Race and ethnicity
The same word means different things in different records. This page shows who records race, where the codes changed, and how CrimeDB shows the result.
What we show
Figures on CrimeDB use six groups: White+MENAMiddle Eastern or North African, Black, Hispanic, Asian, AIANAmerican Indian or Alaska Native, and NHPINative Hawaiian or Other Pacific Islander, plus race unknown. Hispanic is of any race, and the other five leave out Hispanic people, so nobody is in two groups. That is the form the 2024 federal standard uses, so the labels don't repeat "not Hispanic". A source that joins Asian and Pacific Islander in one code shows Asian+NHPINative Hawaiian or Other Pacific Islander; a source with no ethnicity field keeps everyone in their race and says so.
We write Hispanic, not "Hispanic or Latino". The federal standard added "or Latino" in 1997 because people in the West tended to use it and people in the East tended to use Hispanic. Both name the same group, and every source we hold counts them the same way.
The group needs both fields: a race and a Hispanic origin. Police often record the race and leave Hispanic origin blank. CrimeDB splits those people into Hispanic or not, and every bar shows the split part lighter, so what police recorded and what CrimeDB filled stay apart.
Why White+MENA, Hispanic reads high
Police record race and Hispanic origin in two fields, and the FBI's race tables use race alone. Police code almost every Hispanic person White: 93% of Hispanic victims in the 34 states that record ethnicity well. So "White" in a race table is mostly two groups.
A rate by race alone makes it worse. It divides by the Census count of White people, and the Census counts only 80% of Hispanic people White alone. The top of the rate gets nearly every Hispanic victim, but the bottom gets only some of them.
Dropping people with a race but no ethnicity has the opposite problem. Most of them are coded White, so White+MENA would read low. That is why CrimeDB splits them rather than leaving them out. In 2025, 125,675 victims of violent crime were coded White with no ethnicity.
FBI NIBRS incident records. MENA means Middle Eastern or North African. Records with no race are left out.
How CrimeDB shows race
One way everywhere: Hispanic of any race is one group, and each race leaves Hispanic people out. Where police recorded a race and no ethnicity, CrimeDB splits those people into Hispanic or not, by stated rules or a statistical model, whichever proved closer.
| Group | Count | Split | Share | Rate |
|---|---|---|---|---|
| White+MENAMiddle Eastern or North African | 351,064 | 83,686 | 34% | 209 |
| Black | 342,864 | 80,892 | 34% | 935 |
| Hispanic | 249,752 | 43,826 | 25% | 407 |
| Asian | 27,226 | 4,840 | 3% | 137 |
| AIANAmerican Indian or Alaska Native | 13,249 | 4,488 | 1% | 643 |
| NHPINative Hawaiian or Other Pacific Islander | 2,728 | 515 | <1% | 444 |
| Race unknown | 33,828 | – | 3% | – |
| Group | Count | Split | Share | Rate |
|---|---|---|---|---|
| Black | 402,349 | 129,256 | 41% | 1,097 |
| White+MENAMiddle Eastern or North African | 243,388 | 76,822 | 25% | 145 |
| Hispanic | 217,245 | 50,103 | 22% | 354 |
| Asian | 13,376 | 2,773 | 1% | 67 |
| AIANAmerican Indian or Alaska Native | 11,877 | 4,490 | 1% | 576 |
| NHPINative Hawaiian or Other Pacific Islander | 2,966 | 642 | <1% | 483 |
| Race unknown | 84,259 | – | 9% | – |
Hispanic is of any race; the other groups are not Hispanic. White+MENA includes Middle Eastern and North African people.
FBI NIBRS incident records and CrimeDB's split of records with a race and no ethnicity. Split: the part of each group CrimeDB placed. Rate: per 100,000 people in the group, the Census group scaled to the share of people the reporting agencies police; – means no group to divide by.
Police coding against the Census
States where police record ethnicity for 80% or more of violent crime victims with a race, 2025. Police: the share of Hispanic victims coded White. Census: the share of Hispanic people counted White alone.
| State | Hispanic victims | Police: White | Census: White alone |
|---|---|---|---|
| TX | 47,448 | 97% | 89% |
| NV | 3,457 | 97% | 78% |
| HI | 127 | 57% | 36% |
| AZ | 5,485 | 97% | 84% |
| FL | 9,049 | 94% | 87% |
| NJ | 5,101 | 93% | 72% |
| NY | 21,695 | 77% | 62% |
| CO | 7,809 | 98% | 81% |
| IL | 5,709 | 94% | 79% |
| CT | 1,331 | 95% | 72% |
| RI | 480 | 78% | 65% |
| GA | 2,402 | 94% | 74% |
| MA | 3,998 | 84% | 69% |
| DE | 393 | 91% | 71% |
| VA | 2,206 | 96% | 73% |
| KS | 2,411 | 97% | 83% |
| NE | 746 | 98% | 78% |
| LA | 513 | 94% | 74% |
| AR | 907 | 97% | 79% |
| WY | 129 | 100% | 79% |
| PA | 3,031 | 89% | 67% |
| SC | 1,281 | 95% | 75% |
| IN | 1,471 | 95% | 80% |
| TN | 3,546 | 97% | 78% |
| WI | 1,642 | 96% | 80% |
| AL | 756 | 89% | 74% |
| IA | 620 | 97% | 82% |
| MS | 102 | 94% | 71% |
| SD | 152 | 74% | 69% |
| MO | 1,593 | 95% | 78% |
| KY | 645 | 96% | 79% |
| ND | 142 | 92% | 74% |
| MT | 184 | 89% | 73% |
| NH | 110 | 91% | 78% |
Timeline
- 1790 to 1950The census taker decidesEnumerators recorded race by sight. A person of mixed White and other parentage went to the other race. Before 1940 everyone was put in a named race, so there was no Other.Census
- 1930Mexican is a race for one censusBefore 1930 the census counted people of Mexican descent as White. In 1940 it moved them back into White. In Arizona, which count you use moves a 1930 White rate by 43%.Census, lynching records
- 1960 to 1979The FBI's own listWhite, Negro, Indian, Chinese, Japanese, and Other, with no ethnicity field.FBI arrests from 1974, homicide reports from 1976
- 1977The first federal standardFour races (White, Black, American Indian or Alaskan Native, Asian or Pacific Islander), with Hispanic origin asked as a separate question. White covers people from Europe, the Middle East, and North Africa.Every federal source
- 1980The FBI switchesArrest and homicide records move to the four races and add Hispanic origin. The arrest file holds ethnicity counts only through 1986, then leaves the slots empty until 2017.FBI arrests, homicide reports
- 1981 and 1990Hispanic origin reaches population countsCensus estimates by Hispanic origin start in 1981 for states and 1990 for counties. Any rate by Hispanic origin before then has no denominator.Census estimates
- 1991Incident records startOn the siteNIBRS records the race and ethnicity of each victim and arrestee. Ethnicity is optional, and up to two thirds is unknown in the first years.NIBRS
- 1993The survey asks people themselvesOn the siteThe victimization survey records race and Hispanic origin as people report them, including crime never reported to police.NCVS
- 1997The second federal standardNative Hawaiian or Other Pacific Islander splits from Asian, and the ethnicity question becomes Hispanic or Latino. Agencies had until 2003. The FBI changed in 2013.Every federal source
- 2002The earliest sentencing file we holdFrom the first file, FY2002, the Sentencing Commission puts every Hispanic person in Hispanic in its own combined category. By race alone, most people it codes White are Hispanic: 55% in FY2002, 74% in FY2025.U.S. Sentencing Commission
- 2013The FBI adopts the 1997 standardOn the siteAsian and Pacific Islander split. Offender ethnicity starts, blank for 99.8% of known offenders in its first year.NIBRS, hate crime
- 2017Homicide and arrest files rebuiltHomicide reports split Asian from Pacific Islander only now, four years after the FBI's switch, and arrest records carry ethnicity again. The arrest file never gets a Pacific Islander code.FBI homicide reports, arrests
- 2021NIBRS onlyOn the siteThe FBI stops taking summary counts. The share of records missing ethnicity doesn't move.NIBRS
- 2024The third federal standardOne combined question, with Hispanic and Middle Eastern or North African as groups beside the races. Agencies must comply by September 2029. We found no FBI plan for it.Every federal source
What to keep in mind
- NCVS and USSC race is self-reported.People answer the survey for themselves, and federal defendants answer the probation officer. Police race isn't.
- Police race is the officer's call, often by sight.No FBI manual says how an agency decides an arrestee's race.
- Offender race is often what a victim or witness described.An offender nobody arrested has the race someone saw, or none.
- Finer categories don't mean the code is more accurate.More boxes give more detail, not a better match to the person.
- Middle Eastern and North African people are counted White.Every standard the FBI and the Sentencing Commission have used puts them in White, so no record we hold has a count of them. That is why we name the group White+MENA. CrimeDB builds no separate MENA count until both the Census and police records report one.
- Police code almost every Hispanic person White.In the 34 states that record ethnicity well, police coded 93% of Hispanic victims White by race. The Census counts 80% of Hispanic people White alone.
- A missing ethnicity mostly has a White race.Of people with a race but no ethnicity, police coded about two thirds of victims and more than half of offenders White. Some of them are Hispanic, and nobody knows how many.
- Missing ethnicity is agency practice, not chance.In 2025, 2,718 agencies never recorded an offender's ethnicity. Six states record almost none for offenders: Delaware, Kansas, West Virginia, Oklahoma, Vermont, and Kentucky.
- Los Angeles records no race for its victims.In 2024 and 2025 the Los Angeles Police Department gave no race for any violent crime victim. Any race figure that takes in Los Angeles is a floor.
- More Hispanic offenders partly means more recording.Offender ethnicity was blank for 99.8% in 2013 and 38% in 2025. A rise in Hispanic offenders over those years is partly agencies starting to record it.
How we handle it
- Codes stay as each source recorded them, unknowns included.
- Each source has one series transformer: for each code edition, how its race and ethnicity codes map to our groups. A code that holds several, like Asian+NHPI before 2013, is never split, and its label names each group it holds.
- Where a source records ethnicity, Hispanic of any race is one group and each race leaves Hispanic people out. Where police recorded a race and no ethnicity, CrimeDB splits those people by stated rules or a statistical model, whichever missed less when whole agencies' records were hidden, and marks the split part.
- Where a source records no ethnicity with race (summary arrests before 2021, NCIC), each race holds its Hispanic people and says so. From 2021, arrests by race are split by NIBRS arrestees' recorded shares.
- In a two-way table, each state's split for a race, offense, and role applies to every age, sex, and relationship alike, and where both people in a pair have a race and no ethnicity, the two splits are taken as independent.
- Rates divide by the Census count of each group. People of two or more races who are not Hispanic are in no group and out of the total a group is compared with; Hispanic people of two or more races are in Hispanic.
- Middle Eastern and North African people are counted as White by every standard the FBI and the Sentencing Commission have used. No record we publish from has a separate code for them, so the group is labeled White+MENA, and CrimeDB builds no separate MENA count until both the Census and police records report one.
- The victim survey puts Pacific Islanders in Asian, and its public files don't split them, so survey figures have no NHPI group.
- A trend shows a break where the codes changed.
The rules every figure follows are under How we count.