Bryan Caplan  

Who to Fear

A Strange Critique of Libertar... Desrochers on synthetic dyes...
While writing The Case Against Education's section on education and crime, I came across an interesting FBI table on victim/perpetrator demographics for homicide.  "Homicides" includes murder and manslaughter, but excludes "justifiable killings" by authorities or private citizens.

First, the racial breakdown for 2011 homicides:

Race of victim


Race of offender

















Other race






Unknown race






The big shocking fact: The vast majority of homicides are either white-on-white (43%) or black-on-black (40%).  Consistent with stereotypes, white-on-black homicide is rare - 3% of the total.  But contrary to stereotypes, black-on-white homicide is also rare - 7% of the total.  In fact, since blacks are roughly 13% of the U.S. population, black-on-white homicides are almost exactly as common as you'd expect if black murderers randomly selected their victims from the U.S. population.  Murder is overwhelmingly intra-racial.

The results for gender look entirely different:

Sex of offender

Sex of Victim




















Now, male-on-male (61%) and male-on-female homicides (26%) predominate.  Female-on-female homicides are almost invisible - just 2% of the total.  Women are over three times as likely to kill men as women.


1. Gender patterns elegantly fit stereotypes: men commit almost all homicides, and women almost never kill each other. 

2. Racial patterns, in contrast, often clash with stereotypes.  Whites have little to fear from blacks, and blacks have even less to fear from whites.

3. Racial stereotypes would probably be more accurate if you focused solely on homicides between strangers.  But these are rare.  Our focus on such crimes probably reflects the illusion of control; since we personally control who we associate with, we imagine that our associates pose no threat.

4. Prediction: The results by nativity closely mimic the racial results.  When natives think about immigrant crime, they picture immigrant-on-native crime, rather than the statistically predominant immigrant-on-immigrant crime.  Testing my prediction against aggregate statistics is welcome in the comments; testing it against recent media circuses is not.

5. Yes, the government decides whether its own homicides are "justifiable."  But even if you count all "justifiable homicides" by the authorities as cold-blooded murder, they won't change the overall numbers much.  There are only about 400 such killings per year.  Allegedly justifiable homicides by private citizens are even rarer - about 300 per year

The overall lesson: Contrary to bipartisan fear-mongering, don't worry about getting killed by members of other races, police, or gun nuts.  Turn your fear inwards.  Worry about young males of your own race - because that's where the danger is.

Update: On Twitter, Alex Tabarrok and Josiah Neeley point out serious problems with the FBI's statistics on number of people killed by the authorities.

Comments and Sharing

COMMENTS (23 to date)
Nyan Sandwich writes:

>The vast majority of homicides are either white-on-white (43%) or black-on-black (40%).

>black-on-white homicides are almost exactly as common as you'd expect if black murderers randomly selected their victims from the U.S. population.

I can't seem to square these statements. Seems to me that if blacks selected randomly, they would kill a lot more whites.

Further, what do we mean by "white" here? I notice we don't have "hispanic". Might suspect that a disproportionately large chunk of the "white" murderers are actually hispanic.

John Allen writes:

It is interesting though that there are more than twice as many black on white murders as white on black murders. That seems to be the right way to look at it in terms of claims of persecution and gender equity.

Some guy writes:

Please don't take this as an troll comment, but these numbers could to some degree reinforce the stereotype of blacks committing more crime against whites, particularly outside of family members where it's fairly well accepted that a large number of homicides occur. Blacks make up 12.1% of the US population but commit 14.1% of murders of whites meaning they are over represented in this category by 17%. That's without adjusting for murders between family members which you would expect to be almost exclusively white on white crimes. Adjusting for that you would expect to see an over representation well above that 17% mark.

These points are minor compared to the bigger issue. Even more troubling are the staggering number of both victims and offenders that are black given the relatively small size of the population. There are many reasons beyond just race that contribute but we all, regardless of race, need to acknowledge that there are real problems in the black community that need to be carefully studied and addressed.

ed writes:

I believe most murder is among family members or close acquaintances, so the table is not so surprising.

But it is crime by strangers that average people fear the most (this isn't crazy, the reasons are understandable.)

I wonder how the table would look if you limited it to killings by strangers (or near strangers)?

Tatil writes:
There are only 400 such killings [by police] per year
Only 400? When the comparable number is zero or less than a handful in almost all developed democratic nations, one can only hope you used "only" sarcastically.
Tatil writes:

Even though I don't find the results surprising, I am curious: there should be about 15,000 homicides in the US according to UNODC. Your table has about 6,000. Why is there such a big discrepancy? Would the stats on the missing 9,000 victims change the outlook?

Devin writes:

"Racial patterns, in contrast, often clash with stereotypes. Whites have little to fear from blacks, and blacks have even less to fear from whites."

This does not follow from the statistics. Because white people fear black ghetto neighborhoods, they tend to avoid such neighborhoods and act wary when crossing paths with young black men wearing baggy clothes and hoodies. We don't know what the homicide rates would look like if white people did not fear ghetto black people. These statistics might simply show that white fears are useful fears, and are helping white people to not get killed.

Also, stranger homicide, stranger assault, and stranger armed robbery would be much better statistics to look at. If a person knifes his antagonist in a fight at roughneck dive bar, that doesn't really affect my quality of life. But if a gang of youths beats a random innocent to death with hammers in my neighborhood, or if a stray bullet from a drive by shooting hits an innocent pregnant woman, well it only takes one of those innocents to make you want to move to a different neighborhood.

David R. Henderson writes:

The thing that jumped out at me most from this table is how low the absolute number of murders is. I do wonder, though, about Tatil’s point above.

Daublin writes:

I agree with David that above all, the numbers are low. There are 40k-60k deaths per year due to car accidents in America.


Warren Farrell thinks that no one knows whether wives are more likely to kill their husbands than husbands are likely to kill their wives because women's tactics are more sophisticated. Women are more aggressive than men inside home.

Bostonian writes:

As Devin says, you need to normalize the crimes committed by X against Y by the number of interactions of X and Y.

I bet children are killed more often by their parents than their neighbors, but that's because they spend much more time with their parents.

RPLong writes:

@ Nyan Sandwich - The reason "hispanic" is excluded from the tables is that "hispanic" isn't a race. Hispanic is an ethnicity. You'll meet black Hispanics in the Caribbean (if you haven't met them Stateside); you'll meet white Hispanics in California, Puerto Rico, Europe, Argentina, and Chile.

Perhaps you're thinking of Native American Hispanics, in which case they'd most likely be reported as part of the "other races" category, although I do think it's possible that a nontrivial number of them may be included in the white category as well.

Kevin Erdmann writes:

This also seems to be an example of how racism is particularly put into effect through the state. It would take some effort to pull the numbers together, but it seems likely to me that if a black man is killed by a white man, the white man is likely to have been wearing a uniform. That is a staggering outcome.

David Condon writes:

"In fact, since blacks are roughly 13% of the U.S. population, black-on-white homicides are almost exactly as common as you'd expect if black murderers randomly selected their victims from the U.S. population."

That statement is wrong. The relevant figure is percentage of the US population that's white. 13% is the relevant figure for white-on-black, which occurs at roughly half the expected rate. How are Hispanics being treated? Surely they're not other.

mbka writes:

To Tatil and David: If you go to the source pages, the FBI also lists about 14000 murders / year of 2011. But here the fine print says (quote): "This table is based on incidents where some information about the offender is known by law enforcement; therefore, when the offender age, sex, and race are all reported as unknown, these data are excluded from the table. " This leads one to believe that for roughly 8000 murders, or 57% of murders in the U.S., no details whatsoever are known about the offender. That somewhat weakens Bryan's point, unfortunately, especially since it now looks plausible that half the murders are perpetrated by true strangers.

Silver lining: this table shows the staggering and ongoing decline in violent and other crimes in the U.S. since 1992. The murder rate nation fell wide by half. Even property crimes decreased.

Brian writes:

Nyan and David,

Yes, what Bryan meant to say is that homicides are almost exactly as common as you'd expect IF WHITE VICTIMS RANDOMLY SELECTED THEIR MURDERERS FROM THE U.S. POPULATION.

Mm writes:

Good post, but if one has a reasonably well adjusted family your major concern IS stranger violence- so the concern about stranger violence isn't irrational(and the stats are hard to discern from the info provided). Many of the intrafamily homicides are deeply troubled families & well known to the system, but we aren't as good as we would like at intervention in those cases

Dain writes:

"Contrary to bipartisan fear-mongering, don't worry about getting killed by members of other races…"

But of course both sides don't have equal control over the media megaphone. The subtext of most reports is that the country is suffering from white-on-black injustice, and not the reverse, when the numbers you cite would support the latter interpretation (if forced to push one narrative or the other).

Here's a representative example from WaPo: "Darren Wilson, the white officer who fatally shot an unarmed black teenager in Missouri in August…"

Nathan W writes:

Consider the case of sexual abuse, which most often hits close to home and among people who know each other (I'm half watching a movie which includes abuse of a physically disabled person at the moment).

I think there are very few people who just go out and kill peopel with no reason. And for sheer opportunity, the amount of time you spend with people you know compared to the amount of time you spend with people you don't know, in addition to the fact that it is more likely for a disgareement to get out of hand with someone you know than with someone you've never met, then all you ned to consider is that people stlil manifest some degree of segregation (a statistical "preference" for their own colour, but I wouldn't actually think of it this way) in their social cirlces, and you easily achieve the results.

But I'm thinking that if black people are so much more likely to hang with black people, and whites with whites, etc., at this point in time, then an appearance of no effect could actually cover up some underlying tension.

Consider it this way. Let's say that despite being 13% of the population, group A only spends 1% of their total time with people from that group, yet members of that group still feature in 5 times as many murders, five times higher than you would predict if people were equally likely to become homicidal in any given year of interaction (as a unit of statistical comparison). In this situation, the statistics cover up significant evidence of racial tensions by not considering the fact that white and black people both tend to have their own skin colour overrepresented in their social circles. There are probably lots of caveats and the analogies etc don't all quite line up in the right ways, but I think the basic argument is coherent enough and basically represents an important part of the picture.

Nevertheless, even though I argue that it's less innocuous from a race relations perspective than the above presentation of the data, I still think that white people are too paranoid a lot of the time. Just having a little respect for a lifestyle where you work for nearly minimum wage in retail or some factory job and struggle to provide meaningful opportunity to your children due to your own educational limitations, lack of workplace connections, etc., and just respect that these people are mostly working pretty hard (that's not what they're doing when you meet them at home), and just smile and treat people like people, wondering about their hopes and dreams and not whey the haven't managed to transform themselves into whatever you conceive of as an ideal role model for them to be, then generally the sorts of situations where someone might apply related knowledge would become generally more pleasant situations.

caryatis writes:
Contrary to bipartisan fear-mongering, don't worry about getting killed by members of other races, police, or gun nuts. Turn your fear inwards. Worry about young males of your own race - because that's where the danger is.

Of course, some young males of my race *are* gun nuts--or police.

Eric Hanneken writes:

From today's Wall Street Journal: Hundreds of Police Killings are Uncounted in Federal Stats. Alas, it is paywalled, but a a summary of its contents is here. Police departments aren't required to report homicides by police officers to the FBI, and some don't. Also, the FBI doesn't count people killed by federal law enforcers.

Steve-O writes:

[Comment removed pending confirmation of email address. Please see in your email. Email the to request restoring this comment.--Econlib Ed.]

Shahe writes:

Hypothesis: I think that a big reason for the same-race murders is that most murders are usually between people that know one another and people are more likely to have friends/family that are the same race than of a different race.

Maybe some one here can find data for or against this.

Comments for this entry have been closed
Return to top