Methodology

Files

• Exempt Organizations Business Master File, regional files eo1.csv through eo4.csv. The IRS page says these regions partition the file, so state files are not added on top. IRS BMF extract. Code labels come from the EO BMF information sheet.
• Publication 78, the list of organizations eligible to receive tax-deductible charitable contributions. Pipe-delimited. Data dictionary.
• Automatic revocation of exemption list. Pipe-delimited. An organization stays on the list after reinstatement. Data dictionary.
• Form 990-N e-Postcard, the latest notice on record. Only the tax year, tax-period end, and organization mailing state are counted. Data dictionary.

The IRS bulk page says the files are updated monthly and does not state the day. This project checks weekly and refreshes when a file's Last-Modified header changes. Page dates on irs.gov can lag the file timestamp, so both are shown when the page was retrieved.

Parsed October 19, 2026 UTC.

Last ingest status: completed. Ingest completed. Addition and removal counts were left blank.

BMF page posting date: 10/13/2026. Page record count: 1,975,975. Page last reviewed: 15-Oct-2026. Publication 78 page date: September 10, 2026. Revocation page date: September 8, 2026. 990-N page date: September 7, 2026.

The parsed row count and the IRS page record count are both shown. One is not edited to match the other.

Diffs

A slim index keeps EIN plus the whitelisted BMF fields (state, subsection, classification, foundation, deductibility, status, NTEE, ruling). It is cached by the ingest workflow and is not committed. The manifest stores its SHA-256. The next ingest uses that cache when the hash matches, sorts both indexes, and walks them in EIN order. An EIN only in the new file is newly listed. An EIN in both files with a different whitelisted field is a field change. Publication 78 uses the same walk for additions, removals, and code changes. Removed Publication 78 rows keep EIN, state, and code from the index; the legal name is not stored there. If the cache is missing, those counts are left blank.

Revocations and reinstatements do not need a prior snapshot. They are grouped by the posting date and the reinstatement date already in the file. The first ingest is a baseline: new-listing and Publication 78 change counts are left blank on purpose.

Row lists cover the latest 12 months when a month has at most 2,500 rows. Larger months keep complete counts by state and subsection. Each HTML page shows 150 rows; further rows are on /page/N. State pages keep the latest 60 revocation rows and 60 reinstatement rows for that state.

2020 revocation-date caveat

The IRS says organizations with an automatic-revocation date between April 1 and July 14, 2020 should have a revocation date of July 15, 2020, because of the COVID-19 filing extension and a system limit. This site shows July 15, 2020 for those rows and keeps the file value beside it. The boundary dates are included. July 15, 2020 and March 31, 2020 are left as the file states them.

Privacy rules

• Published from the BMF: EIN, legal name, city, state, subsection, classification, foundation, deductibility, ruling, status, NTEE.
• Published from the revocation list: EIN, legal name, city, state, subsection, revocation date, posting date, reinstatement date.
• Published from Publication 78: EIN, legal name, city, state, deductibility code.
• Published from Form 990-N: counts by tax year, tax-period end year, and organization mailing state. Not a row for each filer.
• Never stored or rendered: in-care-of name, street, ZIP, sort name, officer name, officer address, or the 990-N website field (it sometimes holds an email address).

Not stated

• The IRS download pages do not state a license. Calling the files U.S. government work is an inference, not a statement on those pages.
• The day of the next monthly IRS posting is not stated.
• Form 990-series XML is not in this version. The bulk page's August 20, 2026 XML posting date was noted when the sources were checked and is not loaded here.
• Whether a particular organization is currently exempt, deductible, or reinstated beyond these fields is not stated. Confirm it in Tax Exempt Organization Search.
• There is no page per organization.
• Officer names, street addresses, ZIP codes, in-care-of names, and sort names are in the IRS files and are not stated here.
• When no prior slim index is in the cache, BMF additions, field changes, and Publication 78 additions and removals are not stated.
• A full row list is not stated for a posting month with more than 2,500 organizations. The count is complete.
• Revocations are not stated by NTEE category, because that code is not on the revocation file.
• No sponsor is named, because none has been accepted.
• A pre-filled EIN search URL inside Tax Exempt Organization Search is not stated. Each EIN links to the search application.
• Classification is stored as up to four digits. A non-zero digit is read against the IRS one-digit classification table. Where that reading is wrong, the raw code is still the field that was compared.

Regional EO BMF files eo1.csv through eo4.csv are a partition. Per-state files are not added on top.

Revocation dates from 1 April 2020 through 14 July 2020 are shown as 15 July 2020, matching the IRS bulk-download caveat. The file value is kept as revocationDateRaw.

ICO, STREET, ZIP, SORT_NAME, and Form 990-N officer name and address columns are never written.

Row lists are kept for posting or reinstatement months in the latest 12 months when the month has at most 2,500 rows. Larger months keep complete counts only.

NTEE is not a column on the revocation file, so category pages do not assign revocations to a major group.

Snapshot: Slim EIN indexes are not committed. The ingest workflow caches data/snapshot and this manifest stores each index SHA-256. Diffs run only when that cache matches. Source file hashes and Last-Modified are the record of what was parsed. 28.4 MB across the three indexes.