Tuesday, September 07, 2004

causality bites, video-killed-the-abstinence-star edition

From CNN.com:
CHICAGO, Illinois (Reuters) -- Teenagers who watch a lot of television with sexual content are twice as likely to engage in intercourse than those who watch few such programs, according to a study published Tuesday.

The study covered 1,792 adolescents aged 12 to 17 who were quizzed on viewing habits and sexual activity and then surveyed again a year later. Both regular and cable television were included.

'This is the strongest evidence yet that the sexual content of television programs encourages adolescents to initiate sexual intercourse and other sexual activities,' said Rebecca Collins, a psychologist at the RAND Corp. who headed the study.
'The impact of television viewing is so large that even a moderate shift in the sexual content of adolescent TV watching could have a substantial effect on their sexual behavior,' she added."
The alternative, of course, is that teenagers who opt (and who have families that allow them to opt) into watching television with a lot of sexual content are more likely to engage in more teenage sexual behavior.

Methodological note for the interested: this is a study where the putative causal inference is based on repeated measures of the same individuals, which generally does provide a logically stronger grounds for inference than a study based on a single-point-in-time survey of individuals. However, the study--unless there is an unusual, important wrinkle not included in the media report--actually provides an example of a principal danger of longitudinal data, which is that it is easy to overestimate the extent to which it allows one to make inferences about cause-and-effect. What the longitudinal data really allows you to rule out is the possibility that the correlation is driven by a tendency for teens who have start having sex to then be inspired to increase the amount of sexual content TV they watch, but that's not exactly the worrisome alternative explanation here.

6 comments:

Goesh said...

I've sopped up enough grease in too many truck stops over the years and heard too many stories by young men about their sexual escapades to question the validity of self reporting by young males on this subject matter - but then what the hell does a truck driver know about bell curves and means and factor analysis anyway? What if Mom or Dad was looking over sonny boy's shoulder as he reported on his 'activities' with young ladies, eh? Would he fess up or not? Give the same kid a burger in a truck stop with some of us older bucks around and you will hear a different story. It boils down to really knowing what you're hauling and where you're going. I can't speak for the young ladies here.

Tom Bozzo said...

The full paper is available here:

http://pediatrics.aappublications.org/cgi/reprint/114/3/e280

A couple observations regarding the methodological note. It would seem that the longitudinal data would (in principle) allow the authors to make some showing that viewing habits Granger caused subsequent sexual activity, but they don't go all the way (ahem) to making such a demonstration. In particular, they don't directly show by how much incorporating the lagged viewing habits information improves the forecast of behavior.

I'd also expect that for this type of study both unobserved causal variables and measurement error in the dependent and independent variables would not be inconsiderable. Their model description is too sketchy to determine how they might have tried to use the longitudinal data to deal with the former, and it's not obvious that the measured effects are necessarily attenuated due to the latter. So the claim that large changes in behavior result from (relatively) small changes in viewing habits strikes me as worthy of skepticism.

Goesh said...

By God! That's puttin' it in high gear. You don't haul toxic waste, do ya' buddy?

Tom Bozzo said...

Goesh, amazingly I'm not a toxic waste hauler.

Regarding your original comment, for this type of study it matters less *that* the respondents might lie about their sexual activity than *how* they lie. There are potentially bigger problems that would result from misreporting of TV watching habits and all of the other explanatory factors.

Tom Bozzo said...

As a last word, the most obvious problem I see with the analysis is that it doesn't seem to control for exposure to sexual content in other media, so it's not clear that what they're measuring is a pure-TV effect.

Goesh said...

I've skewed too many curves in big rigs and as I wobbled in that dangerous dance of covariance, scattering pesky anonomlies which I could not account for, I longed for tighter Control of the wheel and prayed to safely Regress back to that straight Linear line from whence I had started. No sir, loose coefficients like loose tires don't make for good trucking.