Reference

Data

Measurements computed from primary regulatory filings, published in full and free to reuse. The fund industry does not publish these about itself, and the reason is not that they are hard to compute.

Updated 2026-10-01 ETF concentration Effective number of holdings for 1069 funds, from SEC Form N-PORT. What a holding count looks like once you weight it. 2019–2026 ETF concentration over time 24,300 quarterly observations across 1049 funds. A holding count barely moves; what the weights do to it does. Updated 2026-10-01 ETF valuation inputs How much of each fund carries a quoted market price rather than an estimate, with issuer country and asset class, from the same filings. 275 pairs ETF overlap What two funds actually share, weighted by position size, not just which tickers appear in both. 572 funds with derivatives Exposure, counterparties, swap financing Swap and futures notional as a share of net assets, the banks on the other side, and the financing leg as filed. The notional, not the position value, is what makes a 3x fund 3x. 24 registrants 13F holdings, as filed What Berkshire Hathaway, Renaissance, Bridgewater, Citadel and others reported on Form 13F, with weights, concentration and quarter-on-quarter changes. Named for the filing, never for a person. Every position, every quarter Point-in-time positions (CSV) The filings' own rows since 2019 — CUSIP, weight, value, collateral and lending flags — one file per fund. Recompute anything here, or measure something we have not. Every country, every quarter Where the money sits, by country Every comparable fund pooled by the issuer country it filed, quarter by quarter since 2019, with a page per country listing the funds that carry it. Per-fund and per-country JSON; the series that reads it is World through ETFs. Reported, not computed Monthly returns from the filings Each fund's own monthly total returns, read from Item B.5 of its N-PORT filings back to 2019, cross-checked against a price-derived series. Where the two disagree the fund is named and no summary figure is shown. One file, 3 MB Download every dataset All eleven derived datasets in one archive, with a README stating the method and limits and a SHA-256 per file. CC BY 4.0 — licence and commercial use. 235,641 pairs ETF overlap, every pair How much of any two funds is the same stock, for all 687 US funds over $1bn — not a chosen subset. 73 of them have a near-identical twin. Whole matrix as CSV. 116 borrowers named Who borrows the stock funds lend The other half of securities lending, from Item B.4.a: for 742 funds, the firms holding their shares, each named with its legal entity identifier and the amount on loan. $176bn at the latest filings. Every filing, what changed Filings received Each fund's newest N-PORT filing against its previous one: positions added and dropped, the largest weight changes, turnover. A feed per fund (/funds/TICKER/changes.xml) and one for all of them. Every field, every change Data dictionary and changelog Definitions for each file, the conventions they share, and a dated log of definition changes. Read this before piping the data anywhere. Daily, published Verification log Ten published funds re-derived from EDGAR every morning; two independent checks before every update. Results here, mismatches included. 199 companies Who owns what The reverse direction: for a given company, how many tracked funds hold it and how many combined dollars sit in it.

Why these exist

Every figure here is one line of arithmetic away from data US funds are legally required to file. Nobody publishes it because the firms with the data sell services to the firms being measured. That conflict is the whole opportunity, and this site does not have it: no affiliate links, no sponsored coverage, no issuer reviews anything before publication.

Each run is archived with its date. Holdings as of today can be fetched by anyone; holdings as of three years ago can only be read by whoever wrote them down.