Friday, January 02, 2004

Random Musing on Guns and Crime

Tim Lambert's always your one stop shop for all things John Lott, and for all things guns&crime. And, this may have been something he's addressed one or many times but I don't quite remember and it's been floating around my brain for awhile.

Anyway, John Lott claimed to show that "More Guns=Less Crime." As has been explained elsewhere, even if Lott's results were to be believed, they don't quite show this link anyway...

But, what defenders and attackers of Lott and/or his central thesis have mostly agreed with (at least the intelligent reasonable ones) is the fact that the current state of the literature seems to show that there isn't much of a relationship betwen guns and crime at all. It could be a bit positive, or a bit negative, or statistically insignificant. So, at first pass this admittedly is more of an anti-gun control result. If more guns don't lead to more crime, then there isn't really any compelling state interest in passing gun control measures, so...

What I do know is Lambert and others have pointed out is that there are serious problems with using the county level data used in the analysis. The specific problem is that the data is just bad. It's measured poorly, or contains what is called "measurement error." In other words, when the true underlying number is, say "100," there's a pretty big probability that the recorded data is "20" or "500." It's just noisy.

The thing about noisy data is that what it tends to do when one tries to estimate relationships between variables increase the likelihood that no relationship is found. Noisy data doesn't just reduce the precision of the estimate, it systematically biases the estimated relationship to 0. In other words, the estimate doesn't suggest that there is no relationship with any confidence- it's just a result of your data being crap.

The point is that if that's where the results are coming from, then it's more correct to say that "current research has lacked the appropriate data to adequately address the question" rather than "current research hasn't shown there to be a relationship either way."

Anyway, this may have been covered elsewhere, and it isn't meant to be a serious attack on Lott's, or anyone else's work, just an attack on the "standard interpretation" which seems to be floating around...

Also, I don't care much - guns really aren't my issue either way.