the platform's optimisation target, and Frame C / D are labelled establishment / centrist-consensus and public-broadcaster proceduralist respectively — never "neutral / impartial / balanced". Procedural neutrality (equal analytical depth across all 8 Riksdag parties; equal evidentiary discipline) is the only neutrality the platform claims.

Input

  • Riksdag MCP: search_anforanden, search_dokument, get_dokument (named-MP frame language; longitudinal frame record).
  • Swedish public press: DN, SvD, Aftonbladet, Expressen, SVT, SR, TV4, Dagens ETC, Kvartal, Nyheter Idag, Samhällsnytt, Riks, regional press sample.
  • International quality press for frame lineage: Reuters / AP / DPA / AFP / Politico EU / FT / NYT / Le Monde / Der Spiegel / Helsingin Sanomat / Aftenposten / Berlingske.
  • State-affiliated outlets — RT/Sputnik/RIA/TASS/CGTN/PressTV — used only as amplification fingerprint, never as factual source.
  • Public CIB / disinformation dossiers: EUvsDisinfo, EU DisinfoLab, Meta CIB removal reports, NATO StratCom COE, GLOBSEC Vulnerability Index, Reuters Institute Digital News Report, Freedom House Nations in Transit, SÄPO/MUST/FRA/EU-EEAS public statements.
  • Ownership / funding data: Nordicom Media Ownership Database (gu.se), Allmänhetens Pressombudsman public registry.

Output — required structure (v2.1)

  1. Frame package inventory — ≥ 3 frames (A government / B opposition / C establishment / centrist-consensus / D public-broadcaster proceduralist; Frame E foreign-overlay ONLY when a state-affiliated or coordinated foreign-amplification signal is observed in window). Per frame: name · carrier (actor / outlet) · keyword markers · first-use date · approximate share of coverage.
  2. Entman functions per frame — problem definition · causal attribution · moral evaluation · treatment recommendation; every cell traces to a dated quote / dok_id / anförande.
  3. Cognitive vulnerability map — bias exploited (Cialdini / Kahneman / Roozenbeek-van der Linden) · mechanism · inoculation lever; primary-literature citation per row.
  4. Manipulation indicators (DISARM) — verbatim T#### codes (Flooding T0049, Astroturfing T0086, Distort facts T0023, Amplify existing narrative T0118, Prepare assets impersonating legitimate entities T0099, Develop AI-generated text T0085, Develop AI-generated images / deepfakes T0088, etc.); [unconfirmed] flag for any indicator without ≥ 2 corroborations; explicit no-signal finding when nothing observed.
  5. Narrative-laundering chain — fringe → alt-media → politician amplification → mainstream → international, with first-observed UTC timestamps and reach estimates from public metrics; missing nodes labelled, never invented.
  6. Source ecology / Outlet Bias Audit table — every cited outlet listed with ownership group · funding mix percentages · board-appointment authority · documented editorial lean · Reuters Institute Trust score · PO/PON complaint history · foreign-actor link. No bias audit row = no citation. "Neutral / impartial / balanced / objective" labels rejected. Public-service broadcasters (SVT/SR/UR) carry a licence-fee + politically-appointed-board + institutional-editorial-culture caveat; commercial quality press (DN/Bonnier, SvD/Schibsted) carry an owner-ideology + audience-demographic caveat. Source: Nordicom, Reuters Institute, PO/PON, Förvaltningsstiftelsen, EUvsDisinfo, EU-EEAS / SÄPO.
  7. CIB ABCDE block — Actor · Behaviour · Content · Degree · Effect (Camille François 2020) populated for any state-attribution claim. Observation ≠ attribution; state-attribution requires ≥ 3 ABCDE indicators + published SÄPO/MUST/FRA/EU-EEAS/EUvsDisinfo reference.
  8. Algorithmic-amplification asymmetry — per-platform reach asymmetry; every row must declare the platform's optimisation target (engagement / watch-time / retention / ad-yield) and cite a documented partisan/emotional asymmetry (Huszár et al. 2022 PNAS for X; Rathje, Van Bavel & van der Linden 2021 for Facebook outrage-amplification; TikTok DSA transparency report; Ribeiro et al. 2020 / Hosseinmardi et al. 2024 for YouTube; NATO StratCom COE 2023 for Telegram). There is no neutral algorithm — uncited "the algorithm shows X" rejected at Pass 2.
  9. Comparative-international frame lineage — ≥ 2 cognates traced to prior jurisdictions (FR 2011 Camus replacement frame, HU 2010 Fidesz elite-vs-people, RU 2014 firehose, PL 2015 PiS courts-vs-democracy, IT 2018 Lega/M5S, US 2017 alt-right, IL 2023 judicial reform, MAGA 2016/2020/2024). "No international cognate" for a major story is suspicious and triggers re-do.
  10. Strategic-doctrine detection — match to public-doctrine catalogue (firehose-of-falsehood RAND PE-198 · doppelganger EU DisinfoLab 2022 · gish gallop · reflexive control · active-measures spillover NATO StratCom · interest-group capture · MAGA cognate populism). Detection ≠ attribution; ≥ MODERATE confidence floor for any 🟧/🟥 verdict.
  11. Frame lifecycle / longevity — phase (rising / peak / decay / sleeper / zombie) · half-life days · zombie probability · reactivation trigger; xychart-beta horizon block over 8+ points (T-7d / T-3d / Today / T+3d / T+7d / T+14d / T+30d / T+90d).
  12. RRPA impact conversion — Reach × Resonance × Persistence × Action with ≥ 1 dated action indicator per frame (poll move, petition signatures, donation flows, demonstration permit) or [no action signal yet].
  13. Counter-resilience plan (L1–L5) — prebunking · just-in-time inoculation · lateral-reading prompt · debunking truth-sandwich · algorithmic-friction / DSA Art. 40 escalation; per-frame layer assignment with rationale. Platform reports the ladder, never executes counter-framing.
  14. Forward watchlist — ≥ 5 dated triggers with WEP band + Admiralty grade.
  15. Longitudinal frame record — append today's snapshot so future runs can detect zombie reactivations.

Required Mermaid — frame competition (with foreign-overlay tier when applicable)

graph LR
    classDef gov fill:#1565C0,stroke:#0D47A1,color:#FFFFFF
    classDef opp fill:#D32F2F,stroke:#B71C1C,color:#FFFFFF
    classDef establishment fill:#FFC107,stroke:#F57F17,color:#3E2723
    classDef proceduralist fill:#7B1FA2,stroke:#4A148C,color:#FFFFFF
    classDef foreign fill:#212121,stroke:#000000,color:#FFFFFF

    F1[Government frame<br/>'fiscal responsibility'<br/>dominant]:::gov
    F2[Opposition frame<br/>'welfare erosion']:::opp
    F3[Establishment / centrist-consensus frame<br/>'expert-led fiscal'<br/>moderate<br/>NOT 'neutral' — Bonnier/Schibsted owner ideology +<br/>centrist-urban audience documented]:::establishment
    F4[Public-broadcaster proceduralist frame<br/>'institutional remit framing'<br/>NOT 'impartial' — politically-appointed board +<br/>documented editorial culture]:::proceduralist
    F5[Foreign-overlay frame<br/>'NATO trap'<br/>state-affiliated amplification<br/>⚠️ only if observed]:::foreign

    F1 -.contests.- F2
    F3 -.synthesises.- F1
    F3 -.synthesises.- F2
    F4 -.emergent.- F1
    F5 -.amplifies.- F2

Quality gate (v2.1)

  • Founding "no-neutral-media" doctrine honoured — no outlet, algorithm, or broadcaster labelled "neutral / impartial / unbiased / balanced / objective" in the body without ≥ 1 documented-bias citation in the same paragraph.
  • Outlet Bias Audit populated for every cited outlet — ownership group · funding mix · board-appointment authority · documented editorial lean · Reuters Institute Trust score · PO/PON complaint history · foreign-actor link. No bias audit row = no citation.
  • Public-service broadcaster caveat present — every reliance on SVT / SR / UR / BBC / ARD / NRK / DR / YLE carries the licence-fee + politically-appointed-board + institutional-editorial-culture caveat.
  • Algorithmic-asymmetry rows declare optimisation target — every platform row names the objective function and cites a documented partisan/emotional asymmetry.
  • Frame C / D label discipline — "establishment / centrist-consensus" and "public-broadcaster proceduralist" respectively, never "neutral / analytical / impartial".
  • ≥ 3 frames identified with keyword markers AND Entman 4-function decomposition.
  • Cognitive-vulnerability map cites primary literature per row.
  • DISARM TTP map populated with verbatim T#### codes OR explicit no-signal finding.
  • Narrative-laundering chain produced or absence justified by node-by-node inspection.
  • CIB ABCDE block produced for any state-attribution claim; observation ≠ attribution discipline applied.
  • Comparative-international frame lineage lists ≥ 2 cognates with year, jurisdiction, and source.
  • Strategic-doctrine detection table executed (Yes/No per row).
  • Frame Lifecycle / Longevity table + xychart-beta with ≥ 8 horizon points.
  • RRPA composite per frame with ≥ 1 dated action indicator or honest [no action signal yet].
  • Counter-resilience layer mapping (L1–L5) per frame with rationale; procedural neutrality preserved (equal analytical depth across all 8 Riksdag parties; equal evidentiary discipline; full ownership/funding/lean disclosure) — never claimed as "neutral truth" or "neutral source".
  • Forward watchlist ≥ 5 dated triggers with WEP and Admiralty grades.
  • Frame E discipline: included only with observed signal; otherwise marked absent explicitly.

🛠️ Part 6 — Implementation Feasibility (implementation-feasibility.md)

Purpose

Assess whether a proposed policy can actually be delivered given constitutional, legal, administrative, financial, and operational constraints.

Input

  • The target dok_id and its implementing-agency identification
  • Myndighet (government agency) capacity indicators — staffing, budget, prior delivery record
  • Riksrevisionen prior audits of the same agency / policy area
  • EU-law and ECHR compatibility constraints

Output — required structure

  1. Delivery-path table — which agency owns delivery, legal basis, budget source, first-effect date, steady-state date
  2. Feasibility scorecard — five dimensions × 0–10 score × evidence × confidence:
    • Legal feasibility · Administrative feasibility · Financial feasibility · Political feasibility · Operational feasibility
  3. Risk register — ≥5 delivery risks, each with likelihood × impact × mitigation
  4. Capacity gap analysis — staffing, IT, legal, training gaps identified
  5. Historical-delivery comparison — against a prior similar policy (e.g. RUT, ROT, A-kassa reform)
  6. Feasibility Mermaid — color-coded scorecard bar

Required Mermaid — feasibility scorecard

graph TB
    classDef high fill:#4CAF50,stroke:#1B5E20,color:#FFFFFF
    classDef med fill:#FFC107,stroke:#F57F17,color:#3E2723
    classDef low fill:#FF9800,stroke:#E65100,color:#FFFFFF
    classDef crit fill:#D32F2F,stroke:#B71C1C,color:#FFFFFF

    Legal[Legal feasibility<br/>8/10<br/>🟩 HIGH]:::high
    Admin[Administrative feasibility<br/>5/10<br/>🟧 MEDIUM]:::med
    Fin[Financial feasibility<br/>7/10<br/>🟩 HIGH]:::high
    Pol[Political feasibility<br/>4/10<br/>🟥 LOW]:::low
    Ops[Operational feasibility<br/>3/10<br/>⬛ VERY LOW<br/>bottleneck]:::crit

Quality gate

  • Delivery-path table identifies a specific myndighet
  • Scorecard scores are evidenced, not asserted
  • Risk register ≥5 rows with mitigations
  • Capacity-gap analysis quantified (FTEs, SEK, or time)
  • Historical-delivery comparison present with sources

🔭 Part 7 — Forward Indicators (forward-indicators.md)

Purpose

Pre-register observable indicators whose materialisation (or non-materialisation) will confirm or refute the current intelligence picture. This file is the platform's honesty mechanism — tomorrow's reality checks today's analysis.

Input

  • synthesis-summary.md · scenario-analysis.md · intelligence-assessment.md
  • Committee calendars (motionstider, utskottsmöten) from Riksdag calendar
  • Known government decision points over next 7 / 30 / 90 / 180 days

Output — required structure

  1. Indicator table — ≥10 indicators; per row:
    • Indicator · Horizon (7d / 30d / 90d / 180d) · Observation method · Trigger threshold · Confirms/refutes hypothesis · Confidence in prediction
  2. Time-horizon Mermaid — Gantt / timeline of expected signal dates
  3. Hypothesis-mapping table — which indicator maps to which ACH hypothesis from devils-advocate.md
  4. Null-hypothesis indicators — ≥3 indicators whose non-appearance carries signal
  5. Reliability-of-prediction note — how well previous forward-indicators files have scored (scoring is automated across runs)

Required Mermaid — forward timeline

gantt
    dateFormat YYYY-MM-DD
    title Forward indicators — next 180 days
    section 7-day
    Committee vote on prop 108 :crit, a1, 2026-04-22, 3d
    Opposition response filing  :a2, 2026-04-24, 4d
    section 30-day
    Myndighet report deadline   :b1, 2026-05-05, 7d
    Coalition meeting signal    :b2, 2026-05-10, 3d
    section 90-day
    Budget motion window open   :c1, 2026-06-15, 14d
    section 180-day
    Pre-election kickoff        :milestone, 2026-10-01, 1d

Quality gate

  • ≥10 indicators across all four horizons
  • Every indicator has an observation method (named source)
  • Observation thresholds are concrete (numeric or named event)
  • Null-hypothesis indicators present
  • Hypothesis mapping links back to devils-advocate.md

🛠️ Production Workflow — step-by-step

flowchart TD
    classDef core fill:#1565C0,stroke:#0D47A1,color:#FFFFFF
    classDef step fill:#E8F5E9,stroke:#4CAF50,color:#1B5E20
    classDef gate fill:#FFF8E1,stroke:#FFC107,color:#3E2723
    classDef out fill:#F3E5F5,stroke:#7B1FA2,color:#311B92

    TC[Family A + B complete<br/>→ produce all 7 Family D files]:::core

    E1["election-2026-analysis<br/>(alias: election-cycle-analysis)"]:::step
    E2[voter-segmentation]:::step
    E3[coalition-mathematics]:::step
    H[historical-parallels]:::step
    MF[media-framing-analysis]:::step
    IF[implementation-feasibility]:::step
    FI[forward-indicators]:::step

    G{Quality-gate<br/>per-template}:::gate
    O[Family D complete]:::out

    TC --> E1
    TC --> E2
    TC --> E3
    TC --> H
    TC --> MF
    TC --> IF
    TC --> FI

    E1 --> G
    E2 --> G
    E3 --> G
    H --> G
    MF --> G
    IF --> G
    FI --> G
    G -->|pass| O
    G -->|fail| TC

✅ Family-D Completion Checklist

  • All 7 Family D templates produced this run
  • All 8 parties covered where applicable
  • Polling, seat-math, and coalition-math use verifiable primary data
  • Historical parallels have similarity scores and difference callouts (or an explicit "no-precedent" finding with reasoning)
  • Media frames covered with equal depth across government/opposition
  • Implementation feasibility names a delivery agency
  • Forward indicators include null-hypothesis observations and ≥10 total
  • All outputs use canonical color palette and 5-level confidence scale
  • GDPR data-minimisation observed in voter-segmentation (no sub-1000 cohorts)

🔗 Template bindings

TemplateMethodology sectionCycle anchor
analysis/templates/election-2026-analysis.md — one template. Pre-rollover workflows write election-2026-analysis.md; post-rollover workflows write election-cycle-analysis.md. Both map to the same artifact via FILENAME_ALIASES in scripts/render-lib/aggregator/order.ts; when both exist the aggregator prefers election-2026-analysis.md (first in AGGREGATION_ORDER) — write only one name per run.Part 1 aboveelectionCycleAnchor (article-type registry)
analysis/templates/voter-segmentation.mdPart 2 aboveany
analysis/templates/coalition-mathematics.mdPart 3 aboveany
analysis/templates/historical-parallels.mdPart 4 aboveany
analysis/templates/media-framing-analysis.mdPart 5 aboveany
analysis/templates/implementation-feasibility.mdPart 6 aboveany
analysis/templates/forward-indicators.mdPart 7 aboveany
analysis/templates/cycle-trajectory.md§Cycle-trajectory artefact abovecurrent or next

📐 Cross-references to other methodology layers

🔐 ISMS Alignment

ControlHow this methodology satisfies it
ISO 27001 A.5.23 + A.5.34Voter-segmentation respects GDPR Art. 9 (political-opinion data) + re-identification threshold
ISO 27001 A.5.7Forward-indicators constitutes structured threat-intelligence horizon
NIST CSF ID.RA-6 (Risk responses)Implementation-feasibility lists mitigations
NIST CSF DE.CM-8Forward indicators = continuous monitoring definition
CIS 17.3 (Policy reviews)Historical-parallels maps prior policy cycles
GDPR Art. 9(2)(e)/(g)Political-opinion data processed on public-interest basis
NIS2 Art. 21Implementation-feasibility considers resilience

📄 Document Control

Owner: CEO (Intelligence Program) · Reviewer: CISO + Chief Analyst · Review Cycle: Quarterly Next Review: 2026-07-21 · Related: ai-driven-analysis-guide.md, strategic-extensions-methodology.md, synthesis-methodology.md

Generated following Riksdagsmonitor Electoral & Domain Methodology v1.4 — Family D Lens-Specific Layer.