Plum Crazy

A couple of days ago, Steven Landsburg of Slate approvingly wrote about a study that purported to show that internet porn was a direct cause of a reduction in rape. I first read about it on Feministe. I read the study itself and did a takedown of its methodology in the comments at Feministe. To quote myself:

Ah feh, the entire methodology of the study is flawed. As Kendall himself notes, there is sampling bias in reports of rape, such that stranger rapes are more likely to be reported than acquaintance rapes. As such, you cannot possibly draw any conclusions about the impact of internet access on acquaintance rapes, which are the most common type of rape in the first place.

Second, he couldn't pull apart the use of the internet for porn from the use of the internet for non-porn. Meaning, Kendall cannot really prove anything about the use of pornography on the internet vis a vis even stranger rape.

Third, since they don't have offender profiles, as Kendall himself admits, there is no way to know how much pornography the rapists consumed prior to committing rapes vs. how much non-rapists consume. So he cannot judge the impact of pornography on the rapists themselves.

Fourth, while he does factor in some demographic variables, he doesn't take into account any other changes in societal values over the last x number of years that might explain a reduction in rape vs. other violent crimes.

Fifth, with respect to the lack of statistically significant correlation between internet use and other violent crimes, without examining the factors that contribute to rape vs. other violent crimes, you cannot rule out other explanations for that phenomenon. A lack of correlation to one type of crime does not, as such, prove a correlation to another.

In short, the study contains no proof that internet porn reduces rape. For all we know, all those 15-19 year-olds were at home playing games on the internet, and, as such, weren’t out committing rape. Or they were raping their girlfriends instead, who didn’t report the crimes. Or they weren't committing rape for reasons that have nothing to do with internet access but with societal variables he didn't factor into his analysis. Or all of the above.

Now, however, Melissa Lafsky of Salon has picked up on the Slate article and approvingly quoted one of the more annoying claims made in it (which is saying something, because all the claims made in the article are annoying and completely without actual evidence):

Still, as Landsburg points out, how do we know the Web use-rape link is connected to porn? Maybe online gaming, sports chat rooms or other male-dominated sites are in fact catching the attention of potential rapists. Kendall responds to this argument by offering murder rates; while the data consistently ties increased Internet use to decreased rapes, no such correlation exists for homicides. "It's hard to see how Wikipedia can deter rape without deterring other violent crimes at the same time," Landsburg noted. "On the other hand, it's easy to imagine how porn might serve as a substitute for rape." Coming as no surprise, the effects are strongest among boys ages 14 to 19, a demographic that Kendall (and common sense) identify as the group that relies most heavily on the Internet for porn access.

For the statistically challenged, let me explain why this is an annoying argument, even if it seems to make sense to some. What Kendall, Landsburg, and now Lafsky have done is decided in advance that they believe that porn is a substitute for rape and chosen not to look any further for an explanation for the lack of correlation between internet access and homicide rates. Let me break it down further.

Kendall has proven a statistically significant negative correlation between internet access and rape rates. That much is true. He has not proven any statistically significant negative correlation between viewing internet porn and rape rates. He admits he wasn't able to break down the internet access between porn and non-porn usage. Instead he is relying on indirect methods to come up with his "evidence" for this supposed negative correlation (I'm not saying it does not exist; just that no one has actually been able to prove it does). The first thing is (my words) the "Well, lots of people use the internet for porn, so if there is a correlation between internet use and rape, then it likely is related to porn." Maybe yes, maybe no. It's not an unreasonable hypothesis, but it is just a hypothesis without any actual evidence. In order to make that determination, you would have to study it.

The second indirect method is (again, my words) "Hey, look, no correlation between internet access and other violent crimes, so what can I personally think of that would make that difference? Ooh, look, porn!" Again, there is absolutely no evidence that this is true. Kendall's inability to think of any other distinguishing factors does not mean those distinguishing factors don't exist. He didn't even bother to look at any other factors that distinguish between rape and homicide. Under what circumstances do most homicides occur, for example? I happen to think this is especially relevant given Kendall himself admits he wasn't able to study the impact of internet access on acquaintance rape, meaning that his negative correlation only has been proven for stranger rape (the least common sort of rape). Since most homicides are also committed by friends or family, you would think there might be a connection there worth studying. It's a lot harder to under-report homicide even when it's committed by a friend or family member. Does this mean that many homicides occur at home, such that using the internet wouldn't remove you from a situation in which you were likely to do something that is done outside the home? I don't know, but it's certainly worth considering as a possible explanation for the lack of correlation between internet access and homicide.

Lastly, I would remind them all that correlation does not equal causation. Kendall was only able to prove a correlation. He was not able to prove causation.