The heart of this paper is a survey presented to a sample of 193 college males about pedophilic attraction, fantasies, and urges. A surprisingly large number (at least 21%) reported that they had some sexual interest in children, suggesting that the problem of pedophilia is a lot more common than is believed. About one percent indicated that the fantasies occurred "very often," or that it was "very likely" that they would have sex with a child if they knew they wouldn't be punished or found out.
It is important to interpret this study carefully. Some questions worth asking are: Are the survey questions well designed? Are the survey results reliable? Do they indicate a high incidence of pedophilia, or perhaps of something else? Is there sample bias? These questions must be addressed if we are going to conclude that the study says what it looks like it says.
So, are the survey questions well designed? The questions were a 1 to 7 scale of agreement with the statement "little children sometimes attract me sexually," a 1 to 5 scale (never to very often) on how often they had sexual fantasies about "having sex with a child" and how often they masturbated to these fantasies, and a 1 to 5 scale on how likely they would be to have "sex with a child" if they knew for sure they wouldn't be punished, and that no one would find out. These seem like pretty good questions, but certainly not ones that lead directly to a diagnosis of pedophilia. They do not, among other things, ask whether the attraction to children is primary, or whether the fantasies are insistent.
Are the survey results accurate? There are two types of errors here: underreporting and overreporting. It is reasonable to expect that there is some underreporting, because the subject is in denial, or does not trust the confidentiality of the test procedures, or some other reason. So, if anything, the real figures are higher than those reported. Overreporting might occur because of a deliberate attempt to throw off the results. This is particularly a problem because some of the percentages are so small. It would only take one or two bad answers to throw the numbers into question. On the other hand, there is no real motive to do so, and it seems that most people are straightforward on well designed self-report surveys. So, I would say that overreporting is a significantly smaller problem that underreporting.
Do the results indicate a high incidence of pedophilia, or of something else? The study has two problems here. First, simple sexual attraction to children does not equal pedophilia. A person only has pedophilia when that attraction is central, or when the urges are strong enough to lead to actual sexutal behavior with children. Another problem is with the last question (hypothetical likelihood of sex with a child). There are other reasons besides pedophilia why a person might answer yes, as well as actually molest a child. Examples would be sociopathy, psychopathy, impaired judgement because of alcohol or drugs, and for power dominance. A somewhat more interesting question might have been, "would you have sex with a child if it were legal?"
Finally, what about sample bias? Here is the real strength of this study. It was administered to a group of university males. Most similar studies deal with a "forensic sample," which is essentially a group of convicted child molesters. As a pedophile who is not a child molester, much less a convicted one, I can say with assurance that there is a lot of difference between these groups.
The paper also contained some results correlating the answers to these four questions with other personality factors. However, I didn't find this part particularly interesting.
In summary, the message of this paper is that there are a lot of people out there with some sexual attraction to children.