Nason/qcew-database

QCEW 2025 Annual Employment and Wages

A version-pinned DuckDB edition of the U.S. Bureau of Labor Statistics Quarterly Census of Employment and Wages, 2025 annual averages: 3,096,467 rows across 6 tables (48 columns), all 3,089,813 annual source rows plus five version-matched reference tables and a disclosure-aware view.

Use it for descriptive questions about covered employment, establishments and nominal pay by place, industry and ownership. It is establishment-based industry data, not occupational wages, unique people, an employer panel or a causal analysis.

Built with qcew-database. Published to Hugging Face on 2026-10-18. The dataset is included in the public datapond catalog as qcew. Use normal package discovery and connections, or the immutable direct URL below.

Quick start · Tables · Views · Examples · Caveats · Source · License · Changelog

Quick Start

Choose one route: the datapond package finds the database by catalog ID; the direct Python, R and SQL examples use a fixed artifact URL. These are alternatives, not steps that must all be run. The SQL block requires a DuckDB SQL client.

DuckDB CLI

SET memory_limit = '128MiB';
SET threads = 1;
SET max_temp_directory_size = '0B';
INSTALL httpfs;
LOAD httpfs;
ATTACH 'https://huggingface.co/datasets/Nason/qcew-database/resolve/334c467dba413928233fb1d8cd1c0ff8de5c0105/qcew.duckdb' AS qcew (READ_ONLY);

-- The one published national all-industry, all-ownership total. Never sum the whole table.
SELECT area_title, annual_avg_estabs, annual_avg_emplvl, total_annual_wages, avg_annual_pay
FROM qcew.annual
WHERE area_fips = 'US000' AND own_code = '0' AND industry_code = '10'
  AND agglvl_code = '10' AND size_code = '0' AND year = '2025' AND qtr = 'A';

The database file is 186.5 MiB. Filtered queries fetch only the byte ranges they touch; an unfiltered scan of annual_raw reads most of the file, so download it locally if you plan to scan repeatedly. File size is disk storage, not a RAM estimate. Remote queries still run on your machine and can transfer much of the file. A filter or LIMIT after a join or aggregate does not guarantee a small scan. The 128 MiB setting limits DuckDB's buffer pool, not the whole Python/R process; with spill disabled, a larger query can fail at that limit. Download only if you have room for the file, temporary download bytes and any existing copy.

The examples use this immutable artifact URL:

https://huggingface.co/datasets/Nason/qcew-database/resolve/334c467dba413928233fb1d8cd1c0ff8de5c0105/qcew.duckdb

Python (duckdb)

Install the Python dependency once in the environment that runs your query: python -m pip install duckdb.

import duckdb
con = duckdb.connect(config={"memory_limit": "128MiB", "threads": "1", "max_temp_directory_size": "0B"})
con.sql("INSTALL httpfs; LOAD httpfs;")
con.sql("ATTACH 'https://huggingface.co/datasets/Nason/qcew-database/resolve/334c467dba413928233fb1d8cd1c0ff8de5c0105/qcew.duckdb' AS qcew (READ_ONLY)")
con.sql("SELECT * FROM qcew._metadata").show()
con.sql("""
    SELECT area_fips, area_title, annual_avg_emplvl, avg_annual_pay, disclosure_code
    FROM qcew.annual
    WHERE agglvl_code = '74' AND own_code = '5' AND industry_code = '62'
      AND right(area_fips, 3) <> '999' AND annual_avg_emplvl > 0
    ORDER BY annual_avg_emplvl DESC, area_fips LIMIT 20
""").show()
con.close()

R with DBI and duckdb

Install the R dependencies once: install.packages(c("DBI", "duckdb")).

library(DBI)
con <- dbConnect(duckdb::duckdb(config = list(memory_limit = "128MiB", threads = "1", max_temp_directory_size = "0B")))
dbExecute(con, "INSTALL httpfs")
dbExecute(con, "LOAD httpfs")
dbExecute(con, "ATTACH 'https://huggingface.co/datasets/Nason/qcew-database/resolve/334c467dba413928233fb1d8cd1c0ff8de5c0105/qcew.duckdb' AS qcew (READ_ONLY)")
dbGetQuery(con, "SELECT area_fips, area_title, annual_avg_emplvl, avg_annual_pay
                 FROM qcew.annual
                 WHERE agglvl_code = '55' AND own_code = '5' AND industry_code = '621'
                 ORDER BY annual_avg_emplvl DESC NULLS LAST, area_fips")
dbDisconnect(con, shutdown = TRUE)

In the tested R DuckDB 1.5.5 configuration, BIGINT and DECIMAL results are exposed as doubles. Employment and wage integers here are below 2^53 and exact; location quotients and percentage changes may be approximate in R although the stored DECIMAL values are exact.

datapond package

Install or upgrade the Python client with python -m pip install --upgrade datapond, or install the released R client from CRAN with install.packages("datapond"). Remote examples are tested with Python datapond 0.1.5 (DuckDB 1.5.4) and CRAN R datapond 0.1.0 (DuckDB 1.5.5); older DuckDB reader versions are not verified here. R 0.1.1 is a separate development version, not the CRAN release.

The examples below use the public default catalog. For limits set before attachment, use the direct DuckDB examples above; package examples here set limits immediately after connecting:

import datapond
print(datapond.list())
datapond.describe('qcew', table='annual_raw')
con = datapond.connect('qcew')
con.execute("SET memory_limit='128MiB'; SET threads=1; SET max_temp_directory_size='0B'")
con.sql("SELECT annual_avg_emplvl, total_annual_wages FROM annual WHERE area_fips = 'US000' AND own_code = '0' AND industry_code = '10' AND agglvl_code = '10'").show()
con.close()

library(datapond)
dp_list()
con <- dp_connect('qcew')
DBI::dbExecute(con, "SET memory_limit='128MiB'; SET threads=1; SET max_temp_directory_size='0B'")
dp_describe(con, table = 'annual_raw')
DBI::dbGetQuery(con, "SELECT annual_avg_emplvl, total_annual_wages FROM annual WHERE area_fips = 'US000' AND own_code = '0' AND industry_code = '10' AND agglvl_code = '10'")
dp_disconnect(con)

For repeated scans, download once and connect locally:

import datapond
datapond.download('qcew')
con = datapond.connect('qcew', local=True)
con.sql("SELECT annual_avg_emplvl, total_annual_wages FROM annual WHERE area_fips = 'US000' AND own_code = '0' AND industry_code = '10' AND agglvl_code = '10'").show()
con.close()

library(datapond)
dp_download('qcew')
con <- dp_connect('qcew', local = TRUE)
DBI::dbGetQuery(con, "SELECT annual_avg_emplvl, total_annual_wages FROM annual WHERE area_fips = 'US000' AND own_code = '0' AND industry_code = '10' AND agglvl_code = '10'")
dp_disconnect(con)

Verify the recorded artifact SHA-256 after downloading. In Python datapond 0.1.5 as well as 0.1.1, the optional huggingface_hub download path follows the repository default branch rather than the pinned URL. For an exact-version copy, download the immutable URL shown above and verify its checksum; see consumer verification for the tested configuration.

Default-catalog discovery and representative remote queries were verified after the registry merge. Earlier full-download checks used an isolated candidate registry. See consumer verification for exact client versions, paths, resource limits and download behavior. The direct URL examples above work independently of the catalog. The SQL examples below use the qcew catalog alias from the direct-URL quick start.

Tables

One annual data row is one area × ownership group × industry × aggregation level × size class for 2025. These rows include overlapping totals and detail; choose one level before summing. Start with the annual view so suppressed measures become NULL.

Table	Description	Rows	Columns
annual_raw	2025 annual QCEW rows preserving all 38 source fields, numeric placeholders and disclosure flags. Mixed nested geographies, industries and ownership totals; do not sum unfiltered	3,089,813	38
areas	Distinct exact area code/title pairs from the pinned 2025 BLS by-area archive	4,451	2
industries	Distinct exact industry code/title pairs from the pinned archive	2,159	2
aggregation_levels	Distinct exact aggregation-level code/title pairs	36	2
ownerships	Distinct exact ownership code/title pairs	7	2
size_classes	Code 0, all establishment sizes	1	2

The raw key is (area_fips, own_code, industry_code, agglvl_code, size_code, year, qtr). All key fields are text, including leading-zero FIPS codes; year='2025', qtr='A' (annual, not a fifth quarter), size_code='0'. Reference tables cover codes used in this snapshot, not a historical code list; every code had one exact title. _metadata has one row per data table; _columns carries column definitions.

Pre-built Views

View	What it does
annual	annual_raw with the full key, five labels joined through the unique dictionaries, and the three disclosure flags applied independently

In annual, disclosure_code='N' sets employment, wages, contributions, weekly wage and annual pay to NULL; lq_disclosure_code='N' nulls the six location quotients; oty_disclosure_code='N' nulls the over-the-year changes. The four establishment measures (annual_avg_estabs, lq_annual_avg_estabs, oty_annual_avg_estabs_chg, oty_annual_avg_estabs_pct_chg) are always retained as published. annual_raw keeps every source value, including placeholder zeroes on suppressed rows. The view is not counted in the row and table totals above.

Example Queries

County disclosure coverage for private health care and social assistance, including valid establishment counts on suppressed rows:

SELECT disclosure_code, count(*) AS source_cells,
       count(annual_avg_emplvl) AS published_employment_cells,
       sum(annual_avg_estabs) AS published_establishments_rounded
FROM qcew.annual
WHERE agglvl_code = '74' AND own_code = '5' AND industry_code = '62'
  AND right(area_fips, 3) <> '999'
GROUP BY disclosure_code ORDER BY disclosure_code;

State private ambulatory health care (NAICS 621) with published year-over-year change and its own flag:

SELECT area_fips, area_title, annual_avg_emplvl, avg_annual_pay,
       oty_annual_avg_emplvl_pct_chg, oty_disclosure_code
FROM qcew.annual
WHERE agglvl_code = '55' AND own_code = '5' AND industry_code = '621'
ORDER BY annual_avg_emplvl DESC NULLS LAST, area_fips;

Compare two disjoint subsectors after fixing the grain to one row per county and industry:

WITH selected AS (
  SELECT area_fips, area_title, industry_code, annual_avg_emplvl, avg_annual_pay
  FROM qcew.annual
  WHERE agglvl_code = '75' AND own_code = '5'
    AND industry_code IN ('621', '622') AND right(area_fips, 3) <> '999'
)
SELECT a.area_fips, a.area_title,
       a.annual_avg_emplvl AS ambulatory_jobs, h.annual_avg_emplvl AS hospital_jobs,
       a.avg_annual_pay AS ambulatory_pay, h.avg_annual_pay AS hospital_pay
FROM selected a JOIN selected h ON a.area_fips = h.area_fips
WHERE a.industry_code = '621' AND h.industry_code = '622'
ORDER BY a.area_fips LIMIT 20;

The full set is in examples.sql; saved results are bound to the artifact hash.

Researcher Caveats

1. Suppression flags are independent and partial. Three flags (disclosure_code, lq_disclosure_code, oty_disclosure_code) each govern their own measures. There are 1,860,359 suppressed level rows, 1,872,172 suppressed quotient rows and 1,937,791 suppressed change rows. Suppressed rows still carry valid establishment counts. Never read a suppressed zero as zero employment or pay.

2. Suppressed cells are inside published totals. Summing disclosed detail gives an incomplete total; the view does not estimate or back out suppressed values. Report how many eligible cells have published values (BLS disclosure note).

3. Rows overlap; choose one level. National, state, county, metro and combined-area rows overlap, industry totals nest, and ownership 0 (total), 8 (total government) and 9 (UI covered excluding federal) overlap sectors 1, 2, 3 and 5. Filter on agglvl_code (for example 74 county/sector/ownership, 75 county/3-digit, 55 state/3-digit), not on code length.

4. Geography moves. Connecticut uses planning-region codes from 2024 data; codes ending 999 are unallocated. Check areas before fixing a study universe.

5. Industry is 2022 NAICS. 62 is health care and social assistance; 621 ambulatory care; 622 hospitals. Do not splice historical series without checking classification changes.

6. Employment counts covered jobs, not people. A worker with two covered employers counts twice; the self-employed and other uncovered groups are absent. Annual employment averages 12 monthly levels; establishments average four quarters; rounding can break exact additivity.

7. Pay is a nominal ratio. Average annual and weekly pay are not hourly wages and not additive; do not average county pay to estimate national pay or build weighted aggregates where components are suppressed.

8. Changes are published, not computed. Over-the-year fields come from BLS and reflect revisions, reclassification and coverage changes; they are not causal effects. There is no separate 2024 table.

9. Published zeros can be structural. Some taxable-wage and contribution fields do not apply to particular ownerships, so a zero is not necessarily economic absence.

10. Source anomalies remain visible. DC codes 11000 and 11001 both carry “District of Columbia, not unknown”. Los Angeles County’s private health care/social assistance row has 287,445 establishments and 859,824 jobs; this snapshot does not establish why the establishment count is unusually large.

Data Source and Version

Bureau of Labor Statistics, Quarterly Census of Employment and Wages, 2025 annual singlefile, acquired 2026-10-18. Pinned source ZIP: 62,799,396 bytes, SHA-256 94ebbc589b1b500671c1fd507e2589a9eeb24dcd30eb07f1ce37b0018643f865; labels derived from the official 2025 annual-by-area ZIP (114,997,862 bytes). Both archives were last modified by BLS on August 21, 2026. The national row reports 155,732,363 jobs and $12,259,226,899,972 in wages. BLS may revise annual files; this is the exact acquired snapshot, with no automatic refresh promised.

Artifact: qcew.duckdb, 195,571,712 bytes, SHA-256 89bf1cef54f42f821965df6e9079b24fc7c4c0c86046fadc98e6cedf9ebd8a1d, dataset commit 334c467dba413928233fb1d8cd1c0ff8de5c0105.

License and Attribution

BLS data are public domain; BLS requests source attribution (BLS copyright information). Cite: U.S. Bureau of Labor Statistics, Quarterly Census of Employment and Wages, 2025 annual, acquisition date and artifact hash. This derivative changes representation, adds dictionaries, metadata and the annual view, and preserves all source values in annual_raw. No BLS endorsement is implied.

GitHub, Dictionary and Website

Detailed researcher guide · Build and update instructions · Verified publication

• Source, build scripts, tests and evidence: github.com/ian-nason/qcew-database
• Column dictionary (all 48 columns, source headers, types, NULL shares, join hints): DICTIONARY.md; BLS annual layout
• Catalog: datapond registry; qcew is included in the public catalog and on the datapond website.

Changelog

2026-10-18 — Documentation refresh

• Added Python/R quick starts, table descriptions and column counts, view guidance, navigation, and clearer researcher caveats.
• Database bytes, schema, source locks and validated query results are unchanged. Catalog integration was pending at this documentation checkpoint; it is now complete.

2026-10-18 — Initial release

• QCEW 2025 annual: 3,096,467 rows across 6 tables (48 columns) and the annual view. Artifact 195,571,712 bytes, SHA-256 89bf1cef…edbf8a1d, commit 334c467d.
• Every typed cell verified against an independent read of the pinned source ZIP; disclosure NULLs checked against their own flags; public immutable URL, HTTP range requests and the installed Python and R clients tested after upload.
• Registry entry initially opened as a draft pull request; later merged after integration review.