William M. Briggs

Statistician to the Stars!

USA Homicide Rates: 1950-2010; By Race And Sex

How has the homicide rate changed through time? The Bureau of Justice Statistics of our great government compiles statistics on just this kind of thing.

This data arises from the report “Homicide Trends in United States” by Erica Smith and Alexia Cooper, from their table “Number of homicide victims, 1950-2010.”

This first chart is the rate per 100,000 population. Note that the early 1980s and of course the late 2000s were period of recessions.

Homicide rates per 100,000

Let the theories fly!

The next two figures break down the rate by Whites and Blacks. Note that the scale changes from picture to picture (the White rate is about a tenth of the Black rate). There are two lines in each: the red line shows the homicide rates (the “Killed”). The black line shows rate that each group was the assailant (the “Killers”).

Homicide rates per 100,000 for Whites

Whites killed and are killed by about the same, and falling, rate.

Homicide rate per 100,000 for Blacks

Blacks kill at higher rates than they are killed. Interestingly, the difference in the killer/killed rate appears roughly constant for most years, and narrowing slightly in recent years.

Finally, the same two plots for Males and Females (note the scale change again; Males are about five times higher).

Homicide rates per 100,000 for Males

Males have either a victim deficit or they are killing at rates higher than one would expect if there were no differences in sex.

Homicide rates per 100,000 for Females

After looking at the Female data, we conclude there is a discrepancy in sex: perhaps a government program can address this.


This post was inspired by my friend Charlie Martin who gave me the link for “Homicide Trends in the United States, 1980-2008“. But that website appears down, perhaps slain by too many hits?


  1. Any ideas on why the rate dipped during the second half of the 1990s?

    Plotting all curves on the same scale in one figure (which is just as easy to do in R) would make comparisons easier and clearer and fairer.

  2. Obviously we need more female mass murderers…a government program to ban chocolate should solve this problem.

  3. William Sears

    6 August 2012 at 1:12 pm

    My theory: This looks like the baby boom, offset by 18 to 32 years which is roughly the most violent time of life. To be meaningful the data should be re-normalized by age group.

  4. @Christina,
    The lack of female murders is obviously caused by male sexism and discrimination against females. We need some affirmative action in this matter. The government should set quotas, like title IX in college sports participation.

  5. Lots of theories are flying around to explain the drop.

    Steven Leavitt of Freakonomics fame posits that the much wider availability of abortion in the 1970s meant fewer unwanted children which led to fewer maladjusted adults in the 1990s. He presents evidence that in states where abortion was available earlier, the crime rate dropped earlier.

    Another team (I forget names) recently argued that the phase-out of leaded gas in the 1970s and 1980s led to fewer kids with lead-induced impulse-control problems.

    Yet another team noted that the spike in crime rates followed the great emptying of our state mental hospitals in the 1960s and 1970s and stayed high until we imprisoned enough people that our total institutionalized populations (prisons plus mental hospitals) got back up to where it had been before.

  6. My theory: The Brady Act.

  7. My guess is that the drop was due to the economy. Affordable living gives you social stability.

  8. GoneWithTheWind

    6 August 2012 at 6:05 pm

    Rudy Giuliani served as mayor of NY City from 1994-2001. He (his policies) dramatically reduced all crimes including homicides in NY City. It was so effective it makes you wonder why Chicago and a handfull of other dangerous cities don’t emulate it.

  9. The first graph, homicide rates, pretty closely matches the recent warming and then stabilization of the last 10 years or so. I nominate AGW and by extension CO2

  10. Perhaps females are inherently more likely to adopt untraceable methods and so avoid detection …

  11. In the ’60s and ’70s it was academically known (from rat studies) that population murder rates were proportional to population density. Ghettos, by inference, were death zones due to too many people in too small a space. But the overall murder rate dropping shows this not to be true: as the American population has increased, the area in which they live has not, yet the murder rate has gone down to 50% of its height. One wonders about the academic interpretation of this fact. The interpretation, of course, is demographics. The 15 -24 age group has fallen.

    But I wonder. I have read of writers alive in Harlem in the ’30s, when there were large numbers of poor, urban blacks locked into Harlem with no opportunities, who said that it was a safe environment. Heroin and cocaine were readily available during this time, as well. Drug users and addicts have been around a long, long time and undoubtedly have always been responsible for property crime as they seek to fund their addiction. But murder does not seem to be a necessary tool for that.

    I go for a cultural devaluation of another life. In Johanenesburg, South Africa, the B&E course is now to murder whoever is in the house being burgled. People lock themselves in the room they are, to slow down or avoid contact with someone who has broken in an adjacent room, apparently. In Nigeria, killing is ubiquitous, with the police forces using killing as a tool as well as the “bad” guy. Death has little import within certain cultural groups.

    I don’t think it is just a “nothing to lose” scenario. It is more a mosquito-squashing ethic: we don’t brush off the annoyance of a mosquito, we squash it and move on. A terminal solution for a momentary annoyance – even a potential annoyance, if the mosquito is still in the air. That is a cultural value issue.

    Nature or nurture: clear to me. When nurture says it doesn’t matter a whit, and is more conventient, we go with expediency.

    So why the fall in murder rate? On this basis, our North American society has “advanced” ethically. In spite of video games, military adventures and violent movies. An interesting phenomenen.

  12. William Sears

    7 August 2012 at 1:35 pm

    Of course, there is also this explanation:


    It’s cycles all the way down.

  13. There were more no-fault divorce laws enacted among the states, encouraging couples to split instead of staying stuck together and seething with rage.
    There were also some drug wars going on in the 90’s that someone, apparently, won.

  14. Milton Hathaway

    8 August 2012 at 4:29 am

    At the risk of being irresponsibly disrespectful to the subject under discussion, I can’t help but relate this to past experience in ‘process control’.

    Five to ten humans lost to murder per 100K per year would be a process yield of 99.99% to 99.995% per year. (I once managed to get a process yield up over 98%, with a lot of time and effort, but I couldn’t keep it there for more than a couple months. The fallout from that process could be reworked or recycled or, worst case, scrapped at some nominal cost. So the analogy is on pretty thin ice from the start, but what the heck.)

    In an attempt to understand the factors that determine this process yield, we instinctively are drawn to examining the process fallout in as much detail as possible. I.e., we look at the perpetrators/victims, collecting as much detail about their lives as possible.

    Compare this to industrial process control. While you can’t totally ignore the fallout, you relatively quickly find and fix the systematic problems, but then you are usually still left with a process yield that is well below an acceptable level. The remaining problems are all over the map, each a small bar on the Pareto chart. If you blindly continue to work your way down the Pareto chart, it turns into an unproductive ‘whack-a-mole’ game.

    At this point, you start to look at the fallout as ‘outliers’. If some important aspect of your process is measurable, and a histogram of these measures looks an awful lot like a Gaussian distribution, the fallout is way, way out on the ‘tails’ of the distribution, and you realize that the fallout isn’t telling you much useful about your distribution.

    At this point, I am pretty clueless how to proceed with the analogy. If one could magically come up with a dozen measurable descriptors of humans that prove to have relevance to the murder rate, one could attempt to control (ok, influence, maybe?) the mean and standard deviation of the descriptor distributions over time, as applied to the vast majority of humans who are non-victim/perpetrator. Basically, it seems like you should have a whole lot more potential data to work with.

    When we concentrate our studies on the perpetrators/victim, any solutions we are likely to come up with will probably amount to an attempt to ‘chop the tails off’ of the distributions. Intuition (and experience) tells me that this is a poor substitute for understanding the underlying nature of your process.

  15. OK, I admit, I must be as dumb a a box of rocks, but I have a question. If there are more black killers than black victims, and more white killers than white victims, who are the extra black and white killers killing? Asians? Hispanics? Amerinds?

    Don’t get it.

    (Unless the figures are skewed by multiple convictions related to a single homicide)

  16. Interesting but the limitations of statistics must be recognized.

    And it must be recognized that culture is the determiner, not race, probably not gender. For example:
    – murder rate statistic in the US are skewed by a violent culture adopted by many young black-skinned males, who tend to kill each other.
    – in AB several years ago, the rate of violence against spouses was about equal between male and female (that was all violence, not only murder)
    – in SW BC there is a high murder rate among middle and high level drug dealers, who kill each other. (Reminiscent of the reputation of Hells Angels and the Mafia.)

    One factor can skew results for another. For example, one study a few years ago showed that crime and welfare rates among recent immigrants were slightly lower than average for the total population. But when the young black male phenomenon was removed from the data, rates were average.

    I’d be interested in shooting statistics by gender where gun control is least, expecting that many females would carry a pistol for protection against rapists. But it may be that rapist wannabes don’t try much in those jurisdictions (I know there is an inverse correlation between severity of gun control and total crime rate.) Females are more of a target and are the first protectors of children, but may be protected by males so have less need to kill. (Certainly they can be very nasty, but typically in more subtle ways.)

    (Persons interested in climate science statistical analysis can noodle about the legitimacy of removing the special case of young black males. I suggest it depends on what the analysis is being used for. Those interested in crime prevention will probably do the opposite analysis, looking at rates for different areas and cultures, so they can prioritize for greatest benefit.
    In climate analysis, removing the effect of volcanoes may be comparable to removing young black males from murder statistics, though it is questionable that

    “Culture” is a substantial determiner of both association and thinking. (Physical location is of course the primary determiner of racial distribution since most people don’t move far, and family connections of both distribution and culture.

    Humans are humans, but they often err in thinking and accept bad ideas (which are commonly taught in colleges). A small example is that a bunch of fools in Seattle erected a statue of Lenin, who ruined the life of huge numbers of people. Why would people who benefit so much from the individual freedom in a country like the US believe in the teachings of the most murderous ideology of the 20th century? Religion is another readily identifiable culture, with murderous results in Ireland, the remnants of Yugoslavia, and the Middle East-Persia-SW Asia area for example.

    Blaming race for ills is like blaming humans for climate variation – ignoring the actual causes.

  17. The murder rates start to even out just a few years after Fl (87) and then OR, PA, WV, GA (89) followed by ID, MS (90) and MT (91) adopted nondiscretionary concealed handgun laws. Between 94 and 96, 14 more states adopted those laws as well, which coincides with the period of rapid decrease in the homicide rates. Perhaps the threat of potential victims being able to turn the tables on their attacker should be considered as having an influence on the decline.

Comments are closed.

© 2014 William M. Briggs

Theme by Anders NorenUp ↑