robworks-software/us-k12-schools-directory

US K-12 Schools Directory

A directory of 124,613 US K-12 schools covering all 50 states, DC, and US territories, compiled from federal and state government sources. Each record carries directory information (address, phone, website), enrollment and demographics, and, where a source supplied it, a principal name and email.

This is a compilation of public government data. It is not a survey, and no field was independently verified against the school itself.

Loading

from datasets import load_dataset

ds = load_dataset("robworks-software/us-k12-schools-directory")

texas = ds["train"].filter(lambda x: x["state"] == "TX")
with_principal = ds["train"].filter(lambda x: x["principal_name"] is not None)

The pre-split train/validation/test configuration is a random 70/15/15 partition of the same rows; full.parquet in the repo holds all 124,613 records unsplit. The splits carry no semantic meaning - use full.parquet for directory lookups.

Splits

Split	Rows
train	87,229
validation	18,692
test	18,692
total	124,613

Field coverage

Coverage is uneven, and the gaps are large enough to matter for most analyses:

Field	Populated	Coverage
Phone	124,190	99.7%
Performance data	123,624	99.2%
Demographics	119,209	95.7%
Website	71,117	57.1%
Principal name	63,161	50.7%
Principal email	36,650	29.4%

Sources

• NCES Common Core of Data (CCD) - public school directory, enrollment, demographics
• NCES Private School Universe Survey (PSS) - private school directory
• Civil Rights Data Collection (CRDC) - enrollment demographics
• State education agency directories (22 states) - principal names, emails, websites
• State report cards - graduation rates, accountability ratings
• Urban Institute Education Data API - demographics enrichment
• School website scraping (~28,000 sites) - principal names and emails extracted automatically from school web pages

Limitations

• Principal contacts are half-covered and unevenly distributed. The 50.7% name / 29.4% email coverage is concentrated in the 22 states with a usable state directory. Schools in the other 28 states are much more likely to have no principal at all.
• Scraped contacts are unverified. Roughly 5,700 principal names and 1,300 emails came from automated extraction of school websites, at about a 20% hit rate. These were not checked by hand and some will be wrong, stale, or attached to the wrong person.
• Staff turnover makes contacts decay quickly. Principal assignments change every year; treat any name or email here as a lead, not a fact.
• Mixed vintages. Public school data is 2022-2023; private school data is 2021-2022 (PSS is biennial); CRDC lags 1-2 years; state report cards vary by state.
• State-constructed IDs required remapping. Several state directories build NCES IDs that do not match federal CCD IDs. Those records were matched back by school name and city, which is imperfect - match rates ranged from about 69% to 95% by state.
• Ratings are sparse and third-party. greatschools_rating and niche_rating come from commercial sources with limited coverage and their own methodologies.
• No maintenance schedule. This is a snapshot. It is not updated on a fixed cadence.

License

CC0-1.0 (Public Domain Dedication). The underlying data comes from US federal and state government sources and is public domain under 17 U.S.C. Sec. 105 and state public records laws.

Attribution is appreciated but not required: NCES, US Department of Education Office for Civil Rights, the state education agencies listed above, and the Urban Institute Education Data Portal.

Citation

@dataset{us_k12_schools,
  title  = {US K-12 Schools Directory},
  author = {Robworks Software},
  year   = {2026},
  publisher = {Hugging Face},
  url    = {https://huggingface.co/datasets/robworks-software/us-k12-schools-directory}
}