Gun Control: Four Must-See Graphs

Articles like this one began to appear soon after last month’s tragic massacre in Connecticut. The logic embodied in them is straightforward: Guns are designed to kill people, and more guns mean more dead people, therefore fewer guns mean fewer dead people.

The reasoning is so patently obvious, in fact, that anyone who argues against common-sense gun control like a ban on assault weapons and high-capacity magazines is either an idiot or a murderer-in-waiting, and we can’t afford to let either of these join the conversation. We just don’t have time.

I mean, come on, people! How many more innocent children have to die before we figure this out?

Before we jump any deeper into an emotionally-charged conversation, let’s take our own look at the data. And let’s commit to intellectual honesty. No fudging the facts or conveniently excluding data that do not support our thesis.

Let’s zoom out and consider the issue more broadly than others usually do: Murder of an innocent person in any form is evil, and guns certainly aren’t the only way that bad guys kill. They also use knives, swords, baseball bats, crowbars, rocks, pointy sticks, ropes, lead pipes, candlesticks, automobiles, bombs, poison, fire, water, and even their bare hands, among other things too numerous and too gruesome to list. So why analyze only firearm-related murders?

To the murdered innocent and those who love him the precise manner of his demise is of secondary concern, subordinate only to the principal fact that he is dead, that he has been unjustly deprived of life and this evil deed cannot be undone.

Here is Graph 1, Homicide Rate vs. Firearm Ownership Rate by country. Data sources are here and here:

Homicide vs. Firearms 1

Well, this is troubling. It looks like there are 20 or so countries that have significantly lower firearm ownership rates and homicide rates that are at least as high as that of the United States, as if a higher firearm ownership rate was correlated with a lower overall homicide rate.

How could there be so many murders in those countries with so few guns? Puzzling. The model is probably skewed by all those Spanish-sounding names in the upper left quadrant.

Since we’ve committed to be intellectually honest, we can’t simply exclude these “outliers” and compare the United States against, say, Northern Europe — which is buried in that cluster of blue points in the lower-left quadrant of the graph with practically no guns and practically no murders — even though doing so would clearly support our thesis that fewer guns would make us safer. No, we’ll take the high road and try to find another dataset that controls for factors like living in a narco-state vs. a Scandinavian socialist utopia.

Here is Graph 2, in which we compare Homicide Rate vs. Firearm Ownership Rate by State. The firearm ownership data come from a 2001 survey of 201,881 households by the Behavioral Risk Factor Surveillance System (whose name sounds very safe and trustworthy in spite of containing the word surveillance) and are paired with the 2001 intentional homicide rates by state (compiled from various sources):

Homicide vs. Firearms 2

This whole exercise is starting to get frustrating. This graph exhibits the same troublesome negative correlation between the firearm ownership rate and the homicide rate as the previous graph, but the best-fit regression line is logarithmic, which means that there might be indirect effects for the first few armed households, i.e. their guns may be indirectly protecting more households than one. The R-square indicates that 35% of the variance in the homicide rate is explained by the firearm ownership rate, which is significant even though the factor’s coefficient contradicts our thesis. Regardless, 65% of the variance in the homicide rate across jurisdictions is explained only by factors other than firearm ownership, which doesn’t bode well for our idea that we just need fewer guns to make our society safer.

But most disturbing of all, three of the four jurisdictions with the most-restrictive gun laws and the lowest firearms ownership rates (Washington DC, US Virgin Islands, Puerto Rico) have murder rates that are 2-3x higher than those of the states with the next three highest murder rates (Louisiana, Alabama, Mississippi), all of which have much-less-restrictive gun laws and rates of gun ownership that are 8-10x higher than the more-restrictive states. And this particular cohort pairing shares a number of otherwise blameworthy factors (high poverty rates, high unemployment, poor public education systems, etc.). Not good at all.

Could the United States really be that similar to the rest of the world? These household firearm ownership statistics are probably comparable to the rest-of-world firearm ownership numbers, as they may control for the fact that the US, due to high per-capita GDP, likely has many more households with multiple firearms than does any other country.

Enter Graph 3, in which we compare Homicide Rate vs. Firearm Ownership Rate for the US (by state/territory/district) and the Rest of World:

Homicide vs. Firearms 3

Controlling for US households that own multiple firearms, then, it looks like the distribution of household firearm ownership by state is well within that of the rest of the world. And the same inverse relationship between the firearm ownership rate and the homicide rate seems to hold.

There are exceptions, notably Hawaii and the cluster of blue points that surround it, all of whose names have been invoked by the media with great reverence as examples of the success of strict gun-control laws, but we cannot ignore the reality that more-restrictive gun laws and a reduced firearm-ownership rate can be associated with a significant increase in the homicide rate, both in the United States and the rest of the world.

To say that this observation is counter-intuitive would be a gross understatement.

But really, how bad could it be if we enacted sweeping gun-control legislation at the national level? Let’s assume what looks like the worst-case scenario, with the entire United States adopting the position of Washington, DC:

There would be 3.8 households out of 100 that own firearms (down from a national average of 31.7 per 100, Senator Feinstein would be pleased) and there would be 40.3 homicides per 100,000 people (up from a national average of 4.7 per 100,000 in 2011).

That’s an 8.5x increase in the national homicide rate.

That’s 125,840 homicides per year.

That’s more than 12 Sandy Hook massacres every day.

Perhaps the data do not necessarily support the easy, “intuitive” arguments for gun control after all. It seems that disarming a population is correlated with increasing the variance in the expected homicide rate. There are good outcomes, make no mistake about that. But there are also some very bad outcomes, on the order of 10x worse than the status quo.

But we can still ban those scary-looking assault rifles, right? And those big high-capacity magazines like the terrorists use. There’s clearly no point in having those if you aren’t planning on killing lots of people. They’re no good for hunting, and you don’t need that kind of firepower to defend yourself against a mugger or a rapist or a burglar in the middle of the night.


Wrong. Dead wrong.

It turns out that Washington, DC is far from the worst-case scenario for homicide. And it turns out that gun violence, or violence of any type, is far from the most efficient way to massacre innocent people.

The most efficient weapon in the history of the world is forced starvation, and the Ukrainian language has a word for it: Holodomor (literally hunger-extermination).

The Holodomor was Stalin’s answer to the problem posed by the kulaks, relatively prosperous farmers who owned their land and had resisted collectivization following the Bolshevik revolution. When Soviet economic policies that favored trading grain for farm machinery caused food to become scarce, the government responded in August, 1932 declaring that all food was property of the state and mere possession of food was evidence of a crime. Crops were seized from the kulaks and redistributed to party loyalists under a quota system, while the kulaks and more-affluent peasants were left to starve. By the most conservative estimates the Holodomor killed 3 million people in Ukraine during the following 12 months and 6-7 million during that same time period across all of the Soviet Union.

One of those victims was my wife’s great-great-grandfather, who died in August, 1933 under forced labor on what had previously been his family farm. He had sent two of his sons to the United States prior to the Bolshevik revolution, and today our family is fortunate to have a collection of his letters to those sons, letters that document in excruciating detail the tightening grip of the tyranny that ultimately took his life.

Can you fathom the loss of 3 million lives in one year? How about 3 million out of a total population of 29 million?

That’s a homicide rate of 10,345 per 100,000.

In the United States today, that would be 32.5 million people in a year.

That would be more than 3,300 Sandy Hook massacres every day. For a year.

Here is Graph 4, in which we compare the Holodomor to the overall homicide rates in the rest of the world. The scale of the vertical axis is now logarithmic to accommodate the magnitude of this atrocity. The Holodomor was 100x worse than the homicide rate of modern-day Honduras, currently the world’s most violent country. You will note from the graph that there was no legal firearm ownership in Ukraine in 1932, the Soviets had outlawed firearms in 1929 after years of kulak resistance to collectivization:

Homicide vs. Firearms 4

The actual food confiscation of the Holodomor was carried out not by bureaucrats but by urban thugs loyal to and armed by the regime. They raided farms, built watchtowers over the fields to ensure the peasants weren’t “stealing” grain (the penalty was summary execution), abused the peasants going to and from work, and amused themselves by raping the women who lived alone in the countryside. These mobs had been half-starved themselves and brainwashed by propaganda into thinking they were doing their civic duty, and there was nothing that the disarmed kulaks could do to stop them.

Think this type of massacre couldn’t happen today? You’re wrong. One is brewing right now: Venezuela has long had a strict permit process for firearm ownership. Last year Hugo Chávez enacted sweeping new gun-control measures, banning all retail sales of firearms and ammunition. And just today, his heir-apparent vowed to crack down on “hoarding” and sent troops to take control of food distribution networks. Who will protect the disarmed producers when the loyalist paramilitary mobs come for their food?

Remember the Los Angeles riots of 1992? 53 dead, 2000+ injured, billions of dollars of property damage. 6 days before the National Guard and the Marines were able to restore order to the city. 45% of property damage was inflicted on Korean-owned businesses even though Koreatown was far from the epicenter of the rioting. The business owners had no one to protect them, their families, and their livelihoods — no one except themselves and their neighbors. One gentleman reported firing more than 500 rounds from his rifle — into the ground and into the air — to defend against the mobs of looters after the police had abandoned his block and before the National Guard arrived days later.

This is precisely why law-abiding citizens must never be banned from owning powerful weapons that are designed to defend against large numbers of armed, violent people when tyranny rises.

Sandy Hook was a tragedy, and we should learn from it and do better in the future. We need to look beyond legal firearm ownership as the cause of murder, since the data show that jurisdictions with stricter gun laws and fewer firearms per capita can have significantly higher murder rates compared to states with more firearms.

And as we try to learn from Sandy Hook, let us never forget the 32.5 million reasons why we have the Second Amendment.

35 thoughts on “Gun Control: Four Must-See Graphs

  1. Let me preface my comment by saying that I grew up in a house with guns, know how to use one (and am a pretty good shot), and think responsible gun ownership is a good thing. In order to use my dad’s guns, he required that I learn the different parts of the guns, how to clean the guns, how to safely load and store the guns, and even though I never hunted he asked that I take a hunter safety course. When I was old enough he asked that I take the concealed weapons class (where I earned a permit to carry a concealed weapon in Arizona). That is a lot of training for a person who doesn’t own a gun… Far more than most people who own a gun have. It’s not gun control that I want…it’s ownership training!

    1. Kibbie, great to hear from you. Thanks for reading and for your comment. I think the training you mention is invaluable. I’ve learned a lot from the firearms classes I’ve taken as an adult, and I think others would benefit from them, too. Perhaps we’ll see an increase in demand for training on the coattails of the current gun-buying spree.

  2. Zach, great to see you online. I forwarded your article to a really good friend, a Navy SEAL who ran ST6 for almost 14 years—-he loved it and shared it! Thanks–very informative and good data detail.

  3. Sir, I’m a former Navy SEAL and knowledable in the issues at hand. I forwarded your article and data to some Congressmen for there use…that is if “Executive power” isnt used. Still, in these difficult issues knowledge is ALWAYS required…unbiased and pure. In that alone we can make quality decisions to lead our opinions, our nation and help those who don’t agree…
    With respect.

    mark caracci

    1. Mark, thank you for your service. And thank you for reading, for commenting, and taking the time to forward this article to our representatives. I think there are many who are on the wrong side of this issue only because they don’t know the truth. I appreciate you helping them find it.

    1. Randy, thanks for reading and for your comment. Two thoughts:

      1. I’ll agree that firearm-related deaths are at the mean correlated with the number of firearms when we consider a cohort of “first-world” countries. That’s the same data, cohort, and the implication in the Henry Blodgett article linked in the first sentence of this post. I wrote this article because I think there is more to the story.

      2. For the reasons I stated in my introduction, I think that the overall homicide rate is more important than the narrower firearm-related death rate. Firearms are simply a means to an (evil) end when in the hands of a murderer. If we could magically remove all firearms in an instant, the enmity of man would remain and those bent on homicide would use their human ingenuity to find a substitute for a firearm. And victims killed by knives, clubs, strangulation, bombs, etc. are no less dead than those killed with guns.

      Finally, it might be interesting for you to elaborate on what constitutes a “third-world” country, what factors these countries share with “non-states”, and why those factors matter when it comes to explaining the homicide rates of these jurisdictions rather than merely exclude them from the analysis.

  4. I don’t see any correlation between third-world countries and non-states; I’m just pointing out two different problems with the data. The graph of the 50 states – how would that compare? I’m intrigued by the question of why there’d be a difference for the non-states, but clearly they’re wild outliers that skew the numbers.

    Defining third-world countries: let’s say high poverty. What do homicide rates look like for every country with a median income of X?

    1. There is a lot of analysis yet to be done, and I would caution labeling any cohort as “outliers” and excluding them for the sake of a cleaner-looking model. Particularly when an extreme outlier (like Ukraine 1933) falls roughly in line with a logarithmic regression.

      Have you read The Black Swan by Nassim Taleb? Sometimes randomness isn’t normally distributed, and the rare cases that look like outliers turn out to be massively significant.

      For example, in the life of a turkey every day looks about the same. The sun rises, the turkey eats and gains weight, the sun sets. This pattern continues for roughly 1000 days. The turkey could reasonably conclude that it will continue indefinitely.

      And then on day 1001, the turkey is slaughtered and served for dinner. Although that day was just an outlier.

      1. The clear outliers in your model are non US states which heavily skew the figures – their inclusion appears to show a clear bias on your behalf. Washington DC has a population density of over 10,000 per square mile – to compare this to the likes of New Jersey (the most densely populated state) which has a density around one ninth of D.C’s is frankly moronic. The same goes for the inclusion of third world countries – their levels of corruption, quality of police force, level of education, overall crime, poverty etc etc are in no way comparable to the US or any other first world country.
        Lets see your graphs again but this time with the 50 states and a comparison of the developed first world countries.

  5. Thank you for the article. Sharing!

    I’m surely not a mathematician, but I do need to construct algorithms in the programs I write for my customer. USUALLY, the need and the tendency is to filter OUT anomalies which ‘get in the way’, but if exposing the truth to humans bent on using figures to lie, filtering is an invitation to them, to attack your data – then point and scream, like Donald Sutherland in “Invasion of the Body Snatchers”.

    Since the opposition to our rights uses the cracks and crevices to hide their statistical trickery (lies), I agree with doing the opposite – SHOW IT ALL. We don’t know who will look at this article and data, seeing, in the truth, that which cannot be found in the murdered data presented by the Left. After all, if the problem lies in the oddities (Columbine, Aurora, Sandy Hook, etc…) why filter OUT the anomalies? BRING ‘EM!

    I watched (from the outside, on a few years ago, as people worked to unravel the “hockey-stick” promoted by the UN-IPCC. One guy applied expression after expression after expression to the “raw data”, until, as if by magic, he replicated the “hockey-stick”. He tortured the data to reach the same ends as the liar(s). Was quite an event!

    BTW: I’m kinda watching NUMB3RS right now… A TV show meant for entertainment, yeah, but interesting timing anyway… :)

  6. Real-world example (my real-world anyway) of how one person can change live(s) by wrongly filtering data. (sorry, I get more verbose, as the hour gets later.):

    In grades 8 & 9, I was very unhappy about attending a junior high school, after attending a high school in another state, for grade 7. My algebra teacher determined that my problem was a total disconnect in my brain, when it comes to math. She told my parents (and annotated my permanent school records) that “No matter what, your son should not focus on anything to do with math.”

    Coincidentally, the next semester, I took my first shop class in electronics.

    Moving on to high school: Because of ONE teacher’s assumption, in three years of high school, I was only required to take ONE math class, a semester of “Senior Math” (a.k.a. a joke.)

    What was next? A three year electronic technology program at the local Community College – 27 hours per week of electronics-ONLY classes: half lecture (seemingly everything, but calculus) and half lab. I achieved a 4.0 GPA, in spite of the total arithmetical void of high school.) I didn’t have so much of a challenge overcoming the math, as with overcoming the mental barrier.

    I have spent the subsequent three decades (mostly) self-employed, designing and programming custom robotic control systems and electronic devices for my customers. Math is not a direct, major player in the things I do, but sometimes it is key.

    Expectations of my parents and teachers, were skewed by one teacher. The result was almost skipping college – period, as opposed to following my Dad’s lead: attending a university and going for an engineering degree..

    Do you think my future was served by that ninth grade algebra teacher’s student algorithm[?]

    My point: If data is filtered on presuppositions; assumptions… well, we all know the joke about “ass-u-me”.

    1. Thanks for reading and commenting. I think this is a great example, one that certainly hits close to home. We could talk all day about how the education system is screwing up society by marginalizing the outliers, both good and bad, in any dimension and managing strictly to the mean. I’ll commit to addressing the topic in greater detail in the near future.

    1. Thanks for reading and commenting. I’ve addressed a couple of similar questions about why I didn’t exclude “third-world” nations from the analysis. Generally speaking, there should be a burden of proof for excluding data from a model, not including data. So I’ll ask the converse: What is it about third-world nations that justify their exclusion from a model attempting to explain variance in homicide rates?

      Similarly, I argue above that the *total* homicide rate is what society should be most concerned about minimizing, but the article you linked reports mostly on the “firearm-related death rate”, which excludes homicides by other means and includes firearm-related suicides that are not criminal acts. The “firearm-related death rate” conflates two societal problems that have very different etiologies and appears to show the US on par with Mexico, though 2/3 of “firearm-related deaths” in the US are suicides, 10x more than the percentage of suicides in Mexico.

      This kind of tortured reasoning is precisely what we need to avoid as we study this issue. I’m showing raw numbers for all countries and all states. I’m not aggregating or slicing the data in ways that manipulate it to appear to support some politically-convenient thesis. Though I can’t say I’m surprised to find such an analysis on a blog named after a political icon whose views the author seems to espouse with unquestioning loyalty.

      1. It would seem to me that if you are comparing statistics for a single variable, competing variables should be excluded as much as possible (without polluting the data). I certainly understand that this is never truly possible because life rarely (if ever) gives us the opportunity for true A/B-split testing, we should strive to do so as much as possible. Some of the reason for excluding undeveloped nations include: they generally have smaller governments (less rule of law or at least less ability to enforce any rule of law) and higher rates of poverty. Countries like this, sadly are often run by syndicates who run rampant. These will likely affect homicide rates.

      2. There’s an important difference in experimental design vs. research: In design of experiments you want to control all variables except those you are testing (ideally 1). In research such as this, we’re attempting to build a model, a multi-factor model, that explains a phenomenon based on the relationship between a set of unknown factors. So you add some factors to the model, regress, exclude factors, regress again, add some more factors, and repeat.

        At no time do you exclude data points that do not seem to fit your initial hypothesis. That would be intellectually dishonest. It would be better to address third-world countries with a multi-factor model that includes variables related to these hypotheses to see if they are significant.

        Or you can look for a different (complete) data set that controls for these factors, like living in a banana republic, to see if the same pattern exists. That’s what Graph 2 was about, eliminating all the foreign-country variables and showing that the same pattern exists within the US.

  7. It would be interesting to see if the gun control is the cause or the result of a previous high homicide rate. Since what you show seems to be right, there is a catch. Maybe you should see the historical homicide rate and compare it to the implementations of the policies and see the results, at least in a state like Washington. If it were like this, you would be wrong.

  8. If I am reading this correctly, what you are implying is that the presence of guns don’t make any or a negligible amount of difference to numbers of homicide?

    And thus implying, is a non issue.

    But then again, also that an increase in gun ownership also leads to a decrease in homicides? (as again, a negligible amount)

  9. Hey ZachMo, you lured me in with the Balmer piece and then I discovered the 3 part Gun Control blog. Really great writing and a subject that I have been on the fence about for a while. I think there is too much fearmongering on both sides of this issue and I haven’t read anything quite as pragmatic or as thoughtful as what you have written here. In part one, you do an excellent job of framing the discussion around homicide vs gun violence, it is definitely the right way of thinking – guns are just a means to an end. Is it a faster, easier end? Probably, but would the homicide rate be lower, doubtful – and the data that you use says no. In your analysis, you glaze over an observation that I think has far more meaning than gun posession, that is wealth and education. And really, it’s the disproportion of wealth and education that caught my attention. I don’t have the facts or the data, but I have some experience living in one of the states that rank highest among homocides, Louisiana. What I know about Louisiana is that segregation (and race discrimination) still exists, there is a wide gap between the upper class and lower class with a very small middle class and education and healthcare are ranked near or at the bottom. Interesting that the other top 5 states share many of the same socioeconomic traits with Louisiana (Washington, DC, PR, USVI, Alabama & Mississippi). Hmmm, is there a correlation? When you have time I would love to see wealth and education plotted on the same graph. Regardless, it is an interesting problem – i would submit that Western European countries and Canada don’t have the same homicide problems because there is not such a disproportionate share of wealth and higher education is the norm, an effect of a Socialist economy perhaps? IDK, I am no expert, just a hunch.

    In part 2, the Prisoner’s Dilemma explaination is brilliant. It is obvious but never thought of it this way. It will be lost on some, but most rational human beings go through this decision making matrix naturally, they just don’t realize it.

    Again, great work. I will definately reference it when needed for future debates.

  10. All of this information about how people have been killed is good to know and all but where in all this does it tell us how to stop SPORADIC mass shootings using WEAPONS OF MASS destruction?

  11. Intriguing information. I have a few questions:

    1) Did you look at crime rates or violence rates or death rate by households with guns vs households without guns?

    2) Are the gun ownership figures legal gun ownership numbers or do they attempt to estimate illegal gun ownership as well?

    3) In the high homicide/low gun ownership areas, did the method of homicide change or did the ratio of gun homicides to non-gun homicides remain somewhat constant?

    4) Did you look into suicide rates? Successful/unsuccessful suicide rates in high gun ownership and low gun ownership areas?

    Thanks for the articles.

    1. Thanks for reading and for your thoughtful questions. In order:

      1. We do not have any of that data at the household level. The lowest level we have found for all the variables you mention is at the state level. We could do new research at the household level, but that’s not in the budget at present.

      2. Illegal gun ownership is very difficult to measure for a number of reasons, and I do not know of data that estimates it.

      3. We are looking into this question now. Initial analysis shows that non-gun homicide and gun homicide are strongly correlated at the state level. There appear to be common factors underlying both kinds of homicide.

      4. Looking into this now as well. Suicide rates do not appear to be correlated with homicide rates at all.

  12. ” So why analyze only firearm-related murders?”
    In this case because you are trying to make a case that more fire arms mean less murders, This would be helpful fodder for 2A propagandist, but what does this really show? The issue with guns is not that they kill people, but they are designed to easily and effectively. To really make a point, you have to make an apples to apples comparison, not apples to oranges.,

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