Open data · CC0 · tax year 2023

Download the grant graph

Public domain. Use it for anything, including commercial products and machine learning. Attribution is appreciated but not required.

FileContentsSizeSHA-256
whofunds_ty2023_foundations.csv.gzEvery foundation return read, with matched / unmatched / out-of-scope totals5.0 MB7ae6931b11f994c2…
whofunds_ty2023_grants.csv.gzEvery published grant line, with provenance78.9 MB99e51d78d121d8da…
whofunds_ty2023_organizations.csv.gzRecipient organizations matched to an EIN14.7 MB266f6bb28bd2eaa6…

Grant lines file

1,667,440 rows: every grant paid during the year that WHOFUNDS can publish. Matched lines carry a recipient EIN. Unmatched and foreign lines carry the name as filed, but only where no part of it could identify a person. Lines to individuals aren't in any file.

ColumnMeaning
filer_ein, filer_nameThe foundation that filed the 990-PF
object_id, source_filing_url, source_xml_urlThe exact e-filed return each line came from
section, line_nopaid (Part XV 3a), and the line's position in it
scoperesolved, unmatched or foreign
recipient_ein, recipient_nameThe matched organization (IRS BMF name)
recipient_name_as_filedThe payee as the foundation wrote it (when safe to show)
amount, purposeAs filed (purposes that identify a beneficiary are blank)
match_score, marginName similarity of the chosen EIN, and its lead over the runner-up
runner_up_ein, runner_up_scoreThe best rejected EIN, so a disputed edge can be checked
strategy, matched_variantWhich block found it, and which name of the organization matched

Cite

WHOFUNDS grant graph, tax year 2023 (built 2026-11-09). Public domain (CC0). https://whofunds.systemicaltruism.org/data/

Re-derive it

The pipeline is five standard-library Python scripts reading the GivingTuesday 990 Data Lake (anonymous S3) and the IRS BMF. See method, and the source at github.com/theguysaccount/whofunds-open-grant-graph. A scheduled job checks the sources for new filings and republishes the release and this site when they change.