Number 2 Pencil

Lies, Damn Lies, and anti-gun Social Science

This article by legendary law blogger Eugene Volokh isn't related to testing, but it is a great example of how politically-correct ideology can be used to twist around survey data. In this study, the authors found that gun ownership is correlated with likelihood of dying by gun. Correlation does not imply causation, but the study's authors - and many journalists - consider this to be evident than owning a gun is, in itself, going to increase your risk of dying by gun. Therefore, you should protect yourself by not owning a gun.

Problem is, as Eugene points out, that a third factor is probably confounding these other two. In Statistics 101, I teach that, when two variables A and B are correlated, three situations are possible. A may be causing B, B may be causing A - or a third variable, C, could be causing both A and B to co-occur.

Anti-gun activists would have you believe that owning a gun (A) causes you to die by gun (B). As Eugene points out, multiple studies have shown that large percentages of gun-homicide victims, and smaller percentage of gun suicide victims, have arrest records or other indicators of criminal behavior. In other words, being a criminal - C - is most likely the cause of A and B. The result of ignoring this third variable means these loudly-trumpeted results are most likely meaningless (yet still unfortunately influential in the current gun debate).

A study on gun violence (or any violent behavior, for that matter) which doesn't attempt to identify which of its subjects are criminals is shoddy science. So why would a social science researcher ignore this variable? My theory is that some social science variables are invisible to researchers because of their political incorrectness. I suppose it's considered racist to ask someone if they're a gang member, to ask about past criminal behavior, or to report that gang members (who are overwhelmingly non-white) are more likely both to own guns and die from gunshot wounds. Better to present the data without that and let the conclusion be that

Now that I think about it, maybe this topic is related to testing. After all, we often see statistics which show that minority students do worse on tests. The politically-correct response is to abolish the tests ("get rid of guns"). But it's politically incorrect to point out that

Posted by kswygert at June 17, 2003 04:03 PM