Tuesday, January 11, 2011

Brady Campaign Representative Says Numbers, Tells Us Nothing

In the midst of this story about gun sales in the wake of the shootings in Arizona,
Daniel Vise, senior attorney with the Brady Campaign, said Arizona received a score of two out of 100 on the organization’s rating of state gun laws, and that the rate of gun deaths in the state is one and a half times the national average.
First of all, Arizona's rate of murders by firearms (3.04 per 100,000) is almost exactly the same as the rate for the country as a whole (2.98).  Leaving that aside, though, telling us Arizona's Brady Campaign score and its gun death rates without other data is meaningless.  Here's a scatterplot of 2009 firearm murders rates by state and 2009 Brady Campaign scores:

Arizona is the blue dot in the bottom left corner, right on the regression line.  It has the same Brady Campaign score as Louisiana (10.46 firearm murders per 100,000) and Idaho (0.33 firearm murders per 100,000).  It has about the same firearm murder rate as Illinois (Brady Campaign score 29) and California (Brady Campaign score 79---the highest in the nation).  As the equation for the regression line indicates, there is essentially no bivariate relationship between firearm murder rates and Brady Campaign scores.

As I wrote earlier today:
[S]tate-level variance in gun-related crimes is no-doubt tied up with cross-sectional variance in cultural, economic, and demographic factors as well as public policy choices, all of which are temporally related to one another and to changes in national patterns of crime and policy.  Taking any slice of data and trying to sort out the various causal mechanisms at work in the face of terrible problems of measurement, endogeneity, serial correlation, and cross-level effects pushes the limits of meaningful statistical analysis.
Asserting a causal relationship between gun laws and crime on the basis of a single case is silly at best. 

No comments:

Post a Comment