Enforcement vs reported crime

Published July 2026 · Data window: MPD incidents vs stops (2023–2025) and arrests (2013–2025)

The strongest defense of enforcement geography is that police go where crime is. This page tests it against MPD's own incident data, which publishes only victim-initiated offenses, so police presence cannot inflate the denominator. Measured against the crime residents actually report: frisks in majority-Black districts exceed the crime baseline under BOTH denominators, all victim crime and violent crime alone; stops, arrests depend on the denominator: above the baseline against all victim crime, at or below it against violent crime, which concentrates east of the river.

2.72×

Frisks per reported victim crime in majority-Black districts relative to the rest of DC, the measure that stays elevated under every denominator this page can construct.

Each district shaded by its frisk skew: its share of pat-down searches divided by its share of reported victim crime. Click a district for its full page.

The sensitivity matrix

Majority-Black districts (5, 6, 7) versus the rest, enforcement per reported crime, under both denominators. Showing both is the point: a result that depends on the denominator is reported as exactly that.

Measure	vs all victim crime	vs violent crime only
Stops (2023–2025)	1.44×	0.75×
Frisks (2023–2025)	2.72×	1.41×
Arrests (2013–2025)	2.01×	0.91×

And what those frisks find

Frisks in majority-Black districts found contraband slightly MORE often (14.9% vs 12.7%). The volume skew is not a story of fruitless searches; what it measures is how much more of this intrusion residents of those districts absorb per reported crime. Who gets frisked once police are anywhere is the separate question, answered on the searches page, where hit rates by race are statistically indistinguishable.

Frisks, 2023–2025	With recorded outcome	Something seized	Hit rate
Majority-Black districts (5, 6, 7)	7,685	1,143	14.9%
Rest of DC	4,829	612	12.7%

Same significance gate as the outcome test (two-proportion z-test, z = 3.45); a direction is only ever asserted when it holds.

Districts are listed in numeric order. This is not a ranking: each skew compares a district's enforcement share to its own reported-crime share.

District	Black pop	Victim-crime share	Stop share	Stop skew	Frisk skew
District 1	25.0%	13.5%	14.1%	1.04×	1.04×
District 2	8.0%	17.7%	13.1%	0.74×	0.31×
District 3	22.0%	18.6%	15.6%	0.84×	0.56×
District 4	48.6%	13.3%	11.6%	0.87×	0.65×
District 5	56.7%	16.6%	18.0%	1.08×	0.94×
District 6	88.4%	12.2%	14.0%	1.15×	1.73×
District 7	86.5%	8.0%	13.7%	1.71×	3.07×

Stops and frisks 2023–2025 against victim crime in the same window.

Arrests, thirteen years

District	Victim-crime share	Arrest share	Arrest skew
District 1	14.3%	15.3%	1.07×
District 2	17.8%	7.9%	0.45×
District 3	18.6%	12.7%	0.68×
District 4	13.0%	10.6%	0.81×
District 5	14.9%	16.6%	1.11×
District 6	12.9%	18.9%	1.46×
District 7	8.6%	18.1%	2.11×

How this number is built (and where it's soft)

The denominator cannot be police-made. DC's incident feed publishes only the nine Part I victim/property offenses (theft, theft from auto, motor vehicle theft, robbery, burglary, assault with a dangerous weapon, sex abuse, homicide, arson). Possession, disorderly, loitering, and warrant offenses are not in the dataset at all, so the circularity that plagues crime-baseline studies is structurally excluded.
Two denominators, on purpose. Theft dominates all victim crime and concentrates in commercial corridors; violence concentrates east of the river. Results that flip between the two are reported as denominator-sensitive, not asserted either way.
Reported crime is not occurred crime. Reporting rates vary by neighborhood and offense. If under-reporting is heavier in some districts, their true crime is understated and their skew overstated, and the reverse; this page cannot resolve that, and says so.
Volume, not selection. This page measures where enforcement lands relative to crime. It says nothing about who is stopped once police are somewhere; that is the representation index, the outcome test, and the veil test, which are unaffected by any denominator choice here.
Windows matched. Stops and frisks (2023–2025) against crimes 2023–2025; arrests (2013–2025) against crimes 2013–2025.
Exclusions (reported, not dropped quietly). Rows with no mappable district: crimes 1,043, stops 3,057, arrests 2,256.
Future refinement. 911 calls-for-service as a third denominator, the demand signal MPD itself cites for deployment.