Four cities and national surveys · 2015 to 2025

Police and hospital records agree on the gap in assaults on Black women. The national survey does not.

Police record about 4 times as many assaults on Black women as on White women, per resident, and emergency rooms see about the same. The victimization survey puts it near 1. This report tests why.

Every city report on this site finds the same thing in police records: per resident, far more assaults on Black women are recorded than on White or Hispanic women. Across the 51 large cities whose police record ethnicity, Black women's reported rate is 4.2 times White women's and 2.1 times Hispanic women's.

The National Crime Victimization Survey shows something different. It asks a national sample of households about assaults whether or not anyone told the police. In the survey, Black women's assault rate is about the same as White women's and a little above Hispanic women's. This report tests why the two sources disagree.

Hospitals agree with the police

Emergency department records don't depend on anyone calling the police. In the national hospital survey for 2021 and 2022, adult Black women's rate of emergency visits for assault was 4.6 times White women's. The sample is small, 60 visits by Black women and 62 by White women, so the figure is rough. But it is the size of the police gap, not the survey's. Two independent systems agree on 4 to 5 times; the household survey is the outlier.

What was tested

• Reporting. Black women report a somewhat larger share of their assaults to police than White women do (54% against 47%). That lifts the survey's ratio for assaults reported to police to about 1.1 to 1.4, far short of 4.
• Threats. The survey's "simple assault" includes verbal threats that police record differently. Counting only assaults with an injury, the survey ratio is 1.0 nationally and 1.75 in large places, while police records show about 4.
• Double counting. In DC, which has the largest gap, counting one record per incident instead of one per victim keeps the ratio between 10.4 and 10.7 times.
• The same women reported again and again. In Los Angeles, Black and White women each average about 1.1 reports over four years. Counting women instead of reports moves the ratio from 5.7 to 5.65 times.
• How police write up calls. In New York, Los Angeles and New Orleans, police record about the same number of victims per 911 assault call whatever a neighborhood's racial makeup. In Baltimore, wholly Black neighborhoods record somewhat more victims per call, about 1.2 times.

Assault calls per resident are 1.8 to 2.8 times higher in the Blackest areas than in the whitest. The gap is already there when the phone rings.

Where assaults happen

Police record assaults where they happen; rates divide by where women live. In New York, Black women's recorded rate is highest in Manhattan's whitest precincts, where few Black women live. Within the Blackest precincts, where Black and White women are neighbors, the gap is 2.9 times, against 5.2 times citywide.

What is left

What remains is a gap of about 2 to 3 times between the calls that reach police from the places Black women live and what Black women tell the survey. Two explanations are left, and public data can't settle either:

• The survey may miss assaults on Black women, through who answers, women living outside households, or partner violence not described at home. Each Black woman who answers stands in for 1.25 times as many women as each White woman who answers.
• More assaults on Black women may reach police through other people: neighbors and bystanders calling 911, building staff, hospitals.

The emergency department figures point toward the first. Settling it would take the survey's restricted files, held in federal research data centers, and 911 records that say who made each call.

What this does not show

• Why. This compares what each source counts. It does not say why Black women are assaulted, or why anyone calls the police.
• Precise survey or hospital figures. The survey holds 919 Black women victims over 2015 to 2024, and the hospital figure rests on 60 visits, so both carry wide intervals.
• One definition of race. Race is recorded by officers in police data; respondents report their own in the survey. The two need not agree.
• People. Police records count reports, not people: someone assaulted twice counts twice. The survey caps one person's repeated assaults at 10.

Part of the series Assault victims by race and sex, 6 reports.