The Problem with Averages in Understanding Guns, Violence, and Crime (Take 2)

When I tell my more liberal friends I am studying “gun culture,” they frequently hear me saying “gun violence,” since their primary association with guns is with violence. Although my interest in guns is actually in the culture that surrounds it rather than violence, I still spend a few days on “gun violence” — firearm-related injury and death — in my Sociology of Guns seminar every semester.

An article we are reading this week – “Firearm Injuries in the United States,” published by 4 authors from the Centers for Disease Control and Prevention, it is interesting to note – shows that the firearm homicide rate in the United States was 3.66 per 100,000 from 2010-2012.

Source: Fowler, et al., “Firearm Injuries in the United States,” Preventive Medicine, 2015, p. 10. Note: APC = Annual Percentage Change

Taking an aggregate statistic like this, we often hear about how much higher the homicide rate is in the United States than other “similar” countries.

But there is a problem with such population averages: they gloss over important differences between subpopulations within the United States. For example, according to “Firearms Injuries in the United States,” the firearm homicide rate for those 25-34 is more than four times greater than the rate for those 55-64 (8.01 vs. 1.47). The rate for men is 6.13 and for women 1.15. The rate for non-Hispanic Blacks is 14.78 compared to 0.99 for non-Hispanic Asian/Pacific Islanders.

Of course, these differences in subpopulations are related also to economics, and economics are closely related to residence in the United States. As I have argued previously, the problem with averages is that no one lives in “The United States.”

The second reading for my class this week is by Andrew Papachristos (with Christopher Wildeman), whose work I like so much and about which I have written a number of times (here and here and here and here). His use of social networks to analyze the concentration of firearms injury and death provides some dramatic and easily understood statistics. For example, in one study he found 85% of all firearms injuries in one Boston community took place within one social network of just 5% of the population (depicted below).

Papachristos: Co-Offending network of high-risk individuals in a Boston community, 2008. Red nodes represent the victims of fatal or non-fatal gunshot injuries, and these are clustered within the network.

As the CDC researchers observe, “firearm violence is not evenly distributed by geography or among the populations living in these communities. Rather it is highly concentrated in specific ‘hot spot’ locations and often occurs within high-risk social networks” (p. 11). Citation: Papachristos (among others).

This extreme concentration of firearms violence in the United States was discovered by the gun violence newsadvocacy outlet The Trace last summer and they were reminded of it recently by some work done by The Guardian newspaper. According to The Trace, “While President Trump and Attorney General Jeff Sessions have emphasized the national murder rate when discussing their strategies for tackling violent crime, The Trace has argued that drilling down to the neighborhood level is the best way to understand gun violence.”

Good to know that The Trace is not going to emphasize the national murder rate anymore, since it is unhelpful in solving the problem of crime, injury, and death.

The Trace continues, “St. Louis, as a prime example, had the highest homicide rate per capita in the United States over the last three years. The homicide rates in several neighborhoods in the city are so high that — as we’ve noted — they exceeded those in Honduras, the deadliest country in the world. In other neighborhoods, especially those that are majority white, the risk is negligible.”

Or, as I wrote almost 2 years ago (April 10, 2015):

The problem with averages is that there is no “United States of America” when it comes to guns, violence, and crime, but many Americas. Some of these Americas – like my neighborhood in Winston-Salem – are more like our first world counterparts in the OECD, and some of them are more like the third world politically, economically, and socially.

This is not to say what The Guardian has done isn’t helpful. It is the sense conveyed that they are discovering something that we haven’t known for some time that is bothersome. I very much like the graphical depictions they provide which show the concentration of violence in the United States.


Not quite the 80/20 principle, but same idea…


Burrowing down to particular neighborhoods, e.g., in St. Louis. The murder capital of the United States, but again, some parts more dangerous than Honduras and some part as safe as Switzerland.




  1. I suppose that being against mandated CCW training is also a reflection of your interest in gun ‘culture.’ But if you were ever to come to your senses and change your mind on that issue, the academics like Papachristos and Zimring (who used the Guardian data in his new book) might actually take you seriously and I would give you a footnote in my 60,000-word book that Johns Hopkins University Press is releasing later this year. I was always impressed by the syllabus you created for your class; in fact I recommended it the other day to a faculty member who is doing a segment on gun control in a Freshman writing course, and I have recommended it to others.


    • Mike,

      In those 60,000 words did you manage to come up with any -evidence- supporting statistically significant differences in rates of accident or criminal misuse among permitted carriers in states that mandate training and states that don’t, and/or within states that used to mandate training but dropped it?

      All that data is available, if you are truly interested you could go look at it and support your assertion with some proof. Yet you haven’t. Given the amount of money the anti-rights crowd drops on mendacious studies supporting their positions you’d think if there was any pseudo-scientific claim to be made they’d have made it even if you choose not to.

      The burden is on those stating -mandated- training is necessary to prove that such training, due to its additional infringements on the free exercise of the RKBA, in particular its associated time and monetary costs which disproportionately impact the disadvantaged, has such a positive impact on those rates of misuse as to demonstrate it is substantially related to furthering the government’s public safety interests. Absent such proof claims that -mandated- training is necessary for safety are so much wind.

      Liked by 1 person

      • I suspect you are right, Matt. I wonder if there is any evidence that mandated CCW training statistically reduces bad shootings by permit holders, or reduces accidental shootings. What my instructor told me was that one of the primary outcomes of the required training in New Mexico is that it makes sure those with a permit shoot their guns at least every other year. He implied that many students do the bare bones minimum.

        I suspect that training or not, the simple act of applying for a CCW permit weeds out those who are at highest risk of accidental or criminal misuse of guns. What will be more interesting is seeing if those states that have instituted Constitutional Carry see a difference in criminal/accidental use of concealed firearms in the years ahead.

        I have suggested that if training and background checks are so valuable, they should be free to whoever wants it. If indeed these policies reduce gun violence, then given the high costs of shootings, they should pay for themselves.

        Liked by 1 person

      • Khal,

        I do know from the data we have, across the states that release it anyway, that regardless of those state’s particular license/training requirements, permit holders get permits revoked for all reasons, most of which are either technical (bad address, etc), or not related to firearms violations (say, a DUI), consistently in the low single digits. As far as I can tell that has been the case for decades in some of the early adopters. I think the “self-selecting” has something to do with it, but also that most carriers don’t carry all the time, most will never face a reason to draw even if they are, and, if they obey Tamara Keel’s (she popularized it IMO anyway) “Stop Touching It!” rule, the chance for accidents in public is low, guns being essentially inert.

        To track Permitless Carry State number changes directly, in a meaningful way, we’d have to dig into the incident reports to look at the numbers of persons committing crimes and causing accidents *that were eligible to carry legally* from pre and post Permitless passage.

        The biggest risk factor for crime and accident is a history of such, so even if there was some kind of increase, unless we are assigning some kind of indirect causation to non-prohibited persons carrying, we have to make sure we are only looking at those non-prohibited persons the law is actually impacting.

        Although, I suspect most (otherwise) law-abiding people who felt the need to carry regularly did so even when it was illegal, got a permit when it was available, and continue to carry, with or without permit, after Permitless passage. They are probably pretty safe carriers in any status.

        So the margin we are looking at is the potential increase in people who didn’t/don’t carry regularly *and* who thought getting a permit wasn’t worth the trouble, but who might do it occasionally now that a nominal barrier is gone. That diffidence might be the difference?

        Liked by 2 people

  2. States are extremely heterogeneous units. Most are mixtures of primarily low crime areas and a few very high crime areas, suburbs, rural areas and urban areas, high gun-ownership areas and low gun-ownership areas. Generally speaking, gun ownership rates are lowest in the urban areas where crime rates are highest. The larger the units analyzed, the greater the heterogeneity, and the greater the potential for aggregation bias.
    Let’s say you believe that more guns = more crime. Suppose, however, that in states that have more guns the increases in gun-carrying occurred largely in suburban, small town, and rural areas, while the increases in crime rates occurred in big cities. Surely this would cast doubt on the notion that more guns and the increases in gun-carrying were responsible for the crime increases. State level analysis (or even county-level) makes it impossible to detect these details.
    Unfortunately, the FBI’s national crime data only provides gun murder statistics down to the city level, which masks the clustering of violence within neighborhoods and streets.
    Take a look at Ilinois as an example. If you look at Chicago’s homicide rate, it is 18.5/100k population. When you look at the rest of IL that is NOT Chicago, that rate is 1.5/100k population.
    When gun control advocates suggest it’s because of the weak gun laws in Indiana, what they’re not telling you is that Chicago is not the only portion of IL that shares a border with IN. IL shares borders with many so-called lax gun law neighbors. But it’s only Chicago that has the high rate of firearm murders.
    This is true for most cities and states. Louisiana is also considered a high homicide state, but most of those homicides are heavily concentrated in the St. Roch neighborhood, Elysian Fields Avenue between Brother Martin High School and I-610 and along Old Gentilly Rd, all in New Orleans.
    In Wisconsin, most of the state’s homicide is driven by three neighborhoods in Milwaukee – Metcalf Park, Park West and Concordia.
    Pick a state and it’s almost guaranteed that homicides will be driven by geographically small neighborhoods in one or two cities. In neighborhoods that are majority white, the risk of being shot is negligible.
    This completely demolishes the gun-control advocates’ arguments. Which is why they concentrate on macro instead of micro.

    Liked by 5 people

  3. Big fan of Papachristos work. The CDC found similar relationships in a Wilmington, DE study not long ago. Without such precision in identifying the sub-populations where violence is actually concentrated, we are stuck with feel-good, broad-based “solutions” which waste limited resources and may actually be counter-productive.

    I think it is important to note that the concentration is not just in associations, “hanging out with the wrong crowd” and such, but for the most part involves perpetrators and victims constantly switching roles. Violent people preying, for the most part, on each other. A King County study makes this clear and buttresses the criminal history component from the Wilmington study.

    “Hospitalization for a firearm-related injury is associated with a heightened risk for subsequent violent victimization or crime perpetration. Further research at the intersection of clinical care, the criminal justice system, and public health to evaluate the effectiveness of interventions delivered to survivors of firearm-related injury is warranted.”

    Of course, we see this concentration in every aspect of violent crime, particularly ones with a lot of public attention. It is that concentration among “people not like us” culturally that makes violent behavior so hard to understand and that leaves us feeling hopeless about finding solutions. So hopeless that we look for quick fixes like (trying to) deny those violent people a means of perpetrating their crimes. Even though that is neither possible, nor will it solve the underlying sociological problems even if it was.

    Liked by 2 people

  4. I’d like to see a study on the correlation between the location of high levels of gun crime and the level of government welfare in those areas.

    Liked by 2 people

    • Jack,

      Not sure what that would tell us. We know that poverty, while not causal, is correlated with criminal violence. But as the studies cited above point out, the number of people *even within those poor areas* who are actually committing the vast majority of the violent crime is an infinitesimally small percentage. The sub-culture of violence responsible for a large majority of death and injury with firearms is probably measured in only a few hundred thousand people nationwide. The Police Commissioner in Chicago quite correctly talked about a few *thousand* hardcore repeat offenders driving most of the violent stats in a city of millions.

      Simply equating those violent few to the larger number of people in poverty, on welfare, or any other correlation, is still too broad to meaningfully address the actual problem, IMO.


      • If our goal is to reduce the gun violence no matter where it is exists, then it is essential to identify all factors that may be contributing to gun violence. It is my humble belief that government welfare programs, which may be well-intended, actually perpetuate a cycle of poverty, broken families, lack of parental involvement, failing educational systems, gang influence on youth, drug use and and subsequent gun violence. If it can be shown through impartial study that there is indeed a correlation between the level of government welfare spending in a geographical area and gun violence within those same areas, then it may be worthwhile to further examine if a change in welfare programs would then ultimately cause a reduction of gun violence in those areas.

        Liked by 1 person

      • I see your point, but, we already have that information, more or less.

        Look at Papachristos and the other studies again, the number of actively violent people in any given neighborhood is an infinitesimal proportion of the total number of people in that neighborhood, much less on welfare, we’re talking single/fractional digit percentages. There’s no meaningful statistical correlation, much less any causality, that would support something as broad as modifying “welfare” for all recipients in the hope of trying to impact the few violent people. The violence is concentrated among a very distinct, very small, sub-culture, trying to impact it by using broad changes to welfare is almost certainly destined to be as fruitless and baseless as trying to do so with gun control.


      • I think that we need to be discussing “violent crime” and not “gun violence”. It really makes no difference if a criminal uses a baseball bat to kill or a “assault rifle”. (FBI stats show clubs are used more frequently to kill than any long gun).
        To reduce violent crime, we need to make sure that it does not “pay”. In these dense liberal population areas, crime actually does pay in that the chances of getting caught and serving a long prison term is not very high and the chances of being “STOPPED” by one of the victims is also not very high. We must understand that the first person that is available to stop a violent crime is the victim. So, if the good citizens in these areas were trained, practiced and armed, there would be a real decrease in violent crime. If the police would actually make “removing” gang members and drug users from society a real priority it would also show real results.
        And of course, we need judges that give out the max sentences to criminals that are involved in gangs or drugs.
        I also would support making “pot” legal at the federal level. It should be treated like tobacco. This would leave more resources to stop the more serious drugs like Crack.

        Liked by 1 person

  5. Well, it certainly is not a surprise that areas with major black populations are allways on the top of crime rate statistics in the U.S. for it is the same story everywhere, be it Europe or Africa. Some people wonder why the “post-apartheid” Republic of South Africa is in shambles. Well, the fact is that every African nation ruled by blacks have the same problems: crime, poverty, corruption on all levels.
    So it is obvious that what needs fixing is the mindset of black populations.

    Liked by 1 person

    • You are exactly correct, today. However, in the past other minorities have had that distinction and most likely in the future some other minority will have that distinction. We need to fix the “criminal” and not get side tracked by skin color or race.

      Liked by 2 people

      • Great point and thanks for making it. In the United States, at least (and I suspect elsewhere in the world), a small minority of individuals in any community are responsible for the majority of the harm. And the majority of individuals in those communities who are being harmed are from the same racial/ethnic group. I have no empirical basis for saying this, but I am guessing that the majority of the victims of ISIS are other Muslims.

        Liked by 1 person

      • It is always a small minority of any group. While disproportion between groups shouldn’t be ignored, its utility largely lies in asking why those sub-subcultures of violence may be more likely to appear within a given (sub)culture. And thus how that larger (sub)culture might need to look at itself in terms of what behaviors it may be consciously or subconsciously rewarding. While some of the new research on genetics is interesting, I’m not convinced that culture isn’t the overwhelmingly primary factor.

        My personal take, initially cribbed largely from Bowman, is it reflects the degree to which the “honor culture” elements have been constrained or redirected by other cultural factors. In the West, that pacification of the “warrior” honor cultures was accomplished by the Church and state working together to subvert the “Avenger” role into the “Protector” role by changing the cultural rewards for each. Largely to diminish the negative societal effects of personal vendettas and never-ending family feuds as states coalesced and nationalism arose. Yet, even with those centuries of cultural modification, we still see apparent persisting differences in violence rates even among fully Americanized British sub-groups which share some residual “honor culture” heritage.


  6. The research verifies what most of us know instinctively. It’s bad for your health to engage in criminal activity or to associate with criminals.

    Liked by 1 person

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.