The commenters are right that this is not regression. The psychologist interviewed on the Today Show this morning described the method well.
http://today.msnbc.msn.com/id/28415602/

For more info, google "propensity score matching" or "matched sampling".

This is a public health issue, James, not a political issue. The study demonstrates that these abstinence and virginity-pledge programs are actually harmful, in that the teens who participate are more likely to become pregnant and to be infected with an STD (which is what happens when people have sex with multiple partners without condoms or using birth control.) That's a little more important than your nonsensical CYA claim that "the study compares A's to non-A's when, in reality, we're merely comparing A's to B's." That is just meaningless gibberish.

Again, I press this issue because it is you who are mischaracterizing the results of the study. Actually the two reports you linked to did a pretty good job, except they didn't go the step further to observe that they programs are actually harmful, though they hinted at it.

Of course, children of involved parents, who take them to church, are less likely to have sex early.

Evidence, please.

Of course, children of involved parents, who take them to church, are less likely to have sex early. They are also more likely to sign the pledge. Both of these are the results of the parents' attitudes and actions.

It never ceases to amaze me how much time sociologists spend (and taxpayers' money) to prove the obvious. What a racket.

So unless you're positing that, absent the the Pledge, Group A would turn into SCTs, somebody just wasted a bunch of money getting kids to sign Pledges. And I'd characterize that effort as 'ineffective'.

Agreed. My point here is a narrow one: The reporting and commentary out there is based on the misconception that the study compares A's to non-A's when, in reality, we're merely comparing A's to B's.

Again, the study says that among two groups who are likely to delay having sex, one having taken a virginity pledge and one who did not, there was no significant differences in the sexual frequency or practices in the two groups. There was a significant difference in condom use and birth control use, the group who took the virginity pledge having lower usage of both.

I thought both writeups were pretty good, but James Joyner has mischaracterized it, starting with his title.

Group A, which signed the the Pledge.

Group B, which is, in all relevant aspects, the same as Group A except they didn't sign the Pledge.

And then there's everyone else. Let's just call them the Sex Crazed Teens (SCTs).

Altho the article doesn't provide any data to compare group A or B to the SCTs, I think it is reasonable to assume that SCTs have a higher incidence of sexual activity than either Group A or B.

The article seems to indicate that Groups A and B had roughly equivalent outcomes regarding sexual behavior. However, money was spent developing and promoting the Pledge.

So unless you're positing that, absent the the Pledge, Group A would turn into SCTs, somebody just wasted a bunch of money getting kids to sign Pledges. And I'd characterize that effort as 'ineffective'.

"Virginity pledgers and similar non-pledgers don't differ in the rates of vaginal, oral or anal sex or any other sexual behavior," Rosenbaum said. "Strikingly, pledgers are less likely than similar non-pledgers to use condoms and also less likely to use any form of birth control."

"Again, this is good social science if one is attempting to isolate the effects of signing a pledge."

The study is looking at sexual activity and protection used as an outcome among a cohort of kids over a period of time. During that time Group 1 took a virginity pledge, Group 2 did not take a virginity pledge. That is the independent variable. Sexual activity and practices is the dependent variable. The study was stratified by a variety of cofactors, further described in the study.

This method allowed Rosenbaum to compare those who had taken a virginity pledge with similar teens who hadn't taken a pledge but were likely to delay having sex, she said. She added that she didn't include teens who were unlikely to take a pledge.

"But it shouldn’t be interpreted as a simple comparison of pledgers and non-pledgers. Instead, she’s comparing only the kind of people who would be expected to sign pledges to see if actually signing the pledge changes anything."

No. She is comparing kids who were likely to delay having sex.

I am pressing this issue because while rend your hands about how reporters might cover this, you yourself are misinterpeting and misreporting the study as well.

Additionally, when the media covers these studies, they should really include a link to a .pdf of the study or a direct link to the study in the Journal. I am sure it distresses these researchers who spend hundreds of hours designing studies, collecting samples, analyzing the data, and then drawing inferences from the data to have their work subject to all sorts of inaccurate speculation (and that is not a dig at either James- just saying there is no way to know what is going on without having read the actual study and relying only on the MSNBC write-up.).

By the time we get our hands on the actual study, everyone's conceptions regarding the outcome will be hardened, doing a disservice to the work the researcher did.

The problem, however, is that the study is going to be trumpeted as proving that teens who sign virginity pledges are just as likely to have premarital sex — but are more irresponsible about it when they do! — than those who don’t sign. That’s not the case. It’s just that the teens most likely to engage in premarital sex are less likely to sign the pledges to begin with.

No. The study doesn't even look at "teens most likely to engage in premarital sex." It is looking at teens who signed virginity pledges and compares them to similar kids who didn't. Those who signed virginity pledges are no less (or more) likely to have premarital sex, but less likely to use condoms or birth control than similar kids who didn't sign the pledge. A group that is "more or less likely to engage in premarital sex" has nothing whatsoever to do with it.

Your first statement in that paragraph is EXACTLY what the study shows, i.e., "the study is going to be trumpeted as proving that teens who sign virginity pledges are just as likely to have premarital sex — but are more irresponsible about it when they do! — than those who don’t sign." Though "proving" is a little strong.

James, because a direct comparison of the two groups, ignoring any covariates, would be wrong.

Agreed, from the standpoint of social science and a research question of what impact the pledges have.

Please don't judge a study by the headline that some editor puts on it (who probably didn't even read the article, let alone the study, let alone remember enough stats to pass a Statistics 101 test).

I'm not commenting on the study at all but rather on the reporting and interpretation. The study is quite good. The problem is that it's going to be taken as saying that there's no difference between those who take the virginity pledge and those who do or even that the pledges are bad because of STD and pregnancy. The study doesn't say that at all.

Basically, "good kids" are going to be good regardless of a piece of paper and ditto "bad kids." But the "good kids" are more likely to sign the piece of paper than the "bad kids."

James, because a direct comparison of the two groups, ignoring any covariates, would be wrong. Please don't judge a study by the headline that some editor puts on it (who probably didn't even read the article, let alone the study, let alone remember enough stats to pass a Statistics 101 test).