Skip to content
AS / SIGNALS
Back to work

Policy Research

Brooklyn Fare-Evasion Arrests — Microdata Cleaning & Description

Harmonized 2016 Brooklyn Defender Services and Legal Aid Society fare-evasion arrest files, recoded race/ethnicity, and mapped enforcement concentration across subway stations.

Role: Analyst (Columbia SIPA DSPC 7514)

Situation

Public-defender administrative records are a rare window into subway fare-evasion enforcement, but BDS and LAS code race and ethnicity differently and cover only their own clients.

Decision question

How do you make two public-defender arrest files comparable when race and ethnicity are coded differently and coverage is incomplete?

Data and constraints

BDS and LAS 2016 client files only—not all Brooklyn fare-evasion arrests; dismissal observed only in LAS; missing race/ethnicity excluded from some denominators.

Method

Inspected two microdata files (2,246 BDS and 1,965 LAS records), recoded race/ethnicity into a shared mutually exclusive factor, documented coverage and missingness limitations, and tabulated arrest volume and racial composition at high-volume stations.

Tools

  • R
  • tidyverse
  • fastDummies
  • Administrative microdata cleaning

Deliverables

  • Knitted R Markdown analysis (February 2026)
  • Harmonized race/ethnicity coding and station-level descriptive tables

Outcome / decision value

  • In the BDS file, Non-Hispanic Black clients were the largest recoded group (1,359), followed by Hispanic (493) and Non-Hispanic White (165).
  • The sample represents BDS/LAS clients, not all Brooklyn fare-evasion arrests; dismissal is observed only in LAS data.
  • High-volume stations (Coney Island–Stillwell, Jay St–MetroTech, Utica Av, and others) show distinct racial compositions of arrestees.

Lessons learned

  • Harmonized coding is a decision, and it belongs in the write-up next to the counts.
  • Coverage limits are as important as the demographic tables that follow them.

Links

GitHub URL — add in content-data.tsLive demo URL — add in content-data.tsAnalysis PDF