Case study · Collaborative data + policy research

Arrest trends, policy reform, and racial disparities.

A three-city analysis asking what changed from 2013–2022, which shifts aligned with major policy events, and why falling totals do not automatically mean disparities are shrinking.

NYC · DC · LAPython + pandasMay 2025Team project

01 · The question

Public debate moves quickly. Structural patterns do not.

Short windows and single-city stories can make ordinary seasonality look like policy impact, or make aggregate declines look like equitable progress. The project compared adult arrest patterns across Washington, DC, Los Angeles, and New York City while mapping selected reforms to the timeline.

Source scope: the charts discussed here compare 2013–2022. The linked repository describes a broader 2000–2024 project window; those dates should not be read as the coverage of every chart.

02 · My contribution and ownership

A collaborative course research project.

Shared work

The paper uses “we” throughout and represents a team’s combined literature review, data analysis, policy selection, charts, and conclusions.

My contribution

I helped frame the research questions, interpret the city and policy charts, and write the policy synthesis alongside my teammates.

03 · Evidence

See the analysis, not only the summary.

The excerpts below are rendered directly from the submitted paper. Open the PDF for complete methods, citations, limitations, and team context.

Stacked bar charts showing adult arrest declines from 2013 to 2022 in Washington DC, Los Angeles, and New York City, annotated with policy events — **Comparing cities over time**Adult arrests declined across all three city series, with different policy events and local contexts attached to each timeline. Source: paper, page 6.

Two line charts comparing felony, misdemeanor, and infraction proportions before and after California Proposition 47 — **Los Angeles and Proposition 47**The paper finds that the observed felony-share pattern does not support a simple “reform worked as expected” story. Source: paper, page 8.

Line charts of normalized Black arrest proportions in Los Angeles, New York, and Washington DC around marijuana legalization — **Equity after legalization**Normalized Black arrest proportions did not move consistently across the three cities after legalization events. Source: paper, page 10.

04 · Analytical approach

Move from totals to context, composition, and limits.

A falling number on its own answers almost nothing. Getting from a headline total to a defensible claim took four steps.

Step 1

Build comparable series

Organize city, date, and charge-type fields so trend movement can be compared without treating unlike categories as identical.

Step 2

Mark policy events

Place selected reforms, including marijuana decriminalization/legalization and Proposition 47, inside the observed timelines.

Step 3

Test expectations

Use proportion trends and a difference-in-differences comparison to examine whether the direction of change matches the policy claim.

Step 4

Keep caveats visible

Separate association from causation and acknowledge different reporting practices, population context, and data-quality limits.

05 · What emerged

The direction is clear. The explanation is not singular.

Four things stayed true across the three cities, and none of them by itself settles the debate a headline number tends to imply.

Totals fell

Adult arrest totals declined across the three displayed city series from 2013–2022.

Policy effects varied

Reforms aligned with changes in some series, but the mechanisms and timing differed by city.

Composition matters

A lower total does not reveal whether the burden across charge types or racial groups changed equitably.

Implementation matters

The observed trends complicate clean before-and-after narratives, particularly around Proposition 47 and legalization.

06 · Limitations

What the analysis cannot establish.

City agencies may define, collect, and revise arrest categories differently.
An arrest is not the same as a crime occurrence, conviction, or measure of public safety.
Policy timing overlaps with other changes, including COVID-era enforcement shifts.
The paper notes missing population context in parts of the analysis.
Associations around policy events do not by themselves establish causal impact.

07 · Artifacts

Read, reproduce, and question the work.

Full paper (PDF) GitHub repository Notebooks & charts