Regulator filings¶

Some regulators publish the portfolios that funds file with them as open data. These are not issuer files: each row is labelled with the regulator, the filing and the position date, and the holdings are the fund’s own, as of that filing. Two are built (Brazil, Spain); the rest of this page records what was checked in other countries and why it is not built.

Status as tested on 2026-11-03.

Country	Source	Status	Kind of data	Position date in use
Brazil	CVM CDA, with B3’s ISIN register	Working (`cvm_brazil`)	monthly regulatory filing, full portfolio	2026-09-30 (2026-07-31 where newer months are mostly confidential)
Spain	CNMV FONDCART / SOCCART	Working (`cnmv_spain`)	quarterly regulatory filing, full portfolio	2026-06-30
Japan	EDINET API v2	Gated: free API key; client only	securities and semi-annual reports	not built
India	AMFI / AMC monthly portfolios	Not built	monthly portfolio per AMC, per-AMC Excel formats
Australia	ASIC portfolio holdings disclosure	No central dataset	semi-annual, on each issuer’s site
Korea	KRX ETF PDF	Not built: needs a decision	daily ETF portfolio deposit file
Taiwan	SITCA	Not built	no portfolio dataset found

Both working sources are issuer adapters (openbasket.sources.issuers.cvm_brazil and cnmv_spain), so discover() picks them up. Each also has a pipeline (reg_brazil, reg_spain) that refreshes its local store and writes the fund list; the adapters read the same folders.

Every IssuerHoldings from these sources says in notes what it is, for example regulatory filing: CVM monthly portfolio composition (CDA), positions as of 2026-09-30.

Brazil: CVM portfolio composition (CDA)¶

Every Brazilian fund class files a monthly Composição e Diversificação das Aplicações with the Comissão de Valores Mobiliários, which publishes all of them.

Portfolios

https://dados.cvm.gov.br/dados/FI/DOC/CDA/DADOS/cda_fi_<YYYYMM>.zip (listing at the same folder; dataset page https://dados.cvm.gov.br/dataset/fi-doc-cda). One ZIP per month, about 25 MB, of semicolon-separated Latin-1 CSVs:

cda_fi_PL: net assets (VL_PATRIM_LIQ) of every fund that filed;
cda_fi_BLC_1 to BLC_8: positions by kind. 1 federal bonds and repos (with CD_ISIN), 2 fund quotas (with the held fund’s CNPJ), 3 swaps, 4 shares, debentures, options, BDRs (with B3 code and CD_ISIN), 5 bank deposits, 6 private credit, 7 foreign assets, 8 everything else;
cda_fie: structured and index funds (FIDC, FIP, FIAGRO and ETFs, called FIIM), with net assets on every row and B3 codes but no ISIN column;
cda_fi_CONFID / cda_fie_CONFID: positions a fund keeps confidential, one total per kind of asset, with the date they will be released.

Fund ISINs

CVM identifies funds by CNPJ only. B3 is Brazil’s national numbering agency and publishes its whole ISIN register: https://sistemaswebb3-listados.b3.com.br/isinProxy/IsinCall/GetTextDownload/ returns the id of the latest full file (geralPt), and .../GetFileDownload/<base64 of the id> the ZIP (8 MB): NUMERACA.TXT lists every ISIN with a four-character issuer code and EMISSOR.TXT gives each issuer’s CNPJ. Fund quotas have asset type CTF. This is the same download the public page https://sistemaswebb3-listados.b3.com.br/isinPage/ offers.

How an input ISIN reaches its holdings: fund ISIN (e.g. BRBOVACTF003) to issuer code BOVA to CNPJ 10406511000161 to that CNPJ’s CDA rows.

Weights. The CDA has no weight column, so a line’s weight is its market value (VL_MERC_POS_FINAL) over the fund’s net assets in the same filing, times 100. Lines on the liability side are given a negative weight: written options, borrowed securities sold, forward obligations, payables, swap legs payable and “other liabilities” (LIABILITY_APLIC in the module). With that, the median fund’s lines sum to exactly 100.00% and the 5th and 95th percentiles of 25,305 FIF classes are 100.00% too (July 2026); without the signs the 95th percentile was 115.7%.

ISINs come from, in order, and each line says which in isin_source:

reported: the filing’s own CD_ISIN (bonds, shares, debentures);
b3_fund_cnpj: for a held fund, B3’s quota ISIN for its CNPJ, used only when that CNPJ has a single quota ISIN (54,298 CNPJs do; FIDCs with senior and subordinated quotas do not);
cda_ticker: for index-fund lines that only give a B3 code (VALE3), the ISIN that funds report next to that code in block 4 of the same month, if one ISIN has at least 95% of those rows (4,567 codes).

Repos (Operações Compromissadas) carry the collateral bond’s ISIN in the filing; they are not holdings of that bond, so they keep no ISIN and are named “Repo (operação compromissada) backed by …” with the bond in the name.

Confidential positions. A fund may withhold the detail of positions for up to 90 days. In the files published on 2026-11-03, 2.2% of net assets is confidential for July 2026 but 27.8% for September (38% of funds have more than 10% confidential). The store keeps the newest three months; for each fund the newest month with at most 10% confidential is used, else the month with the least, and the notes say so. Confidential totals stay in the frame as lines named “Confidential positions: ” with no ISIN, so they count as unresolved weight.

Privacy. Blocks 6 and 8 name individuals (CPF numbers) as issuers of some private credit. Their CPF and name are dropped at conversion; only company issuers keep their CNPJ and name.

What the data shows (files published 2026-10-31 to 2026-11-03):

	Jul 2026	Aug 2026	Sep 2026
Funds with a filing	26,301	26,200	26,322
Net assets (BRL trillion)	14.30	14.50	14.58
Confidential share of net assets	2.2%	27.9%	27.8%
Weight with an ISIN, one level down	74.9%	59.8%	59.6%

27,950 fund ISINs (26,145 CNPJs) are in the fund list. 24,650 of the 25,625 FIF classes in the July net-asset file match a B3 quota ISIN.

Fund	As of	Lines	Weight sum	With ISIN	Read in
iShares Ibovespa (BOVA11, BRBOVACTF003)	2026-09-30	158	100.00%	97.93%	0.09 s
Brasilprev Top TPF FIF RF (BRBPVTCTF000)	2026-09-30	36	100.00%	94.19%	0.02 s
BB RF Curto Prazo Automático FIC FIF (BR09SYCTF003)	2026-09-30	13	100.00%	100.00%	0.02 s
iShares S&P 500 FIC (IVVB11, BRIVVBCTF001)	2026-09-30	4	100.00%	0.00%	0.12 s

BOVA11’s unresolved 2.07% is receivables and cash. Brasilprev’s 5.8% is repos. The BB fund of funds is one line, 99.9997%, the quota of BR0CC3CTF008, which is itself in the fund list and expands from its own filing. IVVB11 holds the US iShares Core S&P 500 ETF, reported as a foreign asset with an internal code and no ISIN, so it stays unresolved; a proxy to IVV in lookthrough.proxies is the way to expand it.

Converting one month to Parquet takes about 1.2 s with DuckDB. The network test (one month and the register from CVM and B3, plus the Spanish quarter) ran in 32 s.

Freshness. Monthly data; the dataset is regenerated weekly, and old months change as confidential positions are released (the store re-downloads a month when the listing’s timestamp changes).

Terms. CVM’s open data is under the Open Database License (ODbL 1.0), as the dataset page states: credit CVM, and share derived databases under the same licence. Credit line: “Comissão de Valores Mobiliários (CVM), Dados Abertos: Fundos de Investimento, Composição e Diversificação das Aplicações (CDA); fund ISINs from B3’s ISIN register”. B3’s ISIN register is a free public download; its terms were not found on the download page. dados.cvm.gov.br/robots.txt asks for a 10-second crawl delay, which the client keeps (http.set_min_interval); /api/ is disallowed and not used.

Gaps:

Foreign assets (block 7, and foreign lines of index funds) have only a foreign ticker or an internal code, no ISIN.
Bank deposits, private credit and receivables have no ISIN.
Funds whose CNPJ has several quota ISINs are in the fund list, but a fund holding one of them gets no ISIN for that line (fund_cnpj keeps the CNPJ).
ETFs’ daily portfolios are not here; the CDA is month-end.
Futures are reported as filed (the value is usually the margin adjustment, not the notional).

Spain: CNMV quarterly portfolios¶

The Comisión Nacional del Mercado de Valores publishes, per month, ZIPs of “ficheros de información individual de IIC”: the data Spanish funds (FI) and SICAVs file under Circular 4/2008.

Page

https://www.cnmv.es/portal/publicaciones/descarga-informacion-individual (current year; ?ejercicio=YYYY for earlier years) links each month’s ZIP as https://www.cnmv.es/webservices/verdocumento/ver?e=<token>, with the month’s name as the link title. The ZIP of a quarter-end month (March, June, September, December) has the quarterly files; June 2026 is 8.6 MB.

Files used (XML, each with its XSD and an explanatory PDF in the ZIP)

FONDREGISTRO / SOCREGISTRO: every fund and SICAV, its compartments, and the ISIN of each share class or series;
FONDCART / SOCCART: every security held at quarter end: ClaseIF (domestic, foreign, doubtful), DescripcionIF (equity, bonds over or under one year, government debt, IIC, deposits, reverse repos …), CodigoISIN, DescripcionValor, Divisa, ValorMercado in euros;
FONDPATRIMDISVAR / SOCPATRIMDISVAR: net assets per compartment (TotalPatrimonio) split into financial investments, accrued interest, cash (Liquidez) and the rest.

How an input ISIN reaches its holdings: share-class ISIN to (type, register number, compartment) in the registry, to that compartment’s FONDCART or SOCCART lines.

Weights are ValorMercado / TotalPatrimonio * 100 for the same compartment and quarter. Cash, accrued interest and “rest” from the net asset breakdown are added as lines with no ISIN, so the weights add up to the fund: across 2,067 compartments the 5th to 95th percentiles of the sum are 99.65% to 100.12%, median 100.00%. Reverse repos keep the collateral in the name, not as the line’s ISIN.

What the data shows (June 2026, published 2026-10-26): 2,067 compartments of funds and SICAVs, 3,510 share-class ISINs, 130,475 positions, of which 130,220 carry a valid ISIN.

Fund	As of	Lines	Weight sum	With ISIN	Read in
Accion IBEX 35 ETF (ES0105336038)	2026-06-30	39	100.00%	96.97%	0.01 s
CaixaBank Master Renta Variable USA (ES0171963004)	2026-06-30	72	99.86%	86.59%	0.01 s
CaixaBank Master RF Deuda Pública 3-10 (ES0111223006)	2026-06-30	94	100.01%	95.49%	0.01 s
CaixaBank Monetario Rendimiento (ES0138045002)	2026-06-30	445	100.00%	79.25%	0.01 s

The unresolved weight is cash (11.4% of the US equity fund) and reverse repos (15% of the money-market fund). Parsing the quarter’s XML took 3 s.

Freshness. Quarterly, published about four months after the quarter (June 2026 data on 2026-10-26). CNMV states the files are not re-edited after publication, so a quarter is downloaded once.

Terms. CNMV’s legal notice (/portal/Utilidades/NotaLegal) authorises reuse and calculations on the information it publishes, provided it is reproduced faithfully and unaltered, and that anyone selling a product built on it tells buyers it is available free from the CNMV website. Credit line: “Comisión Nacional del Mercado de Valores (CNMV), ficheros de información individual de IIC (FONDCART/SOCCART quarterly portfolios)”. robots.txt only disallows *.shtml.

Gaps:

Derivatives (FONDDERI / SOCDERI) are reported as commitment amounts, not market values, and are left out.
Market values are in euros; the one fund whose currency is not the euro (USD) gets a note that its weights may be off by the exchange rate.
Positions are four to seven months old by the time they are published.
Spanish UCITS ETFs are few (two in the registry); most ETFs sold in Spain are Irish or Luxembourg funds and are not here.

Japan: EDINET (gated)¶

Investment trusts file securities reports (document type 120) and semi-annual reports (160) on EDINET; the financial statements end with a schedule of every security held (有価証券明細表). EDINET API v2 (https://api.edinet-fsa.go.jp/api/v2/documents.json?date=YYYY-MM-DD&type=2, /documents/<docID>?type=5) needs a Subscription-Key; without one it answers 401 “invalid subscription key”. The key is free but needs registering an account (e-mail and phone) on the EDINET API site, which this project does not do on anyone’s behalf.

openbasket.sources.edinet_japan is a client only: with EDINET_API_KEY set it lists the fund reports filed on a day and downloads a document. It does not parse the schedule (that needs real documents to develop against) and no open mapping from a fund ISIN (JP90C...) to EDINET’s fundCode was found, so no issuer adapter is registered. The key is sent as a query parameter and kept out of error messages.

Not built¶

India: SEBI requires every AMC to publish each scheme’s full portfolio monthly (fortnightly for debt schemes) on its own site and AMFI’s. The scheme list with ISINs is open (https://portal.amfiindia.com/spages/NAVAll.txt, which answered), but www.amfiindia.com, where the portfolio disclosure page is, timed out from here and refused connections from a second network, so the central page could not be checked. The portfolios themselves are Excel files in each AMC’s own layout (over 40 AMCs), which needs one parser per AMC.
Australia: ASIC’s portfolio holdings disclosure rules make each fund publish its holdings twice a year on its own website. There is no central dataset at ASIC, so this is an issuer route.
Korea: KRX publishes each ETF’s daily portfolio deposit file (PDF) on its Data Marketplace (data.krx.co.kr). The data is served by an internal JSON endpoint for the site’s own pages, not a documented download, and it was not probed; whether it now needs a log-in was not established. The public data portal (data.go.kr) needs a free service key and was only found to carry ETF prices, not portfolios.
Taiwan: Only SITCA’s fund registers (ISINs, links to monthly reports) were found as open data, no security-level portfolio dataset. Not checked further.
Other EU regulators: CNMV was the only EU national regulator found in this pass that publishes security-level fund portfolios as open data. The others were not checked one by one.