What’s Behind Gun Murders?

Well, I did this analysis about a year ago, but I hadn’t posted it because I knew that people would call me “RACIST!!” … but what I’m posting are simple facts. Plus at this point I’ve been falsely called a RACIST!! so many times it doesn’t even register any more.

Let me start with a much less contentious bit of analysis. This is the correlation between the rates of gun ownership and the gun murder rate. The short answer is … no relationship.

gun ownership vs murder rates.png

As you can see, gun ownership doesn’t correlate with gun crime. In Idaho about 58% of the households have guns … and yet the gun murder rate is less than that of Rhode Island where only about 7% of the households have guns. And North Dakota has about the same gun ownership rate as Louisiana, but in Louisiana the gun murder rate is twelve times as high as in North Dakota.

Now, there are some policy conclusions to be drawn from this total lack of correlation of numbers of guns and the gun murder rate. The main conclusion is this:

Reducing the number of guns will NOT reduce the number of gun murders.

My other conclusion comes from the fact that the District of Columbia, where about 28% of the homes have guns, has far and away the highest murder rate of any state, more than twice that of the next nearest state, Louisiana.

My conclusion from this is that if you have to live around politicians and put up with their unpleasant lunacy every day of the year, sooner or later you’re gonna want to shoot someone …


With that out of the way, let me move to the more controversial analysis. Finding that gun ownership didn’t predict gun murder rates, I gathered a long list of possible factors of all types that might affect gun murder rates. Here are the factors, with all the percentages expressed as a decimal from zero to one.

  • Population Density
  • % Gun Ownership
  • % Religious Attendance
  • % White
  • % Black
  • % Hispanic
  • % Asian
  • % American Indian Alaska Native
  • % Two Or More Races
  • Income
  • % Population Over 18
  • % High School Graduate
  • % Bachelors Degree
  • % Advanced Degree
  • % Democrat
  • % Urban

So given those variables, how can we determine which ones are important and which are not? One way is to use what is called the “Fast Frugal Tree” (“FFTree”) algorithm. This sets up a decision tree where each fork in the tree divides the outcome variable into two groups. Here’s an example of a simple decision tree:

decision tree example.png

That’s a regular decision tree. What the Fast Frugal Tree algorithm does is that it searches among all of the provided input variables to find the smallest number of variables that most correctly divide up the result variable. In my analysis, the output variable is whether the gun murder rate is MORE or LESS than the median of the murder rates of the individual states.

So with that as prologue, here is the result of the Fast Frugal Tree algorithm looking for explanatory variables for gun murder rates, choosing between all of the variables listed above.

fft gun murders.png

So what does this Fast Frugal Tree (FFTree) decision tree mean?

We start at the top. Per the first box [below where it says “FFT #1], if the black population of a state is greater than 9%, the result is that there will be MORE gun murders. The algorithm puts 21 states in that category, of which 19 are correctly categorized as “Hits”, and 2 are incorrect (“False Alarm”)

From there, we move to the second fork in the decision tree. It says that if there are more than 88.4% high school graduates in a state, there will be LESS gun murders. This fork correctly identifies 20 states as having gun murder rates lower than the median.

The final fork of the decision tree looks at the white population of the states. If it is less than 66%, the algorithm correctly identifies 4 states as having MORE gun murders. And when the white population is more than 66%, it correctly identifies 5 states as having LESS gun murders, and it misses with one state.

Overall, the FFT algorithm correctly identifies 48 states, with two “False Alarms” (rated more gun murders, actually less) and one “Miss” (rated less, actually more).

What do I think about this result? First, to me, this FFTree says very little about race. To me, the division lines are all about culture and education. The racial aspect is just a marker for two very different cultures. Sadly, far too much of the inner-city black culture is all about having “street cred” and being “gangsta” … in other words a culture of violence.

It also has a lot to do with our antediluvian drug laws. Folks don’t seem to realize that we have only two choices:

  1. Junkies on the streets, or
  2. Junkies on the streets mugging grandma because junk is illegal.

We don’t have a choice called “No junkies on the streets.” In Singapore, they have draconian drug laws, and they hang people for dealing. Guess what? They still have junkies. We fought the drug war and we lost.

A huge amount of gun violence is a result of local drug turf wars. And sadly, for a host of cultural, historical, and legal reasons, much of that violence involves young black men killing other black people. I discussed here that we should follow Portugal and legalize all drugs.

I would suggest that another important cultural variable has to do with the destruction of the black family by our idiotic welfare system. Back in the time of LBJ, people thought it would be a good idea to support single mothers … what’s not to like? Don’t single moms deserve support?

Unfortunately, if you pay families to not have a man in the house, the result is quite predictable—we’ll see more and more single mothers. At present, the Census Bureau says that 74.3 percent of all White children below the age of 18 live with both parents, but only 38.7 percent of African-American minors can say the same.

I would say that all of these cultural factors have combined to give us the result of that first Fast Frugal Tree.

Finally, what happens when we remove all the racial information from the input variables and repeat the FFTree analysis? Here’s that result:

fft gun murders no races.png

Given that having a high-school education appeared in the first FFTree, it’s no surprise that the first fork in this tree is the number of high school graduates. States with more than about 90% high school graduates have fewer gun murders. 20 correct, 1 wrong.

The second fork was a surprise to me … if more than about a third of the people regularly attend religious services, the state has MORE gun murders, not less. After some consideration and research, I think that the answer to the conundrum is that most of those states are in the South and Southeast, where both religion and the gun culture are strong, and where people often tend to settle disputes in … well … a less than tranquil fashion. 14 right, two wrong.

Finally, state population—less populated states have a lower rate of gun murders. 3 right, 2 wrong. More populated states have a higher gun murder rate. 7 right, 2 wrong.

Overall, this is less accurate than the previous analysis. That analysis had only three states miscategorized. This one is more than twice as bad, with seven states miscategorized.

So … what does all of this mean regarding trying to reduce the number of people killed by guns? Several conclusions:

Reducing the number of guns will do little to reduce the gun murder rate.

Getting a high school education is crucial.

Our welfare and drug laws badly need overhauling.

It will be a long, slow process because the problem is cultural, and changing cultures is a very gradual process.

That’s what the facts show. Me, I think it has to do with culture rather than race, but if you want to call me “RACIST!!”, well, all I can do is laugh at such foolishness.

My warmest regards to everyone, black, white, or any other color. For me, the only color that matters is red.

Why red? Because if you bleed any other color, I get nervous real fast …

w.

ANALYSIS NOTES: I’ve put a copy of the data I used here so people can do their own analysis if they wish.

For doing the actual analysis, I used the R computer package FFTrees.

37 thoughts on “What’s Behind Gun Murders?

  1. it is so sad and tragic that so many of the black population is self-destructing. MLK Jr. must be turning over in his grave.

    Like

  2. “The final fork of the decision tree looks at the white population of the states. If it is more than 66%, the algorithm correctly identifies 4 states as having MORE gun murders. And when the white population is less than 66%, it correctly identifies 5 states as having LESS gun murders, and it misses with one state.”. But when I look at the diagram, it seems to indicate the reverse: that a white population of greater than 66% indicators less murders.

    Like

  3. With any analysis like this, I think that you need to go a little deeper than per state, there are a few large cities that have drastic effects on the overall stats for the state and they need to be looked at separately. I believe that I’ve seen stats that said that if you removed Chicago, the rest of the state would end up with drastically different ratings for example.

    it may make sense to evaluate by county

    Like

    • Thanks, David. You may be right about the big cities. However, with thousands of counties, I fear such an analysis wouldn’t have enough data in each county to make statistical sense.

      So I tried to approach that obliquely by using the amount of urbanization, figuring that if big cities upset the apple cart that would show it. But the urbanization didn’t get picked by the FFTree algorithm.

      w.

      Like

  4. The difference between a homicide and an aggravated assault is the speed of the ambulance and the skill of the surgeon. It is why I always question homicide statistics, especially when the margins are thin.

    Having said that, your analysis is spot on.

    Like

  5. Nice work, but you miss the point completely. The point is a cultural evolution. At a nation educated by western-style movies, a low level gun violence (probably generating most of your data) is simply a fact of life. A younger generation, educated on violent video games, does not want merely to kill an opponent – they want a glory directly proportional to the number of killings. We object to becoming a target of random shooting every time we go to school or a concert or a shopping mall. And after each mass shooting, we hold hands, to prevent a repetition – but it does not seem to work.

    Like

  6. W, I have been reading your blog here and over on Anthony watts site. Great insight and data driven. I appreciate all the work.

    We hear lots of times about violent video games every time there is a mass shooting. The data just doesn’t support it. Here are FBI actual statistics. Im sure the violent games do cause some impacts, but the assault rifle is a small portion of the overall murder rate. Handguns, drugs, alcohol, hands/feet and clubs/hammers all caused significantly more deaths than assault rifles or murders due to video game violence.

    So in 2016 there were 11,004 fire arm murders, 374 by rifle (including all rifle types and assault rifles), 47,000 suicides, 70,000 drug overdoses, 10,497 dui deaths. There were 656 by hand or feet and 472 by clubs and hammers. Seems like drugs, cars and alcohol are much more dangerous than assault rifles. Maybe we should ban them.

    There are 300 million guns in the united states, assuming that each murder was committed by one unique and individual gun, that’s .0037% of guns were used for murder. Or 99.9963% of guns were not used for murder.

    Like

    • First off, there are few dumb engineers. Second, welcome to the blog. Third, I’d say you are right about video games. If they were a big issue, the FFTree algorithm would have picked out the variable I had in there for age … but it didn’t, even after removing all the race-baced variables.

      And finally … welcome to the blog.

      w.

      Like

  7. Too bad the data doesn’t distinguish between long guns, hand guns, and the most popular “bad” guns (the ones that look military). Sniper murders are not very common, the “bad” guns get lots of press, but I’d wager most murders in the US are done with hand guns (and that countries without many guns are not also without murders). I wonder if that would make a difference. What happens if you do the same analysis using crime rate instead of gun ownership? Would it give the same answers?

    Like

    • Another factoid, via Wikipedia: only 15% of murders in the U.S. are committed by strangers.
      Also via Wikipedia, intimate partner violence makes up 14% of all homicides in the US.
      A person could get lost in this topic. Comparing with other countries is interesting. The US has a lower murder rate than the world average, but considerably higher than other developed countries. But in some of the countries that are worse, it’s BAD! And also some US cities are very bad.
      https://en.wikipedia.org/wiki/Crime_in_the_United_States

      Like

    • Thanks, John. The big cities have the big murder rates and numbers, and they are overwhelmingly democratic. If I’d found that % Democrat actually got picked up by the FFTrees algorithm I was planning to add % Republican, but it didn’t get picked up. Finally, independts are generally not a huge % of the electorate, so 100% minus % Democrat is a rough proxy for % Republican.

      Interesting question, and welcome to the blog.

      w.

      Like

  8. Thanks for this analysis Willis. So many anomalies to ponder.
    However, the biggest one for me is –
    why do all the metrics change for the better as soon as the focus moves north of the 48th parallel?

    Like

  9. “That’s what the facts show. Me, I think it has to do with culture rather than race, but if you want to call me “RACIST!!”, well, all I can do is laugh at such foolishness.”

    You got a minute?

    I find that there is something I call “differencism” that may be inherent in humans. We all notice differences in others: they are taller/shorter, thinner/heavier, the “opposite sex”, have a different hair color, older/younger, different skin color (including well tanned Caucasians), and much more. Do we act differently to these differences? Sometimes, sure. For example, I would not ask a short person to help me get something off the top shelf of a grocery store. We should not assume that because we notice differences in others that we are automatically racist, ageist, sexist, etc-ist.

    Just an observation.

    Like

  10. Hi, Willis. I looked at your data (thanks for that) but didn’t see where you sourced it. Is it from FBI reports? Thanks for the informative analysis.

    Like

    • Roy, I got it from a host of sources as you might imagine. Sadly, I cannot locate any reference to where I got them … population numbers are from the Census, other than that, not sure. I think that many are at best good estimates, as different sources often give slightly different values.

      Like

  11. One confounding element might be that your analysis uses gun ownership but does it account for guns that are illegal. Inner cities are notorious for illegal guns so the actual number of guns present might be skewed towards those states that are more law abiding about gun ownership. Don’t know how you would estimate the real number of guns legal and illegal.

    Some of the numbers are surprising such as the high percentage of gun ownership in Hawaii. Never would have guessed.

    Like

  12. “I would say that all of these cultural factors have combined to give us the result of that first Fast Frugal Tree.”

    One of the factors you did not mention, and I don’t know how to categorize it except as politics gone awry: bussing; that is, bussing white children into black neighborhood schools and black children into white neighborhood schools. The rationale was that white mothers would fight the powers that be to upgrade the schools in which their children were bussed thereby lifting black children’s education as well. Essentially, bringing white middle class values, culture and values into neighborhood schools. Kind of like a transplant; a new heart or lungs.

    That didn’t work out so well it turns out, bussing acted on cities like a neutron bomb, that is, eliminating most of the people and leaving the shells of buildings to decay. Those left standing are majority black. City services nearly disappeared amongst the purging of white officials and staff including police presence and law enforcement. Black ownership of guns became much more common among many of the every day black folk who are/where not gangsters.

    Now, one of the difficulties of removing guns on a wholesale scale is the opposition by the African American community who have guns and taking their guns away make them feel vulnerable.

    For thorny problems that don’t lend themselves to simple political solutions, un-intended consequences seem the rule and not the exception.

    Like

  13. Hi Mr. Eschenbach, are suicides with guns included or just plain murder? Here in Switzerland we have got 50% of Households owning guns, mostly because after completion of our army duties we are allowed to keep our army gun. So we have 2.125 million guns with a population of 8 million. 231 people killed by guns in 2015; 210 suicides and 21 homicides. If I am not mistaken, that brings us to a murder rate of 0.26 per 100‘000 – not bad? So I am with you, gun ownership has got nothing to do with murder rates. Regards from the mountains of Grisons Freddie Stoller

    Like

  14. Nice, succinct writeup. What is tricky in this sort of analysis is to figure out how correlations (or anti-correlations) play into it. For example:
    – It is possible that a higher percentage of African Americans could identify themselves as religious. This would explain at least part of the religion factor.
    – In a state like New York, you have the odd situation that there is a high rate of gun ownership in the rural areas and small cities, with almost no (legal) gun ownership in big cities like NYC (that accounts for almost half of the population of the state). The murders are not happening in the areas with a high rate of legal gun ownership.

    Like

  15. If you plotted all other Western countries on that chart, there would be a clear correlation, so why not ask a different question
    What would it take to bring the US Gun Murder Rate down to below 1 per 100,000? That would still be an order of magnitude worse than most European countries.

    Like

      • David
        It’s a shame your understanding is not based on any actual facts.
        For instance, the total murder rate in UK is around 1 per 100,000 – lower than just the gun murder rate in any state in US.
        The gun murder rate in UK is around 0.01

        Like

    • AndyL,

      Suggest to the ‘gangbangers’ that they kiss and share their drugs and wealth and show love towards one another. That’s the ticket!

      Like

    • AndyL, to severely reduce the gun murder rate, the best thing to do is to legalise all drugs that are currently illegal. At the same time, you would licence their manufacture, distribution, & sale to consumers, along with regulating the purity & dosage of all active ingredients.

      A few of the many benefits of this move would be: an increase in the taxes received; a reduction in the prison population; a reduction in the corruption caused by a hugely profitable black market at all levels of society; fewer people being killed as a result of that black market.

      Like

Leave a Reply to YMMV Cancel reply

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

WordPress.com Logo

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

Google photo

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

Twitter picture

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

Facebook photo

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

Connecting to %s