uscogdata

A curated R reader for the Civilytics US Census of Governments finance corpus — every dollar that US state, county, municipal and township governments reported raising and spending, from FY1967 to FY2024, in one queryable place.

The Census of Governments is the only nationwide source for local government finance, and it is hard to use: item codes change meaning across vintages, government identifiers were renumbered in 2017, and an absent value means "published zero" in one era and "not reported" in the next. This package handles each of those problems, and it tells you when it has — every result carries provenance describing what was converted, what was aggregated, and which known series breaks intersect your query.

Scope: government types 0–3 (state, county, municipality, township). 56 fiscal years, 46,148,034 rows, ~201 MB. There is no source data for FY1968 or FY1969. Special districts (type 4) and school districts (type 5) are excluded pending validation.

Where the data comes from

The corpus is published and documented at the US Census of Governments Finance API. Start there for how the data was built, how the identifier and item-code reconciliation works, and what the corpus does and does not cover.

• API documentation and walkthroughs — reference, data dictionary, and worked examples such as the Southern states guide
• Live API — the same corpus over HTTP, for Tableau, Python, or anything that isn't R
• Bulk corpus on Hugging Face — CC-BY-4.0; the same parquet files this package reads
• Census Bureau source data — the underlying public files

Install

install.packages("uscogdata",
  repos = c("https://civilytics.r-universe.dev",
            "https://cloud.r-project.org"))

Or from source:

pak::pkg_install("git::https://gitea.civilytics.org/Civilytics/uscogdata.git")

Quickstart

No configuration, no credentials, no download. The package reads the published corpus over HTTPS by default.

library(uscogdata)

# Resolve a place name to a canonical government id
madison <- cog_gov_search(name = "Madison", state = "WI", type = 2)
madison$canonical_govid
#> [1] "552025209777"

# Police spending, inflation-adjusted and per capita
spend <- cog_spending(
  madison$canonical_govid,
  years          = 2012:2022,
  category       = "Police",
  per_capita     = TRUE,
  adjust_to_year = 2023
)

# What did that result do to the numbers, and what should you know about them?
cog_explain(spend)

years is required — there is no implicit full-history default.

Two ways to read the corpus

	Remote (default)	Mirrored
Setup	none	cog_mirror(dest), ~201 MB once
Disk used	0 MB — HTTP range requests only	~201 MB
Opening a session	~7.5 s	~0.1 s
One government, one year	~4 s	~0.05 s
One government, full history	~7 s	~0.1 s
Later queries, same session	~1.5 s	~0.05 s
Good for	trying it out, teaching, one-off questions	repeated analysis, offline work, reproducibility

A local mirror is roughly 60–80x faster, and it is one function call. That is by far the largest difference any of these settings makes. If you are going to ask more than a handful of questions, mirror first.

Measured 2026-08-10 on a 16-core Linux workstation against the published corpus (schema v7, pipeline_commit 3d28ddd), fresh R session per arm. A one-off question costs about 12 seconds end to end remotely and 0.15 seconds mirrored, session setup included.

Two things the per-query rows hide:

• Opening the session is the single largest remote cost — larger than any one query. It fetches the manifest and registers 23 SQL views over HTTPS, and it lands on your first query, not on library(uscogdata).
• The cost is network round-trips, not scanning. A repeat query against partitions this session has already touched is ~1.5 s rather than ~4 s, and a full-history query costs ~7 s whether it runs first or last. What you are paying for is reaching each of the 56 yearly files over HTTPS the first time.

Nothing is written to disk in remote mode: DuckDB fetches the parquet footer, works out which row groups it needs, and reads only those. Nothing is cached between sessions either, so every query goes back to the network — and a session that issues many remote queries in quick succession can be rate-limited by the host (HTTP Error: ... 429). Both are further reasons to mirror for real work.

The default points at a public HuggingFace mirror of the corpus. If you would rather not depend on a third party — for reproducibility, for an air-gapped environment, or on principle — the escape hatch is one function call:

cog_mirror("~/cog-corpus")
Sys.setenv(USCOGDATA_URL = "~/cog-corpus/")

After that, nothing in your analysis touches an external service.

Configuration

• USCOGDATA_URL — corpus root: an HTTPS URL or a local path, trailing slash required
• USCOGDATA_CACHE_DIR — where the manifest is cached (default: user cache dir)
• USCOGDATA_MANIFEST_TTL_SECS — manifest re-fetch interval (default 3600)
• USCOGDATA_DUCKDB_THREADS — cap DuckDB's thread count (default: every visible core)
• USCOGDATA_DUCKDB_MEMORY_LIMIT — cap DuckDB's memory, e.g. "4GB" (default: DuckDB's own)

Each also has an options() spelling — uscogdata.url, uscogdata.duckdb_threads, and so on — and the environment variable wins where both are set.

The two DuckDB caps exist for servers, not laptops. Unset, DuckDB claims every core it can see, which is right for one interactive session on your own machine and wrong when several readers share a box: each claims the whole machine and they fight. Capping costs roughly 5% on a single query and is worth it anywhere the process is sharing hardware.

Amounts are in full US dollars

Every amount column this package returns — amt_nominal, amt_real, amt_per_capita_nominal, amt_per_capita_real — is in full US dollars.

The raw Census source files report thousands of dollars, and the corpus's own amt column preserves that. The verbs multiply by 1000 on the way out, so you never have to. The conversion is recorded in every result:

attr(spend, "provenance")$transformations$units_conversion
#> $applied     TRUE
#> $source_unit "$1,000s (raw Census)"
#> $target_unit "$USD"
#> $multiplier  1000

Do not multiply again. If you have read elsewhere that COG amounts are in $1,000s — which is true of the raw Census files and of the corpus's own amt column — that rule does not apply to anything a cog_*() verb hands you. Applying it twice overstates every figure by 1000x, and the result looks plausible rather than obviously wrong.

Concepts worth understanding before you publish a number

Primary vs Direct vs Total spending

cog_spending(..., expenditure_concept = c("primary", "direct", "total")) controls whose spending a result counts. Concepts are defined as sets of the crosswalk's spend_subtype values, never item-code first letters — the letter Y alone spans revenue, expenditure and balance codes.

• "primary" (default) — the government's own service provision: current operations, capital outlay, assistance payments.
• "direct" — Census's published Direct Expenditure: primary plus interest on debt and insurance trust benefits (e.g. pensions).
• "total" — adds the intergovernmental leg, money handed to other governments to spend. Meaningful for one government's own budget over time, but it double-counts when summed across governments: a state's payment to a county is the same dollar the county reports as its own direct spending.

Rule of thumb: any figure spanning more than one government uses primary or direct. cog_geographic_rollup() and cog_peer_compare() enforce that by refusing "total" outright. Worked examples in vignette("total-spending", package = "uscogdata").

General vs Total revenue

cog_revenue(..., revenue_concept = c("general", "total")):

• "general" (default) — Census General Revenue: own-source taxes, charges and miscellaneous, plus federal, state and local aid.
• "total" — General plus utility revenue (A91–A94), liquor store revenue (A90), and insurance trust revenue.

Census defines these by its own identity:

Total Revenue = General + Utility + Liquor Store + Insurance Trust

Two things to know before switching to "total". Utility revenue is large for cities — measured on the bundled fixture, utility plus liquor store is 15.9% of city revenue, against 1.2% for states and 1.7% for counties. And the employee-retirement (X) codes stop at FY2016, when those systems moved to the separate Annual Survey of Public Pensions, so a "total" series steps down at the FY2016/FY2017 boundary for reasons of collection scope, not revenue (series breaks SB197–SB209).

Reporting coverage: the Census is only sometimes a census

The Census of Governments is a complete enumeration only in years ending in 2 and 7. Every other year is a sample, and the sample varies enormously — measured on the bundled fixture, Wisconsin's 608-city universe rolls up 597 governments in FY2012 and 112 in FY2019.

A statewide total resting on a fifth of the universe looks exactly like one resting on all of it, so every multi-government result now says which it is:

attr(rollup, "provenance")$coverage
# per-year n_units_expected, n_units_collected, n_units_reporting, is_census_year

cog_geographic_rollup(), cog_peer_compare() and cog_find_peers() take a coverage argument — "all" (default), "census" (census years only), or "consistent" (only units reporting in every requested year, a balanced panel).

n_units_reporting is category-conditional: it counts governments with rows for the specific category you asked for, so a government that was surveyed and genuinely spends nothing in that category is indistinguishable from one never surveyed — it is not a response rate on its own. n_units_collected is the number that separates them: governments present in the corpus that year for any category. n_units_collected / n_units_expected is the true collection rate; n_units_reporting / n_units_collected is category participation among collected units.

Absent cells mean two different things

Before FY2012, an absent cell means Census published $0. From FY2012 on, it means not reported. cog_spending(..., complete = TRUE) fills the requested grid and labels every row with which it is, via value_source:

value_source	meaning	amt_nominal
reported	the corpus carries this cell	as published
census_zero	dense-source year (≤ FY2011), absent — Census published $0	0
not_reported	sparse-source year (≥ FY2012), absent — unknown	NA

That NA is deliberate. Filling a modern absence with 0 would invent data.

Series breaks surface on their own

Catalogued breaks that intersect your query appear in provenance whether or not you went looking for them — series_break_refs for breaks in a specific item code, and corpus_break_refs for caveats about the corpus as a whole (dollar precision across the 1976/1977 boundary, the FY2017 identifier change, the FY2012 dense→sparse representation change). cog_explain() prints both.

How to cite

citation("uscogdata")

The corpus itself is published under CC-BY-4.0. Cite it as:

Civilytics Consulting. US Census of Governments finance corpus. https://huggingface.co/datasets/civilytics/us-cog-finance

Contributing

Development happens on Gitea; GitHub is a mirror that accepts issues and pull requests. See CONTRIBUTING.md for how a patch gets from there to here.

License

MIT © Civilytics Consulting LLC. See LICENSE.md.