SWF classification rubric (haircut factors for sovereignFiscalBuffer)

Central rubric for classifying sovereign wealth funds under the resilience sovereignFiscalBuffer dimension. Supports scripts/shared/swf-classification-manifest.yaml.

Every fund in the manifest has three coefficients in its classification: block:

classification:
  access:       0..1
  liquidity:    0..1
  transparency: 0..1

These multiply together to form the haircut the scorer applies when computing effective SWF months of reserve coverage:

effectiveMonths = rawSwfMonths × access × liquidity × transparency
score           = 100 × (1 − exp(−effectiveMonths / 12))

This doc defines what each coefficient value means with named tiers + concrete precedents, so:

1. Every rating in the manifest is defensible by pointing to a tier + precedent.
2. Future manifest PRs that add or revise ratings have an explicit benchmark to evaluate against.
3. A reviewer can audit the manifest without re-deriving the rubric from first principles each time.

Scope boundary. This is a methodology doc, not a ground-truth table. The coefficient values live in the manifest YAML; the rubric here explains the semantic tiers those values live on. Revising a fund's rating is a manifest-YAML edit cited against a tier here, not a rubric edit.

Axis 1 — Access

"How directly can the state deploy fund assets into budget support during a fiscal shock?"

Operationalized as a combination of legal mechanism (is there a withdrawal rule?), political clarity (who authorizes deployment?), and historical precedent (has deployment actually happened?). Deployment SPEED (weeks vs months vs years) is the core signal.

Edge case — fiscal-rule caps

A fiscal rule like Norway's ~3%-of-expected-real-return withdrawal cap creates an ambiguous access signal:

• Positive direction: the rule makes access PREDICTABLE and mechanically available for budget support every year, without political negotiation.
• Negative direction: the rule CAPS how much can be tapped, so in a severe shock the fund cannot be liquidated beyond the rule. The mechanism protects the savings against panic but rate-limits the stabilization function.

Rubric treatment: fiscal-rule-capped funds sit at the 0.5-0.7 boundary. Norway's GPFG at 0.6 (current manifest value) is defensible as "between hybrid-constrained and rule-based stabilization."

Edge case — state holding companies

Temasek-style state-holding-company assets can be deployed for fiscal support only via DIVIDEND FLOW, not via primary-asset liquidation (which would disrupt portfolio companies). This mechanism is slow (dividends are typically annual) and bounded (can't exceed portfolio earnings in a shock year). Rubric treatment: 0.3-0.4 tier, NOT the 0.5 hybrid tier — the mechanical deployment path is materially slower than QIA's amiri-decree route.

Axis 2 — Liquidity

"What share of the fund's AUM is in listed public markets and thus liquidatable within days/weeks without fire-sale discount?"

Operationalized as (public equities + listed fixed income + cash) ÷ total AUM, per the fund's most recent published asset mix. When the disclosure is a range (ADIA publishes 55-70%, not an exact ratio), the rubric uses the upper-bound of the range — the fund's own public statement is that it COULD be up to that figure, and haircut factors are designed to reward disclosed LIQUIDITY CAPACITY, not the conservative worst case. ADIA's 70% upper bound lands in the 0.7 tier (65-85%); if future ADIA disclosures tighten the range so the upper bound drops below 65%, the rubric directs the rating to 0.5.

Edge case — listed real estate

GPFG's listed real estate counts toward its liquidity score; PIF's direct real estate holdings do NOT. The distinction matters for boundary calls (0.7 vs 0.9): listed = liquidatable daily; directly-owned = months to sell at disclosed valuations.

Axis 3 — Transparency

"How well-documented is the fund's governance + financials?"

Operationalized as the Linaburg-Maduell (LM) Transparency Index score, normalized against IFSWF membership status and the granularity of the fund's annual reporting.

The LM index is a 10-point scale (1 = lowest, 10 = highest). IFSWF membership is binary (member / observer / non-member). Annual-report granularity gates tier promotion independently of LM/IFSWF.

Edge case — LM score vs disclosure depth

The LM index measures 10 governance signals (publication of financials, independent audit, public objectives, etc.). A fund can score LM=10 under the index while still publishing only RANGED asset-mix rather than exact holdings (Mubadala, Temasek). The rubric distinguishes these cases: LM=10 + holdings-level disclosure → 0.9-1.0 tier; LM=10 + asset-class-only disclosure → 0.7-0.8 tier. Mubadala's current manifest 0.6 under-rates the LM=10 signal against the rubric.

Edge case — sealed filings

KIA files detailed financials to the Kuwaiti National Assembly but the filings are SEALED from public disclosure. Under the rubric this sits at the 0.5 tier (asset-class disclosed + IFSWF engagement) rather than the 0.3 tier (no AUM), because the AUM is audited and disclosed to the oversight body — just not publicly. Current manifest = 0.4 is at the 0.3/0.5 boundary.

Current manifest × rubric alignment (informational, not PR-changes)

Reviewing each of the 8 current manifest values against the rubric tiers. This PR does NOT edit the manifest. The column "Rubric tier" shows where the rating falls under this rubric; "Manifest value" is the current YAML value; "Aligned?" flags whether the rating fits the rubric or looks off.

Summary of rubric-flagged ratings — 8 coefficients across 5 funds (Mubadala ×2, PIF ×2, KIA ×1, QIA ×2, GIC ×1) out of 24 total (8 funds × 3 axes):

• Mubadala access 0.4 (arguably 0.5); transparency 0.6 (arguably 0.7)
• PIF access 0.4 (arguably 0.5); liquidity 0.4 (arguably 0.3)
• KIA transparency 0.4 (arguably 0.5)
• QIA access 0.4 (arguably 0.5); transparency 0.4 (arguably 0.5)
• GIC access 0.6 (arguably 0.7)

None of these changes are made in this PR. The flags are informational — a future manifest-edit PR (PR 4b per the plan) should evaluate each flag, cite the rubric tier, and either confirm the current rating with a stronger rationale or revise it to match the tier.

Directional impact of the flagged ratings (if revised upward)

• Mubadala 0.4 → 0.5 on access, 0.6 → 0.7 on transparency: the access × transparency product moves from 0.24 to 0.35 (+46%). Combined with unchanged liquidity 0.5: haircut multiplier 0.12 → 0.175. UAE gains material SWF-months.
• PIF access 0.4 → 0.5: modest lift. PIF liquidity 0.4 → 0.3: modest dampening. Net: small.
• KIA transparency 0.4 → 0.5: haircut multiplier 0.7×0.8×0.4 = 0.224 → 0.7×0.8×0.5 = 0.28 (+25%). KW already top-quartile.
• QIA access 0.4 → 0.5 + transparency 0.4 → 0.5: QIA haircut 0.096 → 0.15 (+56%). Material lift for QA.
• GIC access 0.6 → 0.7: haircut 0.432 → 0.504 (+17%). SG lift.

The directional impact analysis is INFORMATIONAL and should NOT be treated as a decision to revise. Per the plan's anti-pattern note, rubric flags shouldn't be motivated by a target ranking outcome. A future manifest PR should revise ratings because the rubric + cited precedents support the change, not because the resulting ranking looks better.

How to use this rubric

When adding a new fund to the manifest

1. Locate each axis value on the tier table.
2. Cite the tier PLUS at least one concrete precedent (annual report page, LM index page, IFSWF profile URL).
3. If the fund sits between two tiers, pick the lower tier and explain the boundary rating in the YAML rationale: block.
4. PR review checks: does the rationale's cited evidence actually land the fund at the claimed tier?

When revising an existing fund

1. Cite what EVIDENCE changed: new annual report, LM score revision, IFSWF membership change, mandate amendment.
2. Map the new evidence to a tier per this rubric.
3. Update BOTH the coefficient AND the rationale: text in the same PR.
4. For PRs that shift multiple coefficients: run the cohort- sanity audit (see docs/methodology/cohort-sanity-release-gate.md) and publish the contribution-decomposition table for the affected countries.

When the rubric itself needs revising

Out of scope for a manifest PR. A rubric revision requires:

1. A separate methodology-decision PR citing the construct gap the revision fixes (e.g., "the current rubric doesn't handle state holding companies well — add a dedicated tier").
2. Re-evaluation of every existing fund under the new rubric (the rubric and the manifest must stay in lockstep).
3. Cohort-sanity audit snapshot before/after.

References

• Manifest: scripts/shared/swf-classification-manifest.yaml
• Scorer: server/worldmonitor/resilience/v1/_dimension-scorers.ts line 1654 (scoreSovereignFiscalBuffer)
• Saturating transform: score = 100 × (1 − exp(−effectiveMonths / 12))
• Linaburg-Maduell Transparency Index methodology: https://www.swfinstitute.org/research/linaburg-maduell-transparency-index
• IFSWF member directory: https://www.ifswf.org/members
• Santiago Principles self-assessments: https://www.ifswf.org/santiago-principles