Louisiana DPSC Data

This repository loads, combines, and cleans public datasets released by the Louisiana Department of Public Safety & Corrections (DPSC). It pulls raw files directly from the DOC website, connects categorical codes using the official codebook, and produces analysis-ready dataframes across six data types.

---

Overview

As of May, the Louisiana DPSC publishes several datasets on incarceration, supervision, and mortality. This pipeline automates the full process of:

1. Downloading raw CSVs and Excel files directly from doc.la.gov.
2. Combining files across time periods into unified dataframes.
3. Decoding coded fields (race, offense, parish, facility, supervision type) via the official codebook.
4. Cleaning and standardizing dates, sentence lengths, flags, and demographics.
5. Saving both raw and processed outputs as .rds files.

---

Data Sources

All files are downloaded directly from the Louisiana DPSC website. The pipeline currently covers:

Dataset	Period Covered	Description
Incarceration Snapshots	Dec 2025 – Mar 2026	Monthly point-in-time views of the incarcerated population
Admissions	2019 - 2025	All prison admissions by year
Releases	2016 – 2025	All prison releases by year
Probation & Parole Snapshot	Feb 2026	Point-in-time supervised population
Deaths	2018 - 2025	In-custody deaths with cause
Facility Census	Jan 2025 – Mar 2026	Monthly headcounts by facility and custody type
Codebook	Latest	Official lookup tables for all coded fields

---

Output Files

Raw (raw/)

Unmodified data as downloaded, stored as named lists of dataframes:

• snapshot.rds
• admit.rds
• release.rds
• pp.rds
• death.rds
• census.rds

Processed (processed/)

Cleaned, decoded, and analysis-ready dataframes:

File	Key Variables
snapshots_clean.rds	Location, race, sex, age, sentence length, offense, flags
admits_clean.rds	Admission date, parish, race, sex, sentence type, time-to-release estimates
release_clean.rds	Admission/release dates, years served, offense, release reason
pp_clean.rds	Supervision type, admission/end dates, offense, flags
deaths_clean.rds	Death date, cause, race, gender, years until death
census_clean.rds	Prison, custody type, facility, monthly population counts

---

Key Cleaning Notes

• Dates are parsed from multiple formats (YYYYMMDD, YYYY-MM-DD, MM/DD/YYYY) via a unified parse_mixed_date() function.
• Sentence lengths are decoded from a packed 7-digit format (YYYMMdd -> years + months/12 + days/365.25).
• Flags (sex_offender, violent, habitual) are standardized from "X" markers to logical TRUE/FALSE.
• Offense, race, parish, and location codes are replaced with human-readable literals via codebook join.
• Census files are cleaned into a tidy facility-level format.