Washington, DC · 2019 to 2025
DC police search Black people they stop far more often. Those searches find something at about the same rate.
Pat-downs of Black people who were stopped run 1.7 to 4.5 times the rate for others, depending on the district. A statistical model puts the bar for searching Black people lower.
Being stopped is one step; being searched is the next. From 2023 through 2025, in every DC police district, Black people who were stopped were patted down more often than other people who were stopped. The ratio ran from 1.65 times in the 4th District to 4.54 times in the 5th. These rates are per person stopped, so they don't depend on how many people police stop in each district.
Protective pat-downs, the search officers use most at their own discretion. Source: MPD stop records via DC Open Data.
What the searches find
If officers searched everyone on the same evidence, searches of each group should find contraband at about the same rate, and they do. Across consent searches, warrant searches and pat-downs with a recorded outcome, 13.9% of searches of Black people and 12.5% of searches of everyone else found something. The difference isn't statistically significant. The gap is in who gets searched, not in what the searches find.
13,406 searches of Black people and 1,081 of everyone else. Probable-cause searches have no outcome field in MPD's data and can't be included. Source: MPD stop records via DC Open Data.
A lower bar to search
Equal hit rates can still hide a lower bar for searching one group, because they average over everyone searched, not just the people officers were barely willing to search. A model built for this problem (Simoiu, Corbett-Davies and Goel, 2017) estimates the bar itself. It puts the evidence threshold for a search at about 0.05 for Black people and 0.08 for everyone else, a gap of 0.04 (95% interval 0.02 to 0.06), and the bar is lower for Black people in every district. In plain terms, searches of Black people go ahead on thinner suspicion.
That result depends on the model's assumptions. The hit-rate comparison above makes none, and the two are reported side by side for that reason.
What this does not show
- Why. The records show who was searched and what was found. The data does not say why: it holds no officer's reasons.
- People. Records count stops and searches, not people: someone stopped twice counts twice.
- Every search. Probable-cause searches carry no outcome in MPD's data (4,342 of them in 2023 to 2025), so they are left out of the hit rates and the model.
- Small groups. Searches of everyone else number 1,081, so their hit rate is less precise than the rate for Black people. "Everyone else" combines every group other than Black.
- Exact group membership. Race is recorded by officers.
Whether this enforcement follows where crime is reported is the subject of the next report.
Part of the series Policing in DC, 5 reports.