Datasetpublic

piercewetter3/irs-990-parsed

IRS 990 Parsed Nonprofit Database Public relational extract of IRS Form 990 / 990-EZ / 990-PF filings, plus the colocated public files we join for address research: CMS NPPES + T-MSIS Medicare spend, FMCSA DOT carriers, OFAC SDN, FEC committees, and the IRS EO BMF. Generated: 2026-08-17Tables: 34Rows (sum): 459,069,505License: CC0 / public domain — derived from U.S. government recordsHub: https://huggingface.co/datasets/piercewetter3/irs-990-parsed Layout Tables… See the full description on the dataset page: https://huggingface.co/datasets/piercewetter3/irs-990-parsed.

sourceHugging Facecc0-1.0updated 1mo agoView on Hugging Face

0likes742downloads

Dataset Card

IRS 990 Parsed Nonprofit Database

Public relational extract of IRS Form 990 / 990-EZ / 990-PF filings, plus the colocated public files we join for address research: CMS NPPES + T-MSIS Medicare spend, FMCSA DOT carriers, OFAC SDN, FEC committees, and the IRS EO BMF.

Generated: 2026-08-17 Tables: 34 Rows (sum): 459,069,505 License: CC0 / public domain — derived from U.S. government records Hub: https://huggingface.co/datasets/piercewetter3/irs-990-parsed

Layout

Tables with ≥ 2,000,000 rows are a directory of ZSTD Parquet shards (~256MB each). Smaller tables are a single Table.parquet. UUID columns are exported as VARCHAR.

sql

INSTALL parquet; LOAD parquet;
SELECT * FROM 'Charities.parquet' LIMIT 5;
SELECT COUNT(*) FROM 'medicare_provider_spending/*.parquet';

Tables

Table	Rows	Files	What
`Addresses`	94,102,876	dir	Normalized streets for every owner type (990, Medicare, DOT, OFAC, FEC)
`AuthoritativeEin`	1,233,305	1 parquet	Canonical name → EIN seeds used by grant matching
`BMF`	1,952,238	1 parquet	IRS EO Business Master File including street/city/state/ZIP
`Backfill`	1,720,379	1 parquet	Einless grantee backfill (recipient EIN resolved after ingest)
`Charities`	3,625,588	dir	Parsed Form 990 / 990-EZ / 990-PF filer years
`Contractors`	1,096,522	1 parquet	Highest-paid independent contractors from 990 Part VII
`Contributions`	0	1 parquet	Incoming contribution rows (currently empty in this build)
`Geocoding`	14,415,456	dir	Geocode attempts and rounded lat/lon for master addresses
`Grants`	26,216,672	dir	Grants paid (Schedule I / PF grants)
`IrsBmf`	1,952,238	1 parquet	BMF without raw street fields (legacy export table)
`Officers`	4,430,012	dir	Officer / key-employee compensation
`PoliticalContributions`	0	1 parquet	527 / political outlays from 990 (currently empty)
`XmlFiles`	3,625,588	dir	One row per IRS XML filing ingested
`ZipFiles`	51	1 parquet	IRS annual ZIP archives fetched
`Zips`	41,706	1 parquet	US ZIP centroids used for zip-slice reports
`base_name_ein`	2,144,568	dir	Distinct grantee-name → EIN rollup from grants
`dot_carriers`	4,454,157	dir	FMCSA motor-carrier census
`fec_candidate_spendings`	703,597	1 parquet	FEC candidate spend (city/state/ZIP, not street)
`fec_committee_transactions`	18,667,435	dir	FEC committee-to-committee transfers
`fec_committees`	6,980	1 parquet	FEC committees with a real line-1 street
`fec_individual_contributions`	29,104,378	dir	FEC individual contributions (city/state/ZIP)
`fec_operating_expenditures`	2,249,158	dir	FEC operating expenditures
`hcpcs_codes`	9,006	1 parquet	HCPCS procedure code lookup
`medicare_provider_hcpcs`	5,876,682	dir	Medicare spend rolled up by NPI × HCPCS
`medicare_provider_rollup`	617,492	1 parquet	Medicare spend rolled up by billing NPI
`medicare_provider_spending`	230,154,264	dir	Medicare T-MSIS line grain (billing × servicing × HCPCS × month)
`medicare_providers`	9,606,683	dir	NPPES provider enumeration
`name_mapping`	928,108	1 parquet	Cleaned name → winning EIN from traditional matching
`noc_codes`	0	1 parquet	NOC code lookup
`nppes_code_values`	885	1 parquet	NPPES codebook (entity type, taxonomy, etc.)
`sanctioned_entities`	19,199	1 parquet	OFAC SDN entities
`sanctioned_identifiers`	22,491	1 parquet	OFAC identifiers (passport, tax id, …)
`sanctioned_names`	49,652	1 parquet	OFAC primary names and aliases
`sanctioned_programs`	42,139	1 parquet	OFAC program codes on each entity

Not uploaded

Ops / scratch / raw extracts stay local:

Table	Why
`PendingCanonicals`	empty address-dedup scratch
`PipelineProgress`	local pipeline job state
`Zips_raw`	source extract; use Zips
`_meta_clustering`	DuckDB clustering bookkeeping
`fec_raw_committees`	1-row ingest leftover; use fec_committees
`grant_update_map`	grant_match scratch
`medicare_raw_spending`	238M raw CMS extract; use medicareproviderspending
`temp_gin_batch`	grant_match scratch

Notes

—Addresses is polymorphic (owner_id + address_type). Types include charity, officer, contractor, grant, medicare, dotcarrier, ofacsanction, and fec_committee.
—medicare_provider_spending is the line grain (~230M). Use medicare_provider_rollup / medicare_provider_hcpcs for cheap totals.
—FEC contributor / spend / oppexp / candidate-spend rows are city+state+ZIP, not street. Only fec_committees has a real line 1.
—BMF keeps IRS street fields; IrsBmf is the older street-stripped copy.
—Full DDL: schema_duckdb.sql. Export inventory: manifest.json.