BMF Data Catalog

The IRS Business Master File (BMF) is the authoritative registry of tax-exempt organizations in the United States. NCCS publishes BMF data in several forms; this catalog organizes them by what you’re trying to do.

For background on what the BMF contains and its scope limitations, see the BMF overview page.

Which dataset should I use?

If you need… Use Why
One row per nonprofit, ever (1989–present) Unified BMF Combines every vintage into a single longitudinal frame
A specific month’s snapshot (2023-06 or later) Transformed monthly BMF Cleaned and standardized; column schema is stable across months
A specific month’s snapshot (1989–2022) Harmonized legacy BMF Same column schema as transformed; aligns historical vintages
The unmodified IRS file Raw monthly archive Untouched IRS releases, for replication of upstream analysis
The original NCCS legacy file Raw legacy archive Original 501CX-NONPROFIT-PX vintages, for reproducibility only
Canonical county FIPS, or to roll up to metro areas Geography crosswalks Optional join layers for the geocoded BMF (TIGER 2023 / OMB 2023)
An NTEE code for an org whose current code is blank NTEE-resolved crosswalk Recovers a per-EIN NTEE across all vintages (current + legacy); join by ein


Unified BMF

One row per EIN spanning 1989 to present, with most-recent values for each organization plus first_vintage_ym, last_vintage_ym, and bmf_source to track provenance. This is the primary product for longitudinal analysis and as a sampling frame. 3,687,435 unique EINs, built from 118 source files across 114 vintages (1989–2026).

The join key is ein (XX-XXXXXXX), unchanged. The Unified BMF additionally carries two coercion-safe renderings of the same key (ADR 0036): ein_prefixed (ein-XX-XXXXXXX, lowercase — provably text, recommended for new joins) and EIN2 (EIN-XX-XXXXXXX, uppercase — a legacy-compatibility alias for merges written against the older Unified BMF / harmonized CORE format). Use whichever fits your existing code; all renderings are bijective with the same underlying 9-digit EIN.

Renamed from “Master BMF” (2026-07-01). This product was briefly published as “Master BMF”; it has been renamed back to Unified BMF, restoring the name the research community originally knew it by (ADR 0037). This is a non-silent supersession: the prior s3://nccsdata/master/bmf/ path stays live and reachable through 2026-09-28 (a 90-day window), after which it moves to a retained, citable archive — it is never deleted. If you bookmarked or scripted against the old path, re-pin to s3://nccsdata/unified/bmf/ before the cutover.

The geocoded variant is the recommended default — it carries the same row schema as the plain file plus lat/lon joined against a deduplicated address table. (The geocoded extension and its per-state marts are a separate product, still named “Master BMF” — they are not renamed by this change.)

Variant Download Dictionary Quality Report Size
Master BMF (geocoded) Download Dictionary Quality report 3253.4 mb
Unified BMF Download Dictionary Quality report 2973.4 mb

Download by State/Territory

For users who only need a single state, the geocoded Master BMF is also published as per-state marts (50 states + DC, U.S. territories, APO/FPO military codes, Compact of Free Association nations, and a ZZ bucket for rows with no resolvable state). Each mart carries lat/lon and shares a schema and dictionary with the geocoded headline file above.

Download Link Size State/Territory
Download 44 mb Alabama
Download 9.6 mb Alaska
Download 51.7 mb Arizona
Download 28.2 mb Arkansas
Download 343.6 mb California
Download 61 mb Colorado
Download 39.2 mb Connecticut
Download 15.6 mb Delaware
Download 29.6 mb District of Columbia
Download 186.9 mb Florida
Download 100.9 mb Georgia
Download 15.5 mb Hawaii
Download 17.4 mb Idaho
Download 132.8 mb Illinois
Download 73.6 mb Indiana
Download 48.9 mb Iowa
Download 31.8 mb Kansas
Download 38 mb Kentucky
Download 41.9 mb Louisiana
Download 16.8 mb Maine
Download 72.5 mb Maryland
Download 73.8 mb Massachusetts
Download 97.6 mb Michigan
Download 66.2 mb Minnesota
Download 26.7 mb Mississippi
Download 74.4 mb Missouri
Download 18.8 mb Montana
Download 24.1 mb Nebraska
Download 22.7 mb Nevada
Download 15.5 mb New Hampshire
Download 88.6 mb New Jersey
Download 19.7 mb New Mexico
Download 198.5 mb New York
Download 96.6 mb North Carolina
Download 11 mb North Dakota
Download 129.4 mb Ohio
Download 38.4 mb Oklahoma
Download 44.5 mb Oregon
Download 134.8 mb Pennsylvania
Download 11.6 mb Rhode Island
Download 50.4 mb South Carolina
Download 12.5 mb South Dakota
Download 64.3 mb Tennessee
Download 244.4 mb Texas
Download 21.4 mb Utah
Download 10.9 mb Vermont
Download 89.1 mb Virginia
Download 72.4 mb Washington
Download 20.7 mb West Virginia
Download 65.5 mb Wisconsin
Download 9.6 mb Wyoming
Download 0.1 mb American Samoa
Download 0.3 mb Guam
Download 0.1 mb Northern Mariana Islands
Download 4.6 mb Puerto Rico
Download 0.9 mb U.S. Virgin Islands
Download 0 mb Armed Forces Americas
Download 0.3 mb Armed Forces Europe
Download 0.2 mb Armed Forces Pacific
Download 0 mb Federated States of Micronesia
Download 0 mb Marshall Islands
Download 0 mb Palau
Download 4.5 mb Unknown / unmapped

Access via S3

s3://nccsdata/geocoding/bmf-master/merged/   # Geocoded variant (recommended)
s3://nccsdata/unified/bmf/                   # Plain Unified BMF (bmf_unified.{parquet,csv} + dictionary + quality report + _manifest.json)
s3://nccsdata/master/bmf/                    # Superseded plain path — reachable through 2026-09-28, then archived (not deleted)
s3://nccsdata/master/bmf/state_marts/csv/    # Per-state geocoded marts


Monthly BMF

Cleaned, schema-stable monthly snapshots covering 1989 to the present. The transformed pipeline (2023-06+) and the harmonized legacy pipeline (1989-2022) write to different S3 prefixes but share a column schema, so you can stack them across the era boundary without reshaping.

Recent Releases

One row per release. The Download column links the CSV (canonical format); a parquet variant is also available for some months at the same S3 prefix.

Year Month Pipeline Download Dictionary Quality Report Size
2026 06 Transformed Download Dictionary Quality report 1785.3 mb
2026 05 Transformed Download Dictionary Quality report 1766.2 mb
2026 03 Transformed Download Dictionary Quality report 1712.8 mb
2026 01 Transformed Download Dictionary Quality report 1704.2 mb
2025 12 Transformed Download Dictionary Quality report 1702.4 mb

Browse Older Releases

The complete archive of every monthly release — data, dictionary, and quality report — is below. (There is no separate centralized index; this catalog is the canonical listing.)

Show all monthly releases (1989–present)
Year Month Pipeline Download Dictionary Quality Report Size
2026 06 Transformed Download Dictionary Quality report 1785.3 mb
2026 05 Transformed Download Dictionary Quality report 1766.2 mb
2026 03 Transformed Download Dictionary Quality report 1712.8 mb
2026 01 Transformed Download Dictionary Quality report 1704.2 mb
2025 12 Transformed Download Dictionary Quality report 1702.4 mb
2025 11 Transformed Download Dictionary Quality report 1681.3 mb
2025 10 Transformed Download Dictionary Quality report 1681.3 mb
2025 09 Transformed Download Dictionary Quality report 1667.8 mb
2025 08 Transformed Download Dictionary Quality report 1693 mb
2025 07 Transformed Download Dictionary Quality report 1685 mb
2025 06 Transformed Download Dictionary Quality report 1681.2 mb
2025 05 Transformed Download Dictionary Quality report 1679.2 mb
2025 04 Transformed Download Dictionary Quality report 1672 mb
2025 03 Transformed Download Dictionary Quality report 1672.1 mb
2025 02 Transformed Download Dictionary Quality report 1664.6 mb
2025 01 Transformed Download Dictionary Quality report 1664.6 mb
2024 12 Transformed Download Dictionary Quality report 1662.3 mb
2024 11 Transformed Download Dictionary Quality report 1657.1 mb
2024 08 Transformed Download Dictionary Quality report 1677.4 mb
2024 07 Transformed Download Dictionary Quality report 1669.2 mb
2024 06 Transformed Download Dictionary Quality report 1662.8 mb
2024 04 Transformed Download Dictionary Quality report 1649.8 mb
2024 03 Transformed Download Dictionary Quality report 1651.7 mb
2024 02 Transformed Download Dictionary Quality report 1645.8 mb
2024 01 Transformed Download Dictionary Quality report 1645.8 mb
2023 12 Transformed Download Dictionary Quality report 1644.1 mb
2023 11 Transformed Download Dictionary Quality report 1639.8 mb
2023 10 Transformed Download Dictionary Quality report 1633.1 mb
2023 09 Transformed Download Dictionary Quality report 1628.8 mb
2023 08 Transformed Download Dictionary Quality report 1252.9 mb
2023 07 Transformed Download Dictionary Quality report 1654.2 mb
2023 06 Transformed Download Dictionary Quality report 1621.9 mb
2022 08 Harmonized legacy Download Dictionary Quality report 774.7 mb
2022 01 Harmonized legacy Download Dictionary Quality report 773.5 mb
2020 04 Harmonized legacy Download Dictionary Quality report 765.3 mb
2019 08 Harmonized legacy Download Dictionary Quality report 763.1 mb
2018 12 Harmonized legacy Download Dictionary Quality report 677.9 mb
2016 08 Harmonized legacy Download Dictionary Quality report 821.3 mb
2016 04 Harmonized legacy Download Dictionary Quality report 816 mb
2016 03 Harmonized legacy Download Dictionary Quality report 814.5 mb
2016 02 Harmonized legacy Download Dictionary Quality report 811.2 mb
2015 12 Harmonized legacy Download Dictionary Quality report 804.4 mb
2015 11 Harmonized legacy Download Dictionary Quality report 804.8 mb
2015 09 Harmonized legacy Download Dictionary Quality report 796.1 mb
2015 07 Harmonized legacy Download Dictionary Quality report 801.7 mb
2015 05 Harmonized legacy Download Dictionary Quality report 795.3 mb
2015 04 Harmonized legacy Download Dictionary Quality report 790.9 mb
2015 02 Harmonized legacy Download Dictionary Quality report 783.6 mb
2014 12 Harmonized legacy Download Dictionary Quality report 777.3 mb
2014 11 Harmonized legacy Download Dictionary Quality report 773 mb
2014 09 Harmonized legacy Download Dictionary Quality report 757.8 mb
2014 06 Harmonized legacy Download Dictionary Quality report 743.8 mb
2014 04 Harmonized legacy Download Dictionary Quality report 741.1 mb
2014 02 Harmonized legacy Download Dictionary Quality report 666.8 mb
2013 12 Harmonized legacy Download Dictionary Quality report 664.2 mb
2013 10 Harmonized legacy Download Dictionary Quality report 663 mb
2013 09 Harmonized legacy Download Dictionary Quality report 662.5 mb
2013 08 Harmonized legacy Download Dictionary Quality report 661.9 mb
2013 07 Harmonized legacy Download Dictionary Quality report 671.9 mb
2013 06 Harmonized legacy Download Dictionary Quality report 670.6 mb
2013 05 Harmonized legacy Download Dictionary Quality report 670.7 mb
2013 04 Harmonized legacy Download Dictionary Quality report 670.2 mb
2013 03 Harmonized legacy Download Dictionary Quality report 668.2 mb
2013 02 Harmonized legacy Download Dictionary Quality report 677.2 mb
2012 12 Harmonized legacy Download Dictionary Quality report 685.4 mb
2012 11 Harmonized legacy Download Dictionary Quality report 684.3 mb
2012 10 Harmonized legacy Download Dictionary Quality report 683.9 mb
2012 08 Harmonized legacy Download Dictionary Quality report 683.3 mb
2012 07 Harmonized legacy Download Dictionary Quality report 682.7 mb
2012 06 Harmonized legacy Download Dictionary Quality report 684.4 mb
2012 04 Harmonized legacy Download Dictionary Quality report 678.8 mb
2012 03 Harmonized legacy Download Dictionary Quality report 680.4 mb
2012 02 Harmonized legacy Download Dictionary Quality report 674.5 mb
2011 12 Harmonized legacy Download Dictionary Quality report 691.6 mb
2011 11 Harmonized legacy Download Dictionary Quality report 690.9 mb
2011 10 Harmonized legacy Download Dictionary Quality report 689.3 mb
2011 09 Harmonized legacy Download Dictionary Quality report 688.6 mb
2011 08 Harmonized legacy Download Dictionary Quality report 688.4 mb
2011 07 Harmonized legacy Download Dictionary Quality report 689.1 mb
2011 06 Harmonized legacy Download Dictionary Quality report 759.2 mb
2010 11 Harmonized legacy Download Dictionary Quality report 765.4 mb
2010 08 Harmonized legacy Download Dictionary Quality report 758.2 mb
2010 07 Harmonized legacy Download Dictionary Quality report 756.3 mb
2010 05 Harmonized legacy Download Dictionary Quality report 755.6 mb
2010 04 Harmonized legacy Download Dictionary Quality report 753.4 mb
2010 01 Harmonized legacy Download Dictionary Quality report 713.5 mb
2009 10 Harmonized legacy Download Dictionary Quality report 708.2 mb
2009 07 Harmonized legacy Download Dictionary Quality report 701.5 mb
2009 04 Harmonized legacy Download Dictionary Quality report 694.3 mb
2009 01 Harmonized legacy Download Dictionary Quality report 687.6 mb
2008 12 Harmonized legacy Download Dictionary Quality report 684.2 mb
2008 10 Harmonized legacy Download Dictionary Quality report 682.5 mb
2008 06 Harmonized legacy Download Dictionary Quality report 674 mb
2008 04 Harmonized legacy Download Dictionary Quality report 660.3 mb
2008 01 Harmonized legacy Download Dictionary Quality report 658.7 mb
2007 09 Harmonized legacy Download Dictionary Quality report 666.2 mb
2007 04 Harmonized legacy Download Dictionary Quality report 664.5 mb
2007 01 Harmonized legacy Download Dictionary Quality report 662.4 mb
2006 11 Harmonized legacy Download Dictionary Quality report 647.2 mb
2006 05 Harmonized legacy Download Dictionary Quality report 601.6 mb
2006 01 Harmonized legacy Download Dictionary Quality report 604.7 mb
2005 11 Harmonized legacy Download Dictionary Quality report 602.5 mb
2005 07 Harmonized legacy Download Dictionary Quality report 599.4 mb
2004 12 Harmonized legacy Download Dictionary Quality report 597.3 mb
2004 04 Harmonized legacy Download Dictionary Quality report 587.8 mb
2003 11 Harmonized legacy Download Dictionary Quality report 584.5 mb
2003 07 Harmonized legacy Download Dictionary Quality report 575.4 mb
2003 01 Harmonized legacy Download Dictionary Quality report 564.8 mb
2002 07 Harmonized legacy Download Dictionary Quality report 554.9 mb
2002 01 Harmonized legacy Download Dictionary Quality report 587.3 mb
2001 07 Harmonized legacy Download Dictionary Quality report 527.1 mb
2000 05 Harmonized legacy Download Dictionary Quality report 506.9 mb
1999 12 Harmonized legacy Download Dictionary Quality report 505.3 mb
1998 09 Harmonized legacy Download Dictionary Quality report 497.1 mb
1997 10 Harmonized legacy Download Dictionary Quality report 483.9 mb
1996 06 Harmonized legacy Download Dictionary Quality report 462.9 mb
1995 08 Harmonized legacy Download Dictionary Quality report 441.1 mb
1989 06 Harmonized legacy Download Dictionary Quality report 249.5 mb

Access via S3

s3://nccsdata/processed/bmf/{YYYY_MM}/         # Transformed (2023-06 -> present)
s3://nccsdata/processed/bmf-legacy/{YYYY_MM}/  # Harmonized legacy (1989-06 -> 2022)

Each month directory contains:

  • bmf_{YYYY_MM}_processed.csv (or .parquet) — the data
  • bmf_{YYYY_MM}_data_dictionary.csv — column definitions
  • bmf_{YYYY_MM}_quality_report.json — machine-readable quality metrics


Geography Crosswalks

Optional join layers for the geocoded Master BMF that turn a messy address into clean geography: attach a canonical county FIPS code to each organization, then roll that county up to its metro/micropolitan area. Reach for these when you need a collision-proof geographic key — Baltimore city (24510) is not Baltimore County (24005) — or want to aggregate nonprofits to metro areas.

These rows link the canonical S3 copies — the same files the pipeline consumes — so the catalog never drifts from the machine contract. The CSV is for humans and spreadsheets; the parquet is the machine-typed copy; each _manifest.json records the row count, sha256, and geography vintage.

How the crosswalks chain together

These are derived, multi-hop artifacts: you don’t start from county_fips, you arrive at it. The geocoded BMF gives you a dirty county label and a lat/lon; the crosswalks walk that to a metro area:

  1. County label → county FIPS. Join the county FIPS crosswalk on (geo_state_abbr, geo_county) — the crosswalk side of the county key is geo_county_raw — to get a canonical 5-char geo_county_fips.
  2. Connecticut detour (by coordinate). CT replaced its counties with planning regions in 2022, so CT’s old county labels are deferred in step 1 (resolution = deferred_ct_planning_region). Resolve them instead from the org’s coordinate: round lat/lon to 2 decimals and join the CT planning-region crosswalk on (lat2, lon2) to get the planning-region FIPS (09110–09190).
  3. County FIPS → metro area. Chain the CBSA crosswalk on county_fips to attach the metro/micropolitan area (and Combined Statistical Area, where applicable). The CBSA crosswalk already keys CT on the new planning-region FIPS, so step 2 lands cleanly here.

All three share one geography vintage (Census TIGER 2023 / OMB 2023) and are designed to be chained — present them together, not in isolation. FIPS/CBSA/CSA codes are strings; preserve leading zeros. Unresolved/ambiguous county labels and rural counties carry NA codes by design, not as missing data.

Crosswalk Role in the chain Geography vintage Rows Download Machine copy Manifest
County FIPS label → FIPS (hop 1) Census TIGER 2023 3,635 (3,615 resolved, 99.4%) CSV Parquet manifest.json
CT planning-region CT by coordinate (hop 2) Census TIGER 2023 14,271 (coordinate grid) CSV Parquet manifest.json
CBSA (metro/micro) FIPS → metro area (hop 3) OMB 2023 delineations 3,228 (1,915 in a CBSA, 1,313 rural) CSV Parquet manifest.json

Geography vintage matters. County boundaries shift — Connecticut replaced its counties with planning regions in 2022 — so a stale vintage is a silent trap. All three crosswalks are built against the 2023 vintage (Census TIGER 2023 boundaries; OMB 2023 CBSA delineations) and carry the year in-column (tiger_year / delineation_year). Filter against the boundary year you intend, and don’t mix vintages across the chain.

County FIPS crosswalk

Maps the geocoder’s dirty county label (geo_county) to the canonical Census county FIPS GEOID + NAMELSAD, so you can canonicalize names and filter by a collision-proof key. 3,635 rows; 3,615 resolved (99.4%) — the remainder are 8 ambiguous city-vs-county labels, 8 CT labels deferred to the planning-region crosswalk, and 4 unresolved.

Column Description
geo_state_abbr 2-letter state postal code (join key)
geo_county_raw The geocoder’s raw, dirty county label (join key)
geo_county_fips Canonical 5-char county FIPS GEOID (string — keep leading zeros)
state_fips 2-char state FIPS (string)
geo_county_canonical Canonical county name (NAMELSAD)
resolution How the raw label was matched (resolved, ambiguous, unresolved, deferred_ct_planning_region)
tiger_year Census TIGER boundary vintage (2023)

CT planning-region crosswalk

Connecticut’s companion to the county crosswalk. Because CT’s old county labels can’t be trusted post-2022, this resolves CT geography from the organization’s coordinate instead of its label. It’s a lookup grid — one row per ~1 km cell (lat/lon rounded to 2 decimals) mapped to one of CT’s 9 planning regions (FIPS 09110–09190). 14,271 rows.

Column Description
geo_state_abbr Always CT
lat2, lon2 Latitude/longitude rounded to 2 decimals (the join key — round the org’s coordinate the same way)
geo_county_fips CT planning-region FIPS (09110–09190; joins to the CBSA crosswalk’s county_fips)
state_fips 09
geo_county_canonical Planning-region name
area_share Fraction of the cell falling in this planning region (1 when the cell sits fully inside one region)
straddle TRUE where a cell straddles a region boundary (529 cells); pick the row with the largest area_share
tiger_year Census TIGER boundary vintage (2023)

CBSA crosswalk

The metro/micropolitan rollup. Maps each county FIPS to its OMB Core-Based Statistical Area, so you can aggregate nonprofits to metro areas. 3,228 rows (one per county GEOID): 1,915 counties fall in a CBSA — 1,252 metropolitan, 663 micropolitan, across 935 distinct CBSAs (184 of which are grouped into Combined Statistical Areas) — and 1,313 rural counties carry NA.

Column Description
county_fips 5-char county FIPS (string; joins to the county crosswalk’s geo_county_fips)
county_name County name
cbsa_code CBSA code (string — keep leading zeros)
cbsa_title CBSA name
cbsa_type Metropolitan Statistical Area or Micropolitan Statistical Area
central_outlying Whether the county is central or outlying within the CBSA
csa_code Combined Statistical Area code (string), if any
csa_title Combined Statistical Area name, if any
delineation_year OMB delineation vintage (2023)

Caveats

  • FIPS, CBSA, and CSA codes are strings. Opening a CSV directly in Excel may strip leading zeros on display — read these columns as text.
  • NA is a value, not a gap. Rural counties (those in no CBSA) carry NA CBSA/CSA codes by design; that’s the correct value. Likewise, ambiguous city-vs-county labels (a bare “Baltimore” or “Roanoke”) are intentionally left unresolved upstream in the county crosswalk.

Access via S3

s3://nccsdata/crosswalks/county-fips/         # county FIPS crosswalk (.csv + .parquet + _manifest.json)
s3://nccsdata/crosswalks/ct-planning-region/  # CT planning-region crosswalk (.csv + .parquet + _manifest.json)
s3://nccsdata/crosswalks/cbsa/                # CBSA crosswalk (.csv + .parquet + _manifest.json)

Each prefix is flat — the vintage is carried in-column and in _manifest.json (use its sha256 to detect changes), not in the path.

Full producer documentation: County FIPS crosswalk · CBSA crosswalk · CT planning-region crosswalk.



NTEE-Resolved Crosswalk

A per-EIN NTEE classification recovered across every vintage of both BMF pipelines (the current monthly file and the legacy 501CX archive). The Unified BMF carries each organization’s NTEE as reported in its most-recent vintage — which silently discards a known older code when the newest one is blank. The IRS sometimes nulls an organization’s NTEE_CD at source: Carnegie Mellon (EIN 25-0969449) is blank in the current BMF but was coded B43 (University) in the legacy vintages. No cleaner can classify a null; only a cross-vintage lookup can recover it.

This crosswalk does that recovery. It aggregates the verbatim raw NTEE code per EIN across all 114 vintages, cleans the distinct codes once with the current NTEE transform, and exposes the full history so you pick the resolution rule that fits your use — it takes no opinionated single pick (per ADR 0016, consumer-composes). It is not a column on the Unified BMF: join it onto the BMF or CORE by ein (XX-XXXXXXX) and apply the rule you want.

The four resolution views (each EIN carries all of them):

  • ntee_current — the code in the latest current vintage. May be NA by design — this is the IRS’s current answer (CMU’s state today).
  • ntee_most_recent — the newest vintage with a non-null code, from either pipeline (CMU → B43). This is the view that recovers IRS-nulled codes.
  • ntee_modal — the most-observed code across vintages (resistant to a one-off miscode; ties broken by recency).
  • ntee_code_distribution — the full per-code history as JSON, for your own custom rule.

Each of the first three carries _subsector (NTEEv2 subsector, e.g. UNI/HOS/EDU) and _nteev2 (full NTEEv2 code) derivatives, plus n_distinct_codes, n_vintages_with_ntee, and ntee_agreement (single/unanimous/mixed) as a confidence signal. To filter for universities, for example, use ntee_most_recent_subsector == 'UNI' — not a substring guess.

Scale & recovery. 3,613,958 EINs (one row per EIN with an NTEE in ≥1 vintage). Of 2,054,414 EINs whose ntee_current is null/blank, 2,031,485 (98.9%) recover a usable code via ntee_most_recent. Agreement across vintages: unanimous 71.5% / mixed 21.5% / single 7.0%.

File Format Size Download
ntee_resolved_crosswalk.parquet Parquet (zstd) — machine contract, recommended ~78 MB Parquet
ntee_resolved_crosswalk.csv CSV mirror ~640 MB CSV
_manifest.json Provenance (ADR 0014) — manifest.json

The 640 MB CSV is not Excel-friendly — it is far larger than a spreadsheet can open, and Excel strips leading zeros and mangles the JSON ntee_code_distribution column. Prefer the parquet (the machine contract); reach for the CSV only for a streaming/awk-style read where parquet tooling isn’t available.

Column dictionary

One row per EIN, 20 columns. NTEE codes are strings (NTEE-CC) — never cast them numeric.

Column Type Description
ein string Formatted EIN (XX-XXXXXXX) — the join key (never null); unchanged
ein_prefixed string ein-XX-XXXXXXX — coercion-safe lowercase-prefixed rendering of ein (ADR 0036)
EIN2 string EIN-XX-XXXXXXX — legacy-compatibility alias of ein (matches the older Unified BMF / harmonized CORE join format)
ntee_current string Cleaned NTEE-CC from the latest current vintage; NA if never in current or IRS-nulled
ntee_current_subsector string NTEEv2 subsector (e.g. UNI/HOS/EDU) for the current code
ntee_current_nteev2 string Full NTEEv2 code for the current code
current_vintage string YYYY_MM of the latest current vintage observed for the EIN
ntee_most_recent string Cleaned NTEE-CC from the newest vintage carrying a non-null code (any source)
ntee_most_recent_subsector string NTEEv2 subsector for the most-recent code
ntee_most_recent_nteev2 string Full NTEEv2 code for the most-recent code
ntee_most_recent_vintage string YYYY_MM that supplied ntee_most_recent
ntee_most_recent_source string current | legacy — pipeline that supplied ntee_most_recent
ntee_modal string Most-observed cleaned code across vintages (ties broken by recency)
ntee_modal_subsector string NTEEv2 subsector for the modal code
ntee_modal_nteev2 string Full NTEEv2 code for the modal code
ntee_modal_n int Vintage-observations supporting the modal code
ntee_code_distribution string JSON keyed on cleaned NTEE-CC: {code:{n,first,last}} — full history
n_distinct_codes int Count of distinct cleaned codes ever observed for the EIN
n_vintages_with_ntee int Total vintage-observations carrying an NTEE
ntee_agreement string single (one obs) | unanimous (all agree) | mixed

There is no separate data-dictionary CSV for this artifact — the contracted truth is the parquet/CSV plus _manifest.json, and the column docs above come from the crosswalk contract.

Provenance

The _manifest.json (ADR 0014 shape) records the published provenance — use its sha256 to detect changes. As of first publish:

Field Value
vintage 2026_06 (newest current BMF vintage at build)
built_at 2026-06-24T19:54:22Z
git_sha 1e1290d
row_count 3,613,958
Inputs current + legacy intermediate parquets (ntee_code_raw), plus the sha256 of transform_ntee_code.R and the legacy 5-char lookup

The crosswalk is rebuilt monthly after each new current BMF lands, so ntee_current tracks the newest vintage (the legacy half is static); uploads are idempotent (unchanged files skipped via the manifest sha256). Always read the live _manifest.json for the current vintage/git_sha/built_at.

Access via S3

s3://nccsdata/crosswalks/ntee-resolved/   # .parquet (machine contract) + .csv mirror + _manifest.json

The prefix is flat — the vintage is carried in _manifest.json, not in the path. Rationale: ADR 0034.



Raw Archives

Show raw IRS and NCCS files (for reproducibility only — most users want the products above)

Raw Monthly BMF

Untouched IRS monthly releases. Use these when you need to replicate analysis built on a specific raw file or verify upstream NCCS transformations.

URL pattern:

https://nccsdata.s3.us-east-1.amazonaws.com/raw/bmf/{YYYY-MM}-BMF.csv

Replace {YYYY-MM} with the desired vintage (e.g. 2026-01). See the IRS BMF documentation for the original schema.

Raw Legacy BMF

Original NCCS 501CX-NONPROFIT-PX vintages. Each row links the unmodified file and its profile (data dictionary). Most users should prefer the harmonized legacy BMF above — it has the same schema as the transformed monthly product.

Year Month File Download Profile Size
2022 08 BMF-2022-08-501CX-NONPROFIT-PX.csv Download Profile 345.8 mb
2022 01 BMF-2022-01-501CX-NONPROFIT-PX.csv Download Profile 377.1 mb
2020 04 BMF-2020-04-501CX-NONPROFIT-PX.csv Download Profile 375.9 mb
2019 08 BMF-2019-08-501CX-NONPROFIT-PX.csv Download Profile 374.9 mb
2018 12 BMF-2018-12-501CX-NONPROFIT-PX.csv Download Profile 330.5 mb
2017 12 BMF-2017-12-501CX-NONPROFIT-PX.csv Download Profile 339.5 mb
2017 09 BMF-2017-09-501CX-NONPROFIT-PX.csv Download Profile 344.8 mb
2016 08 BMF-2016-08-501CX-NONPROFIT-PX.csv Download Profile 368.3 mb
2016 04 BMF-2016-04-501CX-NONPROFIT-PX.csv Download Profile 368.4 mb
2016 03 BMF-2016-03-501CX-NONPROFIT-PX.csv Download Profile 367.7 mb
2016 02 BMF-2016-02-501CX-NONPROFIT-PX.csv Download Profile 366.4 mb
2015 12 BMF-2015-12-501CX-NONPROFIT-PX.csv Download Profile 363.7 mb
2015 11 BMF-2015-11-501CX-NONPROFIT-PX.csv Download Profile 363.5 mb
2015 09 BMF-2015-09-501CX-NONPROFIT-PX.csv Download Profile 359.9 mb
2015 07 BMF-2015-07-501CX-NONPROFIT-PX.csv Download Profile 361.1 mb
2015 05 BMF-2015-05-501CX-NONPROFIT-PX.csv Download Profile 358.5 mb
2015 04 BMF-2015-04-501CX-NONPROFIT-PX.csv Download Profile 356.7 mb
2015 02 BMF-2015-02-501CX-NONPROFIT-PX.csv Download Profile 353.5 mb
2014 12 BMF-2014-12-501CX-NONPROFIT-PX.csv Download Profile 351.1 mb
2014 11 BMF-2014-11-501CX-NONPROFIT-PX.csv Download Profile 349.5 mb
2014 09 BMF-2014-09-501CX-NONPROFIT-PX.csv Download Profile 343.2 mb
2014 06 BMF-2014-06-501CX-NONPROFIT-PX.csv Download Profile 337.5 mb
2014 04 BMF-2014-04-501CX-NONPROFIT-PX.csv Download Profile 336.3 mb
2014 02 BMF-2014-02-501CX-NONPROFIT-PX.csv Download Profile 334.5 mb
2013 12 BMF-2013-12-501CX-NONPROFIT-PX.csv Download Profile 331.7 mb
2013 10 BMF-2013-10-501CX-NONPROFIT-PX.csv Download Profile 339.7 mb
2013 09 BMF-2013-09-501CX-NONPROFIT-PX.csv Download Profile 331.2 mb
2013 08 BMF-2013-08-501CX-NONPROFIT-PX.csv Download Profile 330.7 mb
2013 07 BMF-2013-07-501CX-NONPROFIT-PX.csv Download Profile 335.3 mb
2013 06 BMF-2013-06-501CX-NONPROFIT-PX.csv Download Profile 334.7 mb
2013 05 BMF-2013-05-501CX-NONPROFIT-PX.csv Download Profile 334.2 mb
2013 04 BMF-2013-04-501CX-NONPROFIT-PX.csv Download Profile 333.9 mb
2013 03 BMF-2013-03-501CX-NONPROFIT-PX.csv Download Profile 332.3 mb
2013 02 BMF-2013-02-501CX-NONPROFIT-PX.csv Download Profile 337 mb
2012 12 BMF-2012-12-501CX-NONPROFIT-PX.csv Download Profile 339.4 mb
2012 11 BMF-2012-11-501CX-NONPROFIT-PX.csv Download Profile 338.9 mb
2012 10 BMF-2012-10-501CX-NONPROFIT-PX.csv Download Profile 338.7 mb
2012 08 BMF-2012-08-501CX-NONPROFIT-PX.csv Download Profile 338.7 mb
2012 07 BMF-2012-07-501CX-NONPROFIT-PX.csv Download Profile 338.4 mb
2012 06 BMF-2012-06-501CX-NONPROFIT-PX.csv Download Profile 339.2 mb
2012 04 BMF-2012-04-501CX-NONPROFIT-PX.csv Download Profile 310 mb
2012 03 BMF-2012-03-501CX-NONPROFIT-PX.csv Download Profile 336.1 mb
2012 02 BMF-2012-02-501CX-NONPROFIT-PX.csv Download Profile 308.2 mb
2011 12 BMF-2011-12-501CX-NONPROFIT-PX.csv Download Profile 342.6 mb
2011 11 BMF-2011-11-501CX-NONPROFIT-PX.csv Download Profile 342.2 mb
2011 10 BMF-2011-10-501CX-NONPROFIT-PX.csv Download Profile 341.4 mb
2011 09 BMF-2011-09-501CX-NONPROFIT-PX.csv Download Profile 340.8 mb
2011 08 BMF-2011-08-501CX-NONPROFIT-PX.csv Download Profile 342.1 mb
2011 07 BMF-2011-07-501CX-NONPROFIT-PX.csv Download Profile 340.9 mb
2011 06 BMF-2011-06-501CX-NONPROFIT-PX.csv Download Profile 338.5 mb
2010 11 BMF-2010-11-501CX-NONPROFIT-PX.csv Download Profile 374.9 mb
2010 08 BMF-2010-08-501CX-NONPROFIT-PX.csv Download Profile 382.1 mb
2010 07 BMF-2010-07-501CX-NONPROFIT-PX.csv Download Profile 415.5 mb
2010 05 BMF-2010-05-501CX-NONPROFIT-PX.csv Download Profile 411.6 mb
2010 04 BMF-2010-04-501CX-NONPROFIT-PX.csv Download Profile 410.8 mb
2010 01 BMF-2010-01-501CX-NONPROFIT-PX.csv Download Profile 380.4 mb
2009 10 BMF-2009-10-501CX-NONPROFIT-PX.csv Download Profile 382.5 mb
2009 07 BMF-2009-07-501CX-NONPROFIT-PX.csv Download Profile 306.7 mb
2009 04 BMF-2009-04-501CX-NONPROFIT-PX.csv Download Profile 303.4 mb
2009 01 BMF-2009-01-501CX-NONPROFIT-PX.csv Download Profile 323.3 mb
2008 12 BMF-2008-12-501CX-NONPROFIT-PX.csv Download Profile 299.1 mb
2008 10 BMF-2008-10-501CX-NONPROFIT-PX.csv Download Profile 304.9 mb
2008 06 BMF-2008-06-501CX-NONPROFIT-PX.csv Download Profile 293.3 mb
2008 04 BMF-2008-04-501CX-NONPROFIT-PX.csv Download Profile 291.1 mb
2008 01 BMF-2008-01-501CX-NONPROFIT-PX.csv Download Profile 286.1 mb
2007 09 BMF-2007-09-501CX-NONPROFIT-PX.csv Download Profile 297.7 mb
2007 04 BMF-2007-04-501CX-NONPROFIT-PX.csv Download Profile 304.6 mb
2007 01 BMF-2007-01-501CX-NONPROFIT-PX.csv Download Profile 311 mb
2006 11 BMF-2006-11-501CX-NONPROFIT-PX.csv Download Profile 316.3 mb
2006 05 BMF-2006-05-501CX-NONPROFIT-PX.csv Download Profile 316.8 mb
2006 01 BMF-2006-01-501CX-NONPROFIT-PX.csv Download Profile 318.4 mb
2005 11 BMF-2005-11-501CX-NONPROFIT-PX.csv Download Profile 314.2 mb
2005 07 BMF-2005-07-501CX-NONPROFIT-PX.csv Download Profile 306.7 mb
2004 12 BMF-2004-12-501CX-NONPROFIT-PX.csv Download Profile 300.3 mb
2004 04 BMF-2004-04-501CX-NONPROFIT-PX.csv Download Profile 308.9 mb
2003 11 BMF-2003-11-501CX-NONPROFIT-PX.csv Download Profile 309.3 mb
2003 07 BMF-2003-07-501CX-NONPROFIT-PX.csv Download Profile 314.2 mb
2003 01 BMF-2003-01-501CX-NONPROFIT-PX.csv Download Profile 285.5 mb
2002 07 BMF-2002-07-501CX-NONPROFIT-PX.csv Download Profile 286.7 mb
2002 01 BMF-2002-01-501CX-NONPROFIT-PX.csv Download Profile 289.3 mb
2001 07 BMF-2001-07-501CX-NONPROFIT-PX.csv Download Profile 272.2 mb
2000 05 BMF-2000-05-501CX-NONPROFIT-PX.csv Download Profile 262.5 mb
1999 12 BMF-1999-12-501CX-NONPROFIT-PX.csv Download Profile 253 mb
1998 09 BMF-1998-09-501CX-NONPROFIT-PX.csv Download Profile 249 mb
1997 10 BMF-1997-10-501CX-NONPROFIT-PX.csv Download Profile 242.2 mb
1996 06 BMF-1996-06-501CX-NONPROFIT-PX.csv Download Profile 236.6 mb
1995 08 BMF-1995-08-501CX-NONPROFIT-PX.csv Download Profile 223.4 mb
1989 06 BMF-1989-06-501CX-NONPROFIT-PX.csv Download Profile 123.3 mb


Documentation & Support

Last verified: 2026-07-01