Nason/ussc-database

USSC Federal Sentencing Database

Every federal defendant sentenced from FY2002 through FY2025 (1.72M), from the U.S. Sentencing Commission's public individual offender datafiles: demographics, offense of conviction, guideline calculations, sentence imposed, departures and variances. The Commission's ~22,000-variable fixed-width layout is reshaped into a wide sentences table plus long tables per count of conviction, guideline computation, drug type and departure reason. 13,962,089 rows across 9 tables.

Table	Description	Row Count	Column Count
new_statutes	Statutes of conviction as recorded in NWSTAT slots	3,279,305	4
departure_reasons	One row per stated reason for departure/variance: REAS code + RETEXT text (FY2004+, post-Booker coding) or REASON + REASTXT (FY2002-03, pre-Booker coding)	2,637,999	7
counts	One row per count of conviction: statute (TTSC1/TTSC2/TTSC3 = title/section/subsection), statutory minimum/maximum (SMIN/SMAX) and STA1-3	2,412,926	11
sentences	One row per federal defendant sentenced in the fiscal year: demographics, offense, guideline range, sentence imposed, departures (scalar variables of the USSC individual offender datafile)	1,720,684	1610
guidelines	One row per guideline computation on the case: base offense level, specific offense characteristics and Chapter 3 adjustments (ADJ_*), per-guideline totals	1,706,580	348
drugs	One row per drug type involved: type, weight, units, equivalency (DRUGTYP, WGT, UNIT, MWGT, DRGAM, ...)	749,098	14
pleas	One row per plea/verdict slot (INPLEA, INNOPL)	738,024	5
offense_dates	Offense begin/end dates per slot (OFBEG/OFEND; FY2002-era files only)	501,590	5
chapter_text	Free-text chapter 2/3/4 guideline notes (CHP2TXT/CHP3TXT/CHP4TXT) and change flags	215,883	9

Query it remotely

INSTALL httpfs; LOAD httpfs;
ATTACH 'https://huggingface.co/datasets/Nason/ussc-database/resolve/main/ussc.duckdb' AS ussc (READ_ONLY);

SELECT fiscal_year, COUNT(*) AS defendants,
       ROUND(100.0 * COUNT(*) FILTER (WHERE prison_imposed) / COUNT(*) FILTER (WHERE prison_imposed IS NOT NULL), 1) AS pct_prison_imposed,
       ROUND(AVG(prison_months) FILTER (WHERE term_type = 'months'), 1) AS mean_months_when_term_stated,
       COUNT(*) FILTER (WHERE term_type = 'life') AS life
FROM ussc.v_sentence_terms GROUP BY 1 ORDER BY 1;
-- TOTPRISN alone is neither averageable (9992/9996/9997/9998 are codes) nor a prison
-- indicator (0 = no prison OR under one month); the view separates both.

Or with the datapond packages: pip install datapond / pak::pak("datapond-db/datapond-r").

Build pipeline, the USSC codebook and caveats: https://github.com/ian-nason/ussc-database Source: https://www.ussc.gov/research/datafiles/commission-datafiles (public domain).