# CRDC School Arrest Rate API (v1)

Base URL: https://crdc-api.civilytics.org/api/v1
Bayesian small-area estimates of school-based arrest rates from the US DoE Civil
Rights Data Collection. Cite: Knowles & Miller 2025.

Response envelope: { status, data, error, meta }.
Enums: race=AM|BL|HI|WH; sex=F|M; year=15-16|17-18|21-22;
model=unified_m1..unified_m5|stratified_m1..stratified_m5 (default unified_m2; request a stratified_* model explicitly for per-student-group fits); interval=50|80|95 (default 95).

Endpoints:
  GET /estimates?leaid=&state=&race=&sex=&year=&model=&interval=&limit=&page=
  GET /estimates/{leaid}
  GET /states?state=&race=&sex=&year=&model=&interval=
  GET /states/{state}
  GET /districts?q=&state=    (name/geo lookup -> leaid)
  GET /models
  GET /draws?...              (returns HF Parquet shard URL + DuckDB SQL; bulk only)

Example (R):
  library(httr2)
  resp <- request("https://crdc-api.civilytics.org/api/v1/estimates") |>
    req_url_query(state="TX", race="BL", sex="M", model="unified_m2") |>
    req_perform() |> resp_body_json()

Example (Python):
  import requests
  r = requests.get("https://crdc-api.civilytics.org/api/v1/estimates",
                   params={"state":"TX","race":"BL","sex":"M","model":"unified_m2"}).json()

Raw draws (bulk, DuckDB over Hugging Face):
  INSTALL httpfs; LOAD httpfs;
  SELECT * FROM read_parquet(
    'https://huggingface.co/datasets/civilytics/crdc-school-arrest-rates/resolve/main/parquet/model_id=unified_m2_mod/YEAR=21-22/LEA_STATE=TX/*.parquet',
    hive_partitioning=true);