You probably learned somewhere that correlation does not equal causality. That is to say, just because two variables rise and fall together doesn’t mean that one causes the other. One of the examples is that ice cream cone sales rise and fall with rape rates. It isn’t that ice cream cones cause rape; it is that high temperatures increase demand for ice cream and cause many people to leave windows open to get a cool breeze to enter; instead, a rapist takes advantage of the unlocked window.
In that case, there is a common factor driving both variables: high temperatures. Sometimes the correlation is just a coincidence. Statisticians have ways to figure out whether correlations are coincidental or not — but even then, the best that you can say for any particular correlation is that it is unlikely that the correlation is coincidence.
The bigger problem is when you have a correlation that is not a coincidence, but the question becomes, “Which is the cause? Which is the result?” There’s a strong correlation between smoking and lung cancer, but longitudinal studies have pretty well established that smoking comes before the lung cancer. That’s a pretty strong clue that smoking causes lung cancer — not that people who get lung cancer take up smoking because they are so upset about being sick.
Similarly, if you surveyed American homes, you would find a strong correlation between the presence of legally prescribed syringes and at least one diabetic in the home. If you didn’t know that diabetics use syringes to inject insulin, you might wonder: Do syringes cause diabetes? Or does diabetes cause syringes to appear? This problem is known as determining the direction of causality.
Unlike “chicken or the egg,” longitudinal studies give you a realistic chance of figuring out, when you have two correlated variables, which is the cause and which is the symptom. All this discussion is to introduce one of those reminders that correlation is not enough. There’s a new study coming out, just in time to try and influence the Supreme Court in the upcoming challenge to Chicago’s handgun freeze law.
This new study, to be published in the November issue of American Journal of Public Health, claims that Philadelphians in possession of guns were 4.5 times more likely to be shot than Philadelphians who didn’t have a gun. Correlation: check. Therefore, “suggestions to the contrary, especially for urban residents who may see gun possession as a defense against a dangerous environment, should be discussed and thoughtfully reconsidered.”
What’s the direction of causality here? Are you more likely to be shot because you own a gun? Does anyone seriously believe that buying a gun attracts criminal attackers? Or do people buy guns because they perceive that they are in danger of being attacked? If it is the latter, they guessed correctly. Someone shot them.
Now, it is possible that some of these Philadelphians bought guns and a criminal shot them because they resisted a criminal attack using that gun. But if so, this would be a powerful piece of evidence that guns are dangerous to victims. It does not appear that the study checked to see if the victims had tried to defend themselves. This study was based on 677 individual victims; how hard would it have been to find out if the victims had tried to use their guns in self-defense?
If you know much about Philadelphia, you already have seen the other problem with this study: a lot of the victims of violent crime in big cities are gang members shooting other gang members. It may well be that gang members buy guns with the expectation that members of a rival gang are going to try and shoot them — and they would be completely correct about that. Wouldn’t it have been so much more interesting and useful if this study had checked the criminal histories of the 677 victims and compared them to the criminal histories of the control group? In Milwaukee, for example, not only did 86% of those arrested for homicide have previous arrests, but so had 75% of the victims (see p. 4 of the report). Of course, that might have exposed the real risk factor: being a gang member dramatically increases your risk of being shot.
Many years ago, I was quite amused by one of the few really clever and thoughtful bumper stickers that I have ever seen. It managed to teach this problem of the direction of causality in a simple phrase: “Guns cause crime the way flies cause garbage.” The presence of garbage attracts flies; high crime rates cause decent people to buy guns. Prohibiting guns will no more prevent crime than spraying for flies will make your garbage disappear.