US Hate Crime Statistics, 2020-2022 · 5:47am Nov 18th, 2024
One of my side projects for the past year has been studying the FBI's Uniform Crime Reporting system's hate crime statistics. I'm not nearly done, but I thought that, given how many people on fimfiction are quite worried about hate crime right now, I should give some basic numbers on the frequency of hate crimes by category, so that we have some idea where things stand today. I don't expect the FBI hate crime data for 2025 will be available until 2027, so it will be a long time before I can give you any info on how the election affected them.
I only included victims of aggravated assault and murder because the definitions for lesser crimes vary more from place to place. (I previously ran into trouble comparing the number of assaults in Seattle vs. Baltimore, because Seattle had redefined "simple assault" to include hate speech, making it look like Seattle was the most-dangerous city in America.)
#: # victims of aggravated assault & murder in that category, 2020-2022
fraction: fraction of the US population in that category
# / 1000: number of hate crimes per year per 1000 people in that category
VICTIMS # fraction # / 1000
hetero 5 0.924 5.46591455682365e-06
protestant 8 0.38 2.12652844231792e-05
male 14 0.491 2.88012507971775e-05
catholic 8 0.22 3.67309458218549e-05
phys disability 16 0.238 6.79059502588914e-05
female 42 0.505 8.4008400840084e-05
atheist 4 0.04 0.000101010101010101
jehovah witness 1 0.01 0.000101010101010101
other religion 35 0.27 0.000130939019827909
bisexual 13 0.044 0.000298438934802571
mental disability 46 0.139 0.000334278032119759
buddhist 4 0.01 0.000404040404040404
other christian 6 0.01 0.000606060606060606
hindu 6 0.01 0.000606060606060606
white 577 0.578 0.00100835343049876
orthodox 4 0.004 0.00101010101010101
lgbt 228 0.076 0.00303030303030303
latino 636 0.19 0.00338118022328549
asian 265 0.0577 0.00463911209145178
queer 35 0.0076 0.00465178096757044
jewish 99 0.02 0.005
pacific 10 0.0019 0.00531632110579479
amerind 47 0.0084 0.00565175565175565
islam 56 0.01 0.00565656565656566
lesbian 75 0.01045 0.00724952878062926
arab 53 0.005 0.0107070707070707
black 1807 0.1247 0.0146371493604854
trans 185 0.0114 0.0163919900762006
gay 588 0.014 0.0424242424242424
sikh 33 0.000686 0.0486111313657492I hope the spaces & tabs work the same on your screen as mine!
The same figures, but just for murders:
female|1|0.505|2.000200020002e-06
white|4|0.578|6.99031840900353e-06
mental disability|1|0.139|7.26691374173389e-06
latino|2|0.19|1.06326422115896e-05
buddhist|1|0.01|0.000101010101010101
queer|1|0.0076|0.00013290802764487
asian|9|0.0577|0.000157554750275721
black|36|0.1247|0.000291608952394839
islam|4|0.01|0.000404040404040404
lgbt|33|0.076|0.00043859649122807
gay|8|0.014|0.000577200577200577
trans|12|0.0114|0.00106326422115896
sikh|1|0.000685714|0.00147306458684088
(Sorry; no time to reformat these.)
All of these figures are slight under-estimates, because they don't include numbers from areas that don't report crimes to the FBI. I say "slight" because nearly all hate crimes occur in big cities, and all the big cities that I checked report their numbers to the FBI.
The DOJ report that Georg linked indirectly to says, "Because the UCR program does not include federal law enforcement agencies, SHR data do not measure homicides occurring in federal prisons, on military bases, and on Indian reservations."
I don't know how accurate the numbers are; but the ratios between the numbers for different categories, and hence the ranking, should be more accurate.
We have some problems interpreting the LGBTQ+ numbers:
- These could be under-estimates, because police might not know a person was gay; or know that a bias-motivated crime was bias-motivated; or over-estimates, because police might assume that the murder of a trans person was for being trans. These reports are from different police organizations using different methods for collection & classifying the data.
- There are separate categories for bisexual, lgbt, lesbian, gay, queer, and trans. The "gay" total is greater than the "lgbt" total, so I don't know what they mean by "lgbt". Maybe they use that for crimes against groups.
The third column is based on figures for the fraction of the population who are in that category, and estimates of those numbers vary. My sources are:
- Some of those figures, especially for trans, have been changing rapidly recently. I didn't look at years before 2020 because the trans population was so much higher in 2021 than in 2019 that including 2019 would have given the illusion that hate crime against trans is lower than it is.
Intersectional risks are not always the sum of the individual risks. The risks for black trans individuals are higher than the sum for blacks and trans.
- But that means they're all probably averaged into the "trans" category, giving an over-estimate for all other trans people.
You might think that if you're in multiple categories, your risk would be the total of the risk in all your categories. Not always so, because my query only looks at the first victim category entry for each record, because looking at more would make the SQL monstrous. Only 3.4% of the crime records in the FBI's DB have more than one victim category entered, so obviously the agencies reporting the crimes usually just pick one. If you're an Arab Muslim, it's probably a toss-up whether a hate crime against you would be categorized under 'arab' or 'islam', so you should add those risks together. If you're a white trans, a hate crime against you would probably be filed under 'trans', so just check that entry. I think.
Remember that the "# / 1000" column is not the probability per year that you will be the victim of a hate crime, but 1000 times that probability. So if you're trans, the probability that you had in 2021 of being the victim of an aggravated assault or murder that was judged to be a hate crime was 0.0114 / 1000 = 0.0000114, which is 1 in 88,000; for murder alone it was 0.00106 / 1000, which is 1 in 940,000. The chance of being the victim of any kind of murder in 2021 was about 1 in 15,000.
It would be interesting to also see the stats for perpetrators, i.e. which groups are committing the crimes in what numbers.
Also interesting would be to see stats separated by groups to groups. For example, which group is statistically more likely to commit these crimes against another specific group.
5816622
I have those stats, but decided against posting them because I'd get accused of having political intent. The stats on perps aren't as interesting, because they only have 5 categories: Asian, white, Native American, black, and Pacific Islander.
That's one thing I mean to look at. But I'm mostly interested in where, geographically, hate crimes are committed.
Has the FBI finally released the numbers for the major metropolitan areas that had been (ahem) left out earlier in the year? I seem to remember a great deal of loud proclamations "Crime is down! Look how much crime is down! Whoops. I mean it went up and look at that squirrel over there."
(I've always been suspicious of crime statistics where mens rea pushed by a prosecutor can result in a longer sentence. Seems that prosecutors have some sort of magic power that lets them see into the minds of Bob the Drug Addict to get those extra years of prison for their cases. And that's not even counting the hundreds of wildly overcharged 1512c3 cases that recently got tossed when the Supreme Court finally ruled that the Enron-related paperwork destruction statute could not be used against rioters.)
5816625
That's not an isolated feeling (tl;dr summary from another site), and also introduces many politically-based factors that interfere with collecting data and calculating accurate statistics.
Edit: that wasn't aimed at you, just that the initial data collection is/can be intentionally skewed, which lowers the confidence and usefulness of statistical analysis in this specific regard.
5816625
Fair enough.
5816662
The UCR is compiled from crime reports from all around the nation, and isn't usually released until 2 years later. So stats for last year won't appear until next year.
5816667
Okay, I feel like with your having put that out there, I have to show them, because they aren't as lopsided as the stats for the commissions of other violent crimes.
For racial murders in 2020-2022, there were only 2 categories of offender recorded: white and black. The data is:
white|58|0.000101359616930551
black|35|0.000283508703717204
The ratio of the probabilities of committing such a crime, black / white, is 2.8. But if you read my post "The surprising statistics of hate crime", you'll see that's not bad. Each black person in the US is the victim of much more racial animosity than each white person is, not because of any greater animosity per white, but because there are so many more whites than blacks. It would be unreasonable to expect the ratio to be 1.
Stuff like this is why Fermi estimation works: the bad assumptions in both directions end up canceling each other out. It's not statistically rigorous, but it works an infuriating amount of the time anyway.
I’m not sure reporting perpetrator numbers would be considered any more or less political than reporting victim numbers. Given the categories they used, I guess Asian includes Semitic, and white includes Hispanic?
5816693
Asian doesn't include Semitic; those are reported as Arab and Jewish. Hispanic is trickier, since the US Census recently moved Hispanic from a "race" to an "ethnicity", IIRC, and lets Hispanic people count themselves as white and/or as Hispanic, as they choose. I would expect that Hispanic is filed under Latino, which I why I used an estimate of fraction of non-Hispanic whites in the population (0.578), rather than the latest census figures for white, which I think was 0.659. But notice the 'white' figure is much lower than the census figures for 'non-Hispanic whites' plus 'Hispanic'. It's still difficult to interpret the figures, but probably less-difficult than when they forced people to choose between 'Hispanic' and 'white'.
5816674 I wonder how many data points regarding various crimes from previous years have been moved around. The current administration recently 'adjusted' annual murder rates clear back to 2003. (Admittedly from a website, but looks legit)) Either the numbers are being cooked, or there's some mass murderer with a time machine out there.
Nope! The "code" block doesn't really work on mobile, at least for me, it just crams everything together so it's all just a jumble of letters for me.
5816717
I should have been clearer. I was referring to your reply here 5816625 where only 5 ethnicities are used as categories.
5816777
Oh, right. Your guess is as good as mine. I suggest asking a chatbot.