Saturday, June 30, 2012

Guns and homicide

I am currently putting together a large cross-national data set and will report interesting correlations. Let me begin with the relationship between civilian gun availability and homicide. I calculated the Pearson correlation between the two rates using data from the Small Arms Survey and WHO's recent study of homicide. For a sample of 181 countries, the correlation is -.25. That means that countries with more guns tend to have less lethal violence.

Gun ownership is strongly related to per capita GDP (.48). Countries with many guns like the United States (which is the world's leader with 88 firearms per 100 people) are typically wealthy. We Americans usually think of Europe as being practically gun free, but they have significantly higher levels than many poor countries. And two of the poorest regions are also the most violent--Latin America and especailly Sub-Saharan Africa. An African country might have less than one gun per 100 people, but that doesn't stop it from having a homicide rate many times the global average. I might get different results when I conduct a multivariate analysis--gun availability might lead to more violence when GDP is controlled, for example--but guns are not a serious enough factor to overwhelm more important causes of homicide.  


  1. I did did some correlations on race and crime a while back. They may provide your readers with an appetizer to tide them over until Chef Guhname's main course.

    My data in a nutshell: the correlation between homicide and Blackness (fraction of the population identifying as black or African-American in census data) are higher than the correlation between homicide and Black & Hispanic fraction. This is not a surprise; replacing Black in an area with Hispanics can reduce (local) crime rates.

  2. How about a tasteless update to a well-known expression?

    "Guns don't kill people, NAMs kill people"

  3. Presumably you are aware of the existence of "" and that no matter how many correlations you come up with that point in a particular direction you're going to get "correlation doesn't imply causation" from the sophomoric and "ecological correlations are invalid" from the sophisticated.


Libertarians do not have high IQs

I used General Social Survey data in an attempt to identify libertarians (sample size is around 1,000).  If respondents answered yes to &quo...