I selected the U.S. Census Bureau's 2023 Business Dynamics Statistics release through the allowlisted Census provider, then opened the official dataset catalog, program description, codebook, exact CSV files, and public-access policy. The release covers employer businesses from 1978 through 2023 and was reviewed by the Census Bureau for public access and disclosure protection. Census Bureau data and employee-created works generally are not subject to U.S. copyright protection under the agency's public-access policy; this project retains Census attribution.
The reproducible standard-library Python builder downloads two official files. `bds2023_fa.csv` is 86,502 bytes with SHA-256 `52a36e61791aacfaedc0466a4d9b8c1c94a8b0dfbe05e967840119bf6d5695a1` and contains 552 national year-age rows. `bds2023_sec_fa.csv` is 1,463,735 bytes with SHA-256 `07f3b6c3e34add284c1e5b900605eb6d45b4c22420cc4670ee0a9e0d94661755` and contains 10,488 year-sector-age rows. The builder verifies both byte counts, hashes, schemas, row counts, the complete 1978–2023 year sequence, 12 firm-age labels, 19 sector codes, and the published job-flow identity within the source's rounding tolerance.
The displayed period begins in 2003, the first year when the published `26+` age band can appear beside all younger exact-age bands. Exact ages 1 through 5 are added into one readable group; every other published band stays separate. Counts for firms, employment, average employment, jobs created, jobs destroyed, and net jobs are summed within that group. Rates are recalculated as 100 times the grouped count divided by the grouped BDS average-employment denominator rather than added across rows. The source marks job destruction for firm age 0 as unavailable, and the browser snapshot preserves that missing value. Repeated builds produce the same 726,222-byte JSON file with SHA-256 `0212c56dd72ddbe55025c7d61473eae1dd69361a7107973cac848eb7a524d332`. The page imports that file and makes no runtime request.
BDS job flows compare employment in the payroll period containing March 12 with the prior year's period. Job creation is employment gained at opening or expanding establishments; job destruction is employment lost at closing or contracting establishments. Net change is creation minus destruction. Establishments are grouped by the age of their parent firm, so the age can differ from the age of a particular workplace. `Started before 1977` is kept separate because the data system cannot assign those firms an exact birth year.
These are establishment-level business records, not individual worker histories. They do not count hires and layoffs directly, explain why a job appeared or disappeared, measure hours, wages, job quality, vacancies, self-employment, or every kind of work. Sector mix, company ownership, mergers, openings, closures, expansions, contractions, administrative corrections, and classification changes can affect the values. The Census Bureau applies disclosure protection, and recent years can be revised as more information arrives. The comparisons are descriptive and do not isolate causes or predict the next year.