US Census of Governments Finance API

State and local government finance — revenue, spending, and cash and security holdings — for every state, county, municipality and township the US Census Bureau surveyed from FY1967 through FY2024, as one public, read-only JSON API. No account, no key, nothing to download.

The corpus behind it harmonizes the Census Bureau’s Individual Unit Files across five decades of changing identifiers, item codes and file layouts, so a query for one government returns a continuous series instead of a dozen incompatible vintages. Scope is government types 0–3 — state, county, municipality and township; special districts and independent school districts are not included. FY1968 and FY1969 have no source data: the published years are a set, not an unbroken span, and GET /api/v1/ names the gaps in its scope string.

Quickstart

Every per-government route takes a 12-character canonical_govid in its path, and /governments?q= is what turns a place name into one. Start there.

curl

# 1. Resolve a name to its canonical_govid.
curl -s "https://cog-api.civilytics.org/api/v1/governments?q=Madison&state=WI&type=city"

# 2. Three snapshots of police spending, in 2024 dollars, per resident.
#    years= is a comma-separated list; there is no range syntax.
curl -s "https://cog-api.civilytics.org/api/v1/governments/552025209777/spending?category=Police&years=2014,2019,2024&adjust=real&base_year=2024&per_capita=true"

R

library(httr2)

base <- "https://cog-api.civilytics.org/api/v1"

# 1. Resolve a place name to its 12-character canonical_govid.
gov <- request(base) |>
  req_url_path_append("governments") |>
  req_url_query(q = "Madison", state = "WI", type = "city") |>
  req_perform() |>
  resp_body_json()

govid <- gov$data[[1]]$canonical_govid

# 2. Pull a real, per-capita police spending series for that government.
spend <- request(base) |>
  req_url_path_append("governments", govid, "spending") |>
  req_url_query(category = "Police", years = "2014,2019,2024",
                adjust = "real", base_year = 2024, per_capita = "true") |>
  req_perform() |>
  resp_body_json()

for (row in spend$data) {
  cat(sprintf("%d  %-10s $%.2f per capita (2024 dollars)\n",
              row$year, row$spend_subtype, row$amt_per_capita_real))
}

Python

import requests

BASE = "https://cog-api.civilytics.org/api/v1"

# 1. Resolve a place name to its 12-character canonical_govid.
gov = requests.get(f"{BASE}/governments",
                   params={"q": "Madison", "state": "WI", "type": "city"},
                   timeout=60).json()
govid = gov["data"][0]["canonical_govid"]

# 2. Pull a real, per-capita police spending series for that government.
spend = requests.get(f"{BASE}/governments/{govid}/spending",
                     params={"category": "Police", "years": "2014,2019,2024",
                             "adjust": "real", "base_year": 2024,
                             "per_capita": "true"},
                     timeout=60).json()

for row in spend["data"]:
    print(f'{row["year"]}  {row["spend_subtype"]:<10} '
          f'${row["amt_per_capita_real"]:.2f} per capita')

Key concepts

Enough to keep you out of the common traps. The API’s own docs are the authority on parameters — see the links at the end of this section.

One envelope, always. Every response is {status, data, error, meta, provenance, suggestions}. provenance records the exact underlying call, so a figure in a report can be traced back to the query that produced it.
An empty data is a real answer. status: "success" with data: [] means the government reported nothing matching your filters — a common, genuine condition in this data. It never means a bad ID or an unavailable year; those are 400s with their own messages.
Coverage varies enormously by year. The Census of Governments is a complete census only in years ending in 2 or 7. Every other year is a sample, and the sample varies: of Wisconsin’s 608 cities, 597 reported in FY2012 and 112 in FY2019. The peer endpoints report n_units_expected against n_units_reporting so a thin year cannot mislead you silently.
Money parameters. adjust=nominal|real with base_year=YYYY, per_capita=true|false, and basis=harmonized|raw. Pin base_year explicitly if you want a figure that stays stable as new years land.
Concepts change the number. expenditure_concept=primary|direct|total (default primary) and revenue_concept=general|total (default general) select which Census concept you get. The defaults are the narrower ones: a city’s general revenue excludes utility and liquor-store revenue, which is 15.9% of city revenue across the corpus. Ask for total when you need it.
Unknown parameters are rejected, never ignored. A misspelled filter is a 400 that names the offending key, so a typo can’t quietly widen your query.
Vocabulary. /categories lists the metric dictionary; /recipes lists the harmonization recipes that stitch multi-code, cross-vintage series together.

Full parameter contract: llms.txt (written for agents, but the most complete prose reference), the data dictionary, and the interactive Swagger docs.

Guides and specification

Working with the COG Finance API — a Southern guide — an executable walkthrough for state, county and city analysis across the Census South Region. Twelve worked charts — healthcare access across the states, environmental justice in the counties, economic security in the cities — every figure fetched live from this API at the moment the page was built, with the exact call shown beneath each one. Ends with Tableau connection guidance and a plain-spoken list of known limitations.
API specification (PDF) — the formal reference, describing the API as built. Its endpoint, parameter, category and scope tables are generated from the live server at render time, so the document cannot silently drift from the deployment it describes.

Citation

The corpus and the API are separate artifacts with separate versions; cite whichever you actually used, and cite the Census Bureau as the underlying source in both cases.

The corpus:

Jared Knowles and Hannah Miller. 2026. “US Census of Governments Finance Corpus (FY1967–FY2024).” Civilytics Consulting. Available online at: https://huggingface.co/datasets/civilytics/us-cog-finance

Released under CC-BY-4.0. main is the rolling latest; for an immutable reference, cite the commit SHA (resolve/<sha>/) rather than a version tag.

The API:

Jared Knowles and Hannah Miller. 2026. “US Census of Governments Finance API (v1).” Civilytics Consulting. Available online at: https://cog-api.civilytics.org/api/v1/

The underlying source, which both of the above derive from:

US Census Bureau. Annual Survey of State and Local Government Finances and Census of Governments, Individual Unit Files, FY1967–FY2024. Available online at: https://www.census.gov/programs-surveys/gov-finances.html

Acknowledgements

This work was sponsored by the Southern Economic Advancement Project, whose support paid for the multi-year harmonization that makes a single query span fifty-eight years of changing federal file formats. The interpretations here are Civilytics’ own.