Body Cameras and Police Stop Disparities

ISEF Category: Behavioral and Social Sciences

Subcategory: Sociology and Anthropology  ·  Difficulty: Advanced  ·  Setup: Home Setup  ·  Time: 1 to 2 Months

The Hook

A camera on a uniform can change more than what people see. It can change how often a stop ends in a search, a warning, or an arrest. Your project asks a real question with public data, did body-worn cameras line up with measurable changes in stop-outcome gaps?

What Is It?

Body-worn cameras record police and civilian interactions during stops. Cities rolled them out at different times, which gives you a natural comparison. If one city adds cameras in year one and another city adds them in year three, you can compare how each city changed over time.

Difference-in-differences is the tool that makes that comparison useful. Think of it like measuring two plants, one gets sunlight later than the other. You do not just ask whether the treated plant grew. You ask whether it grew more than the other plant grew over the same stretch of time. In this project, the camera rollout is the treatment, and stop outcomes are the growth you measure.

Why This Is a Good Topic

This is a strong science fair topic because the data are public, the question is clear, and the answer can be tested with careful stats instead of guesswork. You also connect a real policy change to outcomes that matter in daily life, such as searches, arrests, citations, and warnings. You can learn data cleaning, causal inference, and how to check whether a result is real or just a timing coincidence.

Research Questions

• How does the introduction of body-worn cameras change search rates during traffic stops in rollout cities?
• What is the effect of body-worn camera rollout on arrest, citation, and warning rates after a stop?
• Does the gap in stop outcomes between Black drivers and White drivers change after cameras arrive?
• To what extent do treated cities show different pre-trend patterns than comparison cities before rollout?
• Which stop outcomes move the most after camera rollout, search, arrest, citation, or warning?
• How does the effect vary across cities with high, medium, and low stop volumes?

Basic Materials

• Laptop with internet access and enough storage for downloaded data.
• Stanford Open Policing data files and city metadata.
• A spreadsheet app or notebook app for tracking city rollout dates.
• Python or R for cleaning data and running regression models.
• A simple codebook or notes file for recording how you define each outcome.

Advanced Materials

• R with panel-data and causal-inference packages.
• Python with pandas, statsmodels, and plotting libraries.
• A local copy of the stop-level data, city rollout dates, and demographic controls.
• GIS software for checking city location patterns and regional balance.
• Access to a statistics adviser who can review fixed-effects and event-study choices.

Software & Tools

• Python: Cleans stop-level records, runs regression models, and draws coefficient plots.
• R: Fits fixed-effects and event-study models with social science packages.
• Jupyter Notebook: Keeps code, notes, and figures together while you iterate.
• OpenRefine: Fixes inconsistent city names, dates, and labels before analysis.
• Google Sheets: Helps you inspect small tables and spot merge errors early.

Experiment Steps

1. Define the treated cities, comparison cities, and rollout dates for each policy change.
2. Choose the stop outcomes you will measure, such as search, arrest, citation, and warning rates.
3. Build a before-and-after timeline and check whether treated and comparison cities move together before rollout.
4. Pick control variables and fixed effects so city size, season, and time trends do not stand in for policy effects.
5. Plan a sensitivity check that tests whether your result survives different comparison groups, date windows, or outcome definitions.

Common Pitfalls

• Using the announcement date instead of the rollout date, which puts stops in the wrong treatment window.
• Mixing cities with different stop coding rules, which makes outcome rates hard to compare.
• Treating raw stop counts as the main outcome when stop volume changes can hide the real pattern.
• Ignoring pre-trend differences, which can make an old gap look like a camera effect.
• Combining searches, arrests, citations, and warnings into one score, which can hide opposite movements across outcomes.

What Makes This Competitive

A strong version does more than compare before and after. You separate stop outcome types, test for pre-trends, and show that your result holds across more than one comparison group. You can also ask whether the effect changes by city size, region, or demographic makeup. That kind of careful design turns public data into a real causal argument.

Project Variations

• Use only cities with clear rollout announcements and compare stop outcomes before and after each one.
• Split the sample by city size and test whether large cities and small cities respond in the same way.
• Compare search rates with warning rates to see whether cameras change enforcement style more than stop volume.

Learn More

• Stanford Open Policing Project: Download stop-level records, codebooks, and city metadata from the project site.
• MIT OpenCourseWare, 14.382 Econometrics: Find free lectures on panel data, fixed effects, and treatment effects.
• UCLA IDRE: Search its free guides for difference-in-differences, regression, and robust standard errors.
• U.S. Census Bureau, American Community Survey: Pull city demographics and controls from the Census data pages.
• Journal of Empirical Legal Studies: Search for peer-reviewed articles on body-worn cameras and policing outcomes.