This article in today's New York Times discusses an unpublished study by the Bureau of Prisons (BOP) suggesting a strong link between viewing child pornography and sexual abuse of children. Specifically, the study reports that 85% of men who admitted to looking at child pornography also admitted to sexually abusing children. There's a big problem though with the study: The sample consisted only of men serving federal prison sentences at the federal prison in Butner, NC.
A crucial step in doing any prevalence study is obtaining a representative sample of the population of interest. Here that population is presumably all people who download child pornography. By definition, a sample that is restricted to inmates of a federal prison is not representative. Think of a comparison study: What if we wanted to know what factors are associated with another illegal behavior, say, using some form of illicit drug. If we only interviewed folks in our county jails we would likely find that being an ethnic minority and poor are risk factors for drug use. But, of course, such findings would be influenced by the fact that minorities and the poor are overrepresented (in a statistical sense) in the criminal justice system. Such a study might tell us a bit about what risk factors are associated with getting caught using illegal drugs and serving time -- but it would not tell us what are the risk factors for using drugs. Lot's of folks use drugs and are never caught.
That's the problem with the BOP study: it neglects to account for all of the people viewing child pornography who are not getting caught and serving time. Such a study would be very difficult to conduct, of course, since obtaining such a sample would be nearly impossible. Nevertheless, one cannot claim a link between these two behaviors -- child pornography and child sexual abuse -- otherwise (no matter how intuitive the finding may be). Judging by the variability in terms of race, age, occupation, and other demographic characteristics of people arrested for downloading child pornography one could argue that this behavior appears somewhat normally distributed in the adult general population. (NB: By normally distributed I in no way mean that this behavior is "normal" -- lots of phenomena appear normally distributed in the general population but are not "normal". Just think of obesity). If child pornography is anything like illicit drugs many more people are engaging in the behavior and not getting caught than those who do it and get caught. What this says about our culture is left to the pundits.
Another fact of the BOP study that deserves caution is the fact that the manuscript was withdrawn after it was accepted for publication in a peer reviewed journal. This is quite unusual and the question remains whether the study will be made public. Without outside eyes looking at the study, we have no way of understanding its strengths of limitations in detail. Hopefully, the BOP will release the study no matter how flawed it is since any data on this morally reprehensible problem is better than our current knowledge-base which is almost zero right now.