The Internet has been tarred and feathered as a haven for sexual predators and pedophiles since its inception; while I don’t mean to dismiss the concern entirely, I’m often troubled by the way that reports of this lurid terror are regularly magnified, overemphasized, and used as justification for regulations of the Internet that go far beyond the scope of the threat.

The latest version of this worry has emerged around social networking sites, especially MySpace, after some cases began to emerge where MySpace played some role in the criminal solicitation of a minor. MySpace has been under pressure recently to develop ways of patrolling its massive site to make it safer for kids; The Associated Press reported today that MySpace has removed 29,000 profiles whose users matched registered sex offenders. North Carolina attorney general Roy Cooper is calling for legislation that would require adults to prove their identity before being allowed to create a profile, and require kids wanting to set up profiles to have their parents prove their identity. (This would likely mean entering a credit card number, though there are other ways. Requiring users to establish identity would make it easier for MySpace to compare their users to the sex offender registry; currently you do not need to indicate your identity to establish a profile, which means MySpace has not necessarily identified all registered sex offenders with profiles.) As Farhad Manjoo notes at Machinist, one of the problems of such a rule is how to define “social networking site” — would it apply to YouTube? del.icio.us? Twitter?

29,000 sounds like a lot, and part of the reason this is news is because MySpace had announced back in May that it had found and removed 7,000 registered sex offenders from its site; the revelation that the number is much larger adds an alarmist ring to the issue. But this quantification of the problem lends a great deal of power to these fears, and is often the reason they can be exploited for political gain. It made me want, despite my skepticism of quanititative evidence more generally, to put these numbers into some perspective. But I find that doing so also reveals how difficult it is to reach reliable numbers. Here goes…

According to a May 2007 report from the National Center for Missing & Exploited Children, there are 602,139 registered sex offenders in the U.S. This number has likely changed since then, but presumably not by an order of magnitude, so let’s use it. The AP report is not explicit about whether MySpace is comparing their users against only this U.S. registry, or if there are also international ones they have access to. So let’s make the assumption that they’re only using the U.S. registry. It’s also not clear from the report whether the 29,000 includes the 7,000 deleted in May, or if they located 29,000 more. If it’s 29,000 total, this means that 4.81% of all U.S. registered sex offenders had MySpace profiles that were located and deleted; if it’s 36,000 total, then that represents 5.98% of the registered sex offenders in the U.S. This feels, at least to me, like a pretty big number; if it was discovered that 6% of registered sex offenders were regularly attending Little League games (maybe they are), would we want to intervene? moreover, this doesn’t tell us how many MySpace user profiles of registered sex offenders did not include their real names and therefore have not been located and deleted — I imagine, if you’re a registered sex offender, that you wouldn’t be too inclined to provide your real name, though I don’t know to what extent there are other services MySpace offers that would require a credit card, making it more likely that real names end up being involved.

The AP report also notes that MySpace has over 180 million user profiles. This number is only an estimate of course, mostly because new profiles are being added all the time. But if we use this estimate, this means that the 29,000 or 36,000 profiles deleted represented approximately 0.016% or 0.02% of all MySpace profiles. These numbers feel small, to me at least. It is important to remember that we can’t necessary use them to stand for a percentage of all MySpace users, just a percentage of profiles; a single user may have multiple profiles. We could take this to suggest that, while there may be more profiles of registered sex offenders, the 29,000/36,000 at least represents that many people; maybe some of the profiles they have yet to discover are these same people; we might also take it to mean that some of these 29,000/36,000 may be multiple profiles of the same person, meaning there were fewer users identified as sex offenders (though if this were the case, to a significant degree, it’s likely that MySpace would make it known in order to downplay the concern.) Also, profiles remain even if their users logged on once and then never came back; the number of active profiles is significantly smaller, though this also applies to the number of sex offender profiles, though perhaps to a different degree.

Of course, we have obviously not accounted for sexual predators who are not registered as sex offenders for whatever reason — they’ve never been caught, they’ve yet to do something criminla but are about to, they were indicted as a minor or under some other circumstance that does not require them to register — or their crimes were in countries that do not maintain a registry, or a registry that MySpace has access to.

Just to add to the tangle, it’s probably important to recognize that part of the fear is based on the assumption that any child who has a MySpace profile is at risk of being accosted by one of these sex offenders. According to research reported in the Journal of Adolescence (blogged about here) a random sample of 2423 profiles by users under 18 years old, 948 were set to “private,” which means they can only be seen by the user’s friends. of the remaining “public” profiles, only 8% revealed their full name on the site, only 1% provided an email address. (Of course, I’m relying here on the abstract of the research and a secondary retelling of the published results of a research project; lots of details about how these numbers were reached are unavailable to me.)

I find it frustrating that the rhetorical power of a number depends in part on its clarity, which means its ability to obscure the complexity of the issue. And developing other numbers to counter it is an exercise in continued frustration. They always depend on unknowns, which must then be ignored, estimated, or explained away — or represented in ways that then undercut the rhetorical power of the number produced. I could say “29,000 sex offenders were removed from MySpace’s 180 million profiles” or I could say “roughly either 29,000 or 36,000 user profiles that likely represented some of the profiles maintained by sex offenders registtered in the U.S. were removed from MySpace’s approximately some number of active user profiles less than 180 million and changing every day.” Both are “true;” in some way the second statement seems like a fairer and more accurate representation, but it’s pretty clear which has more rhetorical oomph, which one will make it into the abstract of the paper, which will get it published, which will get picked up by the press.

And, these numbers always represent another attempt to make compelling numbers for the sake of some argument. A report that says 36,000 sex offenders were removed from MySpace is different than a report that says that 0.016% of MySpace profiles represent sex offenders, which is different than a report that says that 99.984% of MySpace profiles are people without records of sex offenses. Each of these claims is loaded, an “accurate misrepresentation” of the issue.

Update: “The Numbers Guy” at the Wall Street Journal has raised some similar concerns.