What the assault-victim gap measures

Police record assaults on Black women about 4.2 times as often as on White women. Hospitals see the same gap (4.6 times). Only the victimization survey, where women describe their own assaults, puts it near 1.

The question

Is the 4x gap in police records about assaults, about reporting, or about how police record?

Not reporting: Black women report 54% of assaults, White women 47%.
Not recording: police write up the same share of victims per 911 call in every neighborhood in New York, Los Angeles, New Orleans; a little more in Baltimore.
Not repeat counting: about 1.1 reports per woman in both groups in Los Angeles.
Hospitals agree with police: 4.6 times, with no call involved. What is left open is why the survey sees so little.

At a glance

Police records

4.2x

51 cities

Emergency departments

4.6x

women, 2021 to 2022

Victimization survey

1x in large places

Recording adds nothing

3 of 4

cities

Three sources

Black women's assault rate over White women's. A bar ending at 1 would mean no gap.

Victims recorded per 911 call

Flat = police record the same share everywhere.

Calls per resident

Rising = the gap is already in the calls.

Testing explanations

By who did it

The police gap is larger in every category.

Every explanation tested

Could the police gap be...	Tested in	Answer
Black women report assaults more often	Survey	A little: 54% against 47%; lifts the survey ratio to about 1.4, not 4
Survey counts threats as assaults	Survey, injury only	No: with an injury required the ratio is 1 nationally
One assault becomes several records	DC	No: 10.4x to 10.7x however records are counted
The same women reported again and again	Los Angeles	No: 1.11 and 1.10 reports per woman; ratio 5.7 to 5.65
Police write up more victims per call in Black neighborhoods	New York, Los Angeles, New Orleans, Baltimore	Flat in New York, Los Angeles, New Orleans; up to 1.2x in Baltimore
Public housing and shelters	New York	Partly: 17% of recorded assaults on Black women are in public housing against 3% of those on White women; outside it the ratio is 4.5x, not 5.2x
Assaults recorded where they happen, residents counted where they live	New York	Large: within the Blackest precincts the gap is 2.7x, citywide 5.2x
Hospitals see what police see	NHAMCS	Yes: 4.6x for women, 4.3x for men

The cities

City	Victims per call, whitest to Blackest fifth	Recording effect	Calls per resident, Blackest over whitest
New York 77 precincts	0.622 to 0.535	1.07 (SE 0.065)	2.77x
Los Angeles 680 districts	0.887 to 0.844	1.09 (SE 0.107)	2.1x
New Orleans 17 zips	0.988 to 1.065	1.00 (SE 0.119)	1.78x
Baltimore 133 neighborhoods	0.371 to 0.404	1.22 (SE 0.07)	2.32x

Emergency departments, 2021 to 2022

Per 100,000 adults a year; 60 Black and 62 White women's visits sampled.

	Black	White	Hispanic	Black vs White, 95% interval
Women	1,654	357	337	4.63 (3.25 to 6.61)
Men	1,857	436	728	4.26 (3.01 to 6.04)

Method

Survey: BJS NCVS, women 18 and older, 2015 to 2024. Hospitals: NHAMCS 2021 and 2022. Police: the 59-city run. Calls: each city's open data by neighborhood. Every number is generated from out/*.json; the README holds the full tables.

Limits

Small survey cells. 919 Black women assault victims in the survey over ten years; the hospital figure rests on 60 sampled visits. Intervals are approximate.
Definitions differ. Survey simple assault includes threats; police and hospital counts do not. The injury-only checks address this.
Place is not residence. Assaults are recorded where they happen and residents counted where they live, so the within-neighborhood ratios are rough too.
Race recorded three ways. By officers in police data, by respondents in the survey, by hospital staff in NHAMCS. The race-coding bound from the national run applies to the police side.
What is left open. Why the survey sees so little: non-response, coverage of shelters and institutions, or what women say at home. And who makes the calls. Neither can be settled with public data.
What, not why. Nothing here measures causes, offenders or circumstances.