abigailhaddad/usaspending-bulk-awards

USAspending bulk awards — contracts & assistance

Clean, partitioned, query-ready Parquet mirror of the public USAspending Award Data Archive (prime contract and financial-assistance transactions, FY2007–present, all agencies).

The source publishes 4,600 per-agency ZIP/CSV files (830 GB uncompressed). This dataset normalizes them to typed, zstd-compressed Parquet (~8× smaller) with amount columns as double and date columns as date, partitioned for fast predicate-pushdown querying.

Layout

Two datasets (the schemas differ — contracts 297 columns, assistance 112 columns), Hive-partitioned by fiscal year and agency:

contracts/fiscal_year=YYYY/agency=CODE/part.parquet     # prime contracts (297 cols)
assistance/fiscal_year=YYYY/agency=CODE/part.parquet    # grants/loans/etc. (112 cols)
reference/*.parquet                                     # joinable dimension tables (below)

Reference tables

File	Rows	Description
reference/data_dictionary.parquet	457 rows	Rosetta crosswalk — every column's definition, domain values, and source file
reference/toptier_agencies.parquet	111 rows	Agency code ↔ name crosswalk
reference/def_codes.parquet	51 rows	Disaster Emergency Fund Codes (COVID-19 = L–U, IIJA = Z & 1)
reference/glossary.parquet	151 rows	Plain-language term definitions
reference/assistance_listing.parquet	56 rows	CFDA / assistance program catalog

The data dictionary documents every column in both products.

Quick start (DuckDB)

import duckdb
con = duckdb.connect()
# Top recipients of FY2024 contracts at one agency — partition pruning reads only that slice
con.sql('''
  SELECT recipient_name, sum(federal_action_obligation) AS obligated
  FROM read_parquet('contracts/**/*.parquet', hive_partitioning=true)
  WHERE fiscal_year = '2024' AND agency = '097'
  GROUP BY 1 ORDER BY 2 DESC LIMIT 10
''').show()

Provenance & updates

• Source: USAspending.gov Award Data Archive (U.S. Government public-domain data).
• Coverage: FY2007–present, prime awards only. Subawards and account File A/B/C are not in the source archive (they come from USAspending's Custom Bulk Download API).
• Updates: refreshed when the source re-publishes; only changed (agency, FY) slices are re-processed (tracked by file ETag). Latest snapshot only.
• Not the full USAspending Postgres database dump.