Vox’s misleading gun death graph

In the wake of the mass shooting tragedy in Los Vegas, Vox posted an analysis showing that more guns lead to more gun deaths:

I was immediately suspicious because I’m living in Guatemala and I am constantly being reminded that Latin America has very high homicide rates, and although I see guards with massive guns all over the place, Guatemala has quite low gun ownership rates. Whereas I agree that a lot of data supports the thesis that too many guns contributes to more gun deaths, gun availability isn’t the only thing that leads to gun deaths. Other factors are often more important. Otherwise the US would have the highest homicide rate in the world because the US has by far the most guns per capita.

I can’t figure out any simple rule that explains how they picked the nations they included in the above graph. It isn’t the list of OECD nations. It is pretty close to a graph that excludes all poor nations, but if so, then it should really include Panama and Uruguay which are richer than Argentina (which is included). But that wouldn’t fit the thesis that guns cause more deaths as nicely. The graph originally came from Tewksbury Lab, but they have taken down all reference to it, so I can’t see where they got their data, but it looks like they used data from the 2011 “Estimating Civilian Owned Firearms” by the Small Arms Survey. Nationmaster seems to have taken the Small Arms Survey data and added some estimates for other countries to come up with a more comprehensive data set of gun ownership which I combined with World Bank homicide data to produce this graph:

In fact, it appears that gun availability is actually somewhat correlated with fewer gun deaths. I tried excluding all nations poorer than Spain, but it still doesn’t give much support to the thesis that more guns cause more homicides because there is little correlation again:

Basically, there are two outliers, the US and Qatar, and they are so different from all other rich nations, that they should both probably be excluded. When you do that, again there is almost zero correlation between guns and homicides. That doesn’t mean that guns don’t contribute to the high homicide rate in the US, but it just isn’t a simple matter. Lots of other factors are also important such as environmental lead, criminal justice, inequality, etc.

Below is another graph using data from Wikipedia’s entry about firearm-related deaths which looks more reliable than NationMaster because Wikipedia cites where all their data comes from, and it is more up to date. Still, it tells a similar story.

Guns don’t make that much difference because there are wide differences in the number of homicides per gun:

That is using the Nationmaster & World Bank data again. Each gun kills over 2,000 times more people in Tunisia than in Switzerland. Here is a close up of the left tail of the above graph to show the nations that have low levels of homicides per 1,000 guns like in the USA.  On average, only about 1 in 20,000 guns is used to kill people in America.  That is just a few bad apples.

UPDATE: I found the original graph on the Wayback Machine and the author’s methodology looks sound, but he says he got a lot of criticism about it so maybe he took it down to avoid controversy (UPDATE AGAIN: as he explains in comments below, he unfortunately just happened to take down his old website, so it was hard to find his original work.)

Also, the following chart from Mother Jones is a pretty convincing simple demonstration that more guns are linked to more deaths since US states are all relatively similar in income, culture, demography, criminal justice, and all the other important factors that influence homicides, so the effects of differences in gun ownership can pop out more clearly:

Gun ownership tightly correlates with gun violence.

So, reducing gun access undoubtedly saves a few lives (particularly gun suicides and gun accidents), but lots of other factors are more important for saving lives than guns.  

It may be harder to commit mass murder without guns, but extremists have managed to do it with airplanes, trucks, and explosives too, so gun restrictions help, but they don’t eliminate the problem and fear of mass murderers is completely irrational anyhow.  

Posted in Health, Violence & Peace
8 comments on “Vox’s misleading gun death graph
  1. […] passionate about gun control because although it generally reduces homicides and suicides, other factors are even more important for reducing homicides. Furthermore, there are a lot of other public health issues that are more important for saving […]

    • Thanks George, I updated the text and included your link at the bottom of the original post. I agree that the research overwhelmingly shows that gun control generally saves lives, but after living in Latin America, I can’t say that the homicides that gun control can prevent are anywhere near the biggest public health problem that any country faces and although the gun fanatics are simply wrong to think that greater gun access prevents violence, both sides are actually irrational in overly focusing on their excessive fear of violence because lots of other issues are more important than guns in determining public health or even for determining levels of violence, so we should all take a deep breath and try to focus more on issues where we can more easily find common ground to improve society. Let gun fanatics cling to their irrational fears and instead of trying to get them to be more rational, lets work with them on issues that we can achieve more on that don’t trigger their irrational fears.

  2. […] Once again, Vox Media gets the importance of gun violence all wrong. Brian Resneck says, “America’s teens are [right to be] extremely stressed out about school shootings”. No, no, no. Schools are extremely safe places and mass shooters are extremely rare even in America. America has many more gun massacres than any other country (except countries that are immolated in war), so gun massacres are an odd problem for a rich country like America to have, but it is still a very minor problem that is far from the top of the long list of American public health problems. And mass shootings in schools are especially rare. It is really irrational to worry about them. Driving to school is much more dangerous than mass shooters in school. Suicide is more dangerous than murder for young people. Climate change will probably have a bigger impact on youth than mass shooters. But young people are more worried about mass shootings than about suicide or climate change: […]

  3. Josh Tewksbury says:

    Just found this great thread – thanks for digging up the original posts. To be clear, the data got pulled down because I changed jobs, and I don’t currently have a good place for these kind of posts. Kudos to the way back machine! Here is the link to the second post


  4. Thanks Josh. I updated the links to your original pages in the addendum at the bottom. I thought I had included that second link, but I messed that up, so thanks for the correction.

  5. anon says:

    This is an old comment, but I do believe you have missed quite a few points in your analysis and I would be remiss not to mention them. Vox’s graph is a graph of firearm deaths versus number of firearms, whereas the first two graphs you produced are those of all homicides versus firearms. The statement isn’t so much that more firearms are correlated with more deaths, it is that more firearms is correlated with more gun-related deaths.

    The total homicide rate is thus:

    Homicide rate = (Homicide component correlated with gun ownership) + (Homicide component uncorrelated with gun ownership)

    For a country like the US, where 2/3 of homicides are firearm-related, you can definitely reduce the total homicide rate by a huge margin by reducing the number of guns. It would be hard to propose that all or even the majority of these deaths would have occurred if firearms were not present. It is simply much more difficult to kill someone without a gun than with.

    On a side note, I am interested why you didn’t try removing the poor countries from the wikipedia data – perhaps because it does show a correlation (since the wikipedia data uses gun deaths)? With the poor countries, this is likely confounded by the fact that a lot of perpetrators hold illegal guns, whereas control of illegal guns is much tighter in richer countries.

    • You are right. I agree with you about this. I just think it is politically impossible to reduce gun prevalence in America and that we can also do lots of other things to reduce our ridiculously high homicide rate. Since voters seem to love having more guns per capita than any other nation on earth, we might want to change other parts of our national culture to make our guns less lethal like increasing the tax on alcohol. Alcohol is a bigger public health problem than guns overall and it is less politically controversial so it seems like a place where we could make some progress towards improving public health.
      I didn’t remove poor nations mainly because that would be removing information. I agree that I don’t normally want to compare the US’s challenges with those of poor nations, but I think it is useful to see that gun prevalence isn’t the ONLY thing that predicts high homicide rates. Similarly, in the US time series, homicide rates increased nearly 3X peaking in the 1990s and then dropped back again. Gun prevalence showed no correlation with that wave of homicides, so something else has to explain most homicides in the 1990s. I think it was mostly lead exposure which again has some similar effects upon behavior as alcohol.
      I agree with your concept that part of the homicide rate is caused by gun ownership, but I don’t think the formula is additively separable as you wrote;
      Homicide rate = (Homicide component correlated with gun ownership) + (Homicide component uncorrelated with gun ownership)
      There is also a multiplicative aspect of things like lead exposure and alcohol which make gun ownership more dangerous. I think it is more like this:
      Homicide rate = (Homicide component correlated with gun ownership)alcohol(lead exposure) + (Homicide component uncorrelated with gun ownership)
      Thanks again for the comments and insight.

Leave a Comment

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 )

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.

Enter your email address to follow this blog and receive notifications of new posts by email.

Join 96 other subscribers
Blog Archive