Another Day, Another Gun Study
This one, from The American Journal of Medicine, finds that countries with higher gun ownership (meaning total guns per 100 people) also have higher rates of firearm death, even after accounting for differences in rates of mental illness (meaning major depressive disorders). Meanwhile, it finds no relationship between gun ownership and the overall crime rate.
The study is not particularly helpful for policymakers. To begin with, it's not so much a "study" as "something a nerdy person does in an afternoon while he's bored." (I'm something of an expert on that.) A whopping four variables are involved: gun ownership, mental illness rate, overall crime rate, and rate of firearm death. I've said before that we shouldn't be bowled over by incredibly complicated attempts to disentangle correlation from causation, but that doesn't mean simplistic work is actually superior. Here there are no attempts to control for historical or demographic factors, aside from choosing a sample of "developed" countries -- the U.S. and South Africa, with their complicated and violent racial histories, are lumped right in with Switzerland and Sweden.
Further, even given its simplicity, the methodology is highly questionable. Like the study I wrote about last weekend, it focuses on firearm deaths -- raising the question of whether people who don't have access to guns simply switch to other weapons, instead of refraining from killing altogether. It also includes suicides, which raises thorny questions about the extent to which governments should protect us from ourselves. And as the authors concede, the results don't show that guns cause firearm death; it could be that people who live in violent countries feel the need to arm themselves, or that underlying tendencies explain both phenomena.
Fortunately, it's pretty easy to redo the analysis using homicide rates rather than gun deaths, which at least addresses some of these issues. What follows is no more of a debate-ender than the original study, but it's a different -- and in my view, more useful for crime-fighters -- way of analyzing the data. Warning: wonky.
For starters, here's a basic scatterplot of gun ownership (as reported by the authors of the paper) and homicide rates (as reported here) in the countries studied:
Obviously, there are two outliers here: South Africa (top left) has an astronomical homicide rate, and the U.S. (lower right) has an incredibly high gun ownership rate -- and the second-highest homicide rate to boot. Unsurprisingly, these data don't yield any statistically significant results to explain homicide -- whether I'm using gun ownership alone, mental illness alone, or both together in a multiple regression.
So, here's a new chart with the U.S. and South Africa kicked out:
Here's where things get interesting. In trying to explain homicide, guns still don't give a significant result, but mental illness by itself explains about 20 percent of the variance in homicides (i.e. the r-squared is about .2) and the result is statistically significant (P-value < .05).
And when guns and mental illness are included together, they explain about one-third of the variation. Within this multiple regression, guns fall just short of statistical significance (P = .054), but the coefficient is negative. If the result were significant, it would suggest that, in a population of 100,000 people, every 1,000 guns are associated with 0.02 fewer homicides each year.
So, these data don't suggest that guns make much of a difference. And the fact that the results change so dramatically when one replaces firearm deaths with overall homicides shows how subjective and sensitive this kind of statistical work can be.
Tweet me if you'd like me to send you the spreadsheet I used. I was a journalism major, so any thoughts on my methods are welcome. And you can see some more analysis of guns and homicide on RealClearScience here.
Robert VerBruggen is editor of RealClearPolicy. Twitter: @RAVerBruggen