A weekly letter on the movements of state capital. Free, by email.
No naked numbers. Every figure carries a source, an as-of date, and an evidence label — and only verified or clearly attributed figures publish.
State-owned investment institutions worldwide, in three types:
SWF — sovereign wealth funds · PPF — public pension and reserve funds · SDF — strategic/development holding funds.
Central-bank reserves count only where a distinct investment institution manages them. Sub-national public funds qualify. Member-owned industry schemes do not (universe definition v1.0).
1. Locate the institution’s most recent primary document (annual report, quarterly release, official statistics). 2. Capture the figure in its original currency with its exact as-of date. 3. Convert to US$ at the ECB reference rate on that as-of date — never today’s rate. 4. Label the evidence and record the source URL. 5. Publish — and correct within one business day when better documentation arrives.
What the labels on every figure mean:
sourced The institution’s own published figure, or an official government source — linked directly.
est. The institution does not disclose; the figure is a third-party estimate, attributed by name and vintage, never presented as fund-reported.
† Consolidated holding-company or total-fund basis — broader than portfolio assets and not directly comparable; each entry’s notes explain the basis.
How verifiable is each institution from public evidence alone? Eight criteria, each scored 0, 1 or 2 — a maximum of 16:
| Criterion | Question | |
|---|---|---|
| C1 | AUM disclosure | Does the institution publish its own assets under management? |
| C2 | Audited financials | Are audited financial statements public? |
| C3 | Reporting cadence | How frequently does it report? |
| C4 | Holdings detail | How much of the portfolio is disclosed? |
| C5 | Performance | Are investment returns published as a series? |
| C6 | Governance | Are the governing law, board, mandate and ownership public? |
| C7 | Strategy & allocation | Is the asset allocation disclosed? |
| C8 | Regulatory filings | Does it file recurring public reports (e.g. SEC 13F)? |
A 14–16 B 11–13 C 8–10 D 5–7 F 0–4
Grades measure disclosure, not performance or governance quality. A grade rises the moment an institution publishes more — and we note improvements in the weekly brief.
There is no weighting, no discretion, and no adjustment. The grade is the sum of the eight criteria, banded. That is the whole calculation:
That is a standing constraint on this publication, not a stylistic preference. Every criterion must be answerable from a public document by any reader, and the arithmetic that turns eight answers into a letter is fixed and published below. We hold no editorial view of any institution and apply none. A grade is a count of what an institution has published — a factual claim you can check — and never our opinion of how it is run.
Any future proposal to add a judgment-based axis would require this constraint to be broken deliberately and in public. That is why criteria such as “governance quality” are excluded: they are defensible things to assess, but they are assessments. This grade is a measurement.
total | = C1 + C2 + C3 + C4 + C5 + C6 + C7 + C8 (each 0, 1 or 2 → 0–16) |
grade | = A if total ≥ 14, else B if ≥ 11, else C if ≥ 8, else D if ≥ 5, else F |
If you score an institution differently from the same documents, that is a correction we want to receive.
A frozen, mechanical rubric guarantees that the arithmetic is reproducible. It does not by itself guarantee that each of the eight inputs was read off a document rather than inferred. Those are different claims, and we separate them here.
We audited our own scores on 1 September 2026 and publish the result:
| Criterion | Status |
|---|---|
| C1 · AUM disclosure | cross-checked against the evidence label on every AUM figure — 0 contradictions |
| C5 · Performance | cross-checked against return figures in the source record — 0 contradictions |
| C8 · Regulatory filings | partly verified — 23 institutions checked directly against SEC EDGAR, 6 scores corrected; the remainder still being verified |
| C2 · Audited financials | under audit |
| C3 · Reporting cadence | under audit |
| C4 · Holdings detail | under audit |
| C6 · Governance | under audit |
| C7 · Strategy & allocation | under audit |
In plain terms: five of the eight criteria — 500 of 800 individual scores in this edition — were scored from the research that produced each institution’s entry plus general knowledge of that institution, rather than from a document checked specifically against that criterion. That is inference. It is not yet measurement, and we are not going to describe it as one while we audit it.
Grades may therefore move. They have already moved once: eleven scores changed on 1 September 2026 when C8 was checked against the register rather than assumed, and five institutions moved a full letter. Every change is recorded in the changelog. Until the audit completes, treat a grade as our best current reading rather than a settled figure — and if you are citing one, cite it with its date.
Institutions in this universe report annually, and often late — China’s CIC runs roughly a twelve-month lag, its social security fund closer to sixteen. That is the reporting cycle of state capital, and no amount of diligence changes it. The assets figure on any entry will always be the slowest-moving number we publish.
One exception exists. Managers holding more than US$100 million of US-listed securities must file Form 13F with the SEC each quarter, roughly 45 days after quarter end. That is a statutory obligation rather than a courtesy, which makes it the only reliably quarterly signal in this universe. Where it exists, we show it as a filing activity marker on the institution’s page.
It is not evenly available, so here is the split as we have verified it: of 100 institutions, 28 are confirmed filers and carry the quarterly marker; 14 are confirmed non-filers; and 58 have not yet been checked. Non-filers — among them GIC, ADIA, GPIF, HKMA and Australia’s Future Fund — have no quarterly signal at all, and their entries are monitored only against expected reporting cadence. That is a genuine limitation of this publication, and we would rather print it than let a reader assume uniform freshness.
Form 13F is never used as an assets figure and never enters a grade. It covers US-listed equities and certain related instruments only — excluding fixed income, real estate, direct stakes, private holdings and all non-US listings. For most institutions that is a minority of total assets, and sometimes a tiny one: PIF’s entire 13F is five positions, Alaska’s is ten. It is excluded from the Disclosure Grade by design, because filing is a legal obligation rather than voluntary disclosure, and grading it would penalise institutions with little US equity exposure for something unrelated to transparency.
Rubric v1.0 is frozen as of edition 2026. The criteria, their definitions, and the banding above will not change within an edition — a grade moves only because an institution’s disclosure moved, never because we changed the ruler mid-measurement. Any future change to the rubric ships as a new numbered version, is recorded in the changelog, and states whether prior editions were rescored.
Axes under consideration for a future version: fee and cost transparency (does the institution publish what it costs to run?) and reported-versus-audited reconciliation (does the headline figure tie to audited statements?). Both are disclosure questions answerable from public documents. Fund-level performance measures such as DPI and TVPI are explicitly out of scope — they describe closed-end funds, not asset owners.
No pay-to-rank: grades and ranks cannot be bought, and commercial relationships, if they ever exist, will be disclosed. Institutions may submit corrections in writing with supporting documents to corrections@capitalalmanack.com; verified corrections are applied with a public changelog entry, and disputed estimates are annotated with the institution’s stated position.
Represent one of these institutions — or know the numbers better than we do? Send corrections or newer figures with supporting documents. Verified updates are applied within one business day and credited in a public changelog — institutions that engage are how this record stays accurate.
Email a correction