Guns and Murder Internationally

Our opponents at the Brady Campaign, and other places, are fond of saying that the United States would be the safest country in the world if it were true that more guns meant less crime. They often cherry pick data from favored European Countries, and hope no one bothers to look at the whole picture. Thanks to the folks at Lucky Gunner, I decided to take a look. I’m using this data from the Small Arms Survey. We will take the most wealthy countries, which I’ll define as those who have a per capita GDP of $14,000 or more. We will toss out any countries that are very undemocratic (Qatar, Saudi Arabia, Oman), or that have instability problems (Lebanon), figuring these countries don’t have good incentives to report accurate crime statistics. That gets rid of most of the Middle East. We’ll also get rid of very small countries, like Luxembourg and Malta, figuring they are very small, and because I don’t want to compile that much data. We’ll keep the focus on major, reasonably democratic and wealthy countries. I’ll use the murder rate data found here. My raw data can be found here. But I’ll show the chart:

Each dot represents an individual country. See my data if you’re curious about which countries. If you do an r-squared correlation on the data, it does not correlate whatsoever. That means there is absolutely no correlation between the number of guns in civilian hands in any given country and the murder rate. Murder rates and GDP correlate slightly, and gun ownership and GDP per capita don’t correlate all that much either. Some might complain that I included African countries, which were above the cutoff I chose. That does not improve the correlation in the slightest if you get rid of them. Some might argue I kept the cutoff too low. If you draw the line above Russia, it improves the correlation slightly, but still no real correlation. If you draw the line at Hungary, you get some correlation. In order to get a strong correlation, you have to pretend that Hungary, Poland, the Czech Republic, and Portugal aren’t real counties. In order to get strong correlation, you have to cut the number of countries down to the point where the US can give you the correlation you’re looking for. If you just take the top European Countries (GDP/capita > 28,000) again, there’s no correlation, with only a very slight downward trend.

In short, our opponents can only get correlation by using a very small sample size, so that the United States (which legitimately does have a very high gun ownership, and higher murder rate than most other very wealthy countries) can drive a correlation. If you use more objective criteria, you don’t get what they want.

8 thoughts on “Guns and Murder Internationally”

  1. Also, Sebastian, there is really no basis for equating figures from the US with those from western Europe. IIRC, figures in the US represent all people initially believed to have been killed by someone else, whether or not the death was murder, manslaughter, accidental homicide, or justifiable homicide, or just a mistake by law enforcement at the scene. Figures in western Europe represent actual murder convictions.

  2. One thing to watch for is whether such correlations (and I suspect they should be done as ranks rather than parametrically) use “murder rate” or whether they use “gun death rate.” For the most part, what I see coming out of gun control advocates is the use of death rate (includes accidents and suicides) and what I see coming out of gun rights advocates is the use of murder or homicide rate.

    And generally, any such correlations seem to be weak (if present).

  3. It looks to me (from my little android screen) that the only countries with really high murder rates are those with less than 50 firearms per 100 in population.

  4. kenno271 hit my main point: international murder/crime rate comparisons are meaningless, unless the countries in question use the same criteria both for defining a crime and including it in statistical reports. IIRC, Formerly Great Britain, for example, only includes crimes where there has been a conviction, whereas the US includes all crimes that have been reported.

    The GB figures will be artificially low when compared to the US, simply because the US includes ongoing investigations and prosecutions, and even cold cases, where GB only counts completed cases with successful convictions.

    There are other questions, too. Does the other country define murder the same way the US does, for instance. Does the other country investigate and prosecute the same murders the US would (i.e., Iran and murders of gays, for instance, or a hardline sharia law country where a man killing a woman is not a crime)?

    The comparisons are meaningless.

  5. Very impressive analysis and comments. I observe also that the United States is huge and varied. The murder rate in Chicago (far greater than national average) varies sharply from the murder rate in Peoria (less than national average) although the two cities share the same state laws. Rockford just west of Chicago does not even begin to approach Chicago. My own city is slightly higher than the national average—but far less than Chicago’s— while smaller Columbia, TN, exhibits a negligible murder rate.
    It is statistical nonsense to attempt to draw inferences comparing any other country to the United States. We are, after all, a collection of sovereign states, with varying laws and differing subcultures, not some homogeneous whole.
    Finally, I have to add that the gun controllers’ statistics completely ignore the millions of disarmed Europeans killed by other Europeans during the last 100 years. Those people somehow don’t count in measuring “violence.”
    Thanks for working up the numbers, Sebastian.

  6. The problem with the small arms survey data is quite simple. They are from “official sources,” and are heavily edited to make the regime look better. England calls a murder “a matter of interest to the police” until someone has been convicted and his appeals are exhausted. Poland tabulates and reports only a fraction of the crimes known to the police. And so on and so forth and so fifth.

    Using data that is even in the ballpark obtains a vastly different result. One in which the incidence of violent crime is inversely proportional to the number of guns in private citizens hands.


Comments are closed.