Sunday, November 11, 2007

There is no Valid Evidence that Early Sex causes delinquency.

The claim that it does is based on correlation. Obviously there is omitted variables bias (I just mean correlation does not imply causation). Paige Harden et al study twins one of which had sex earlier than the other. They actually find a negative correlation between early sex and later delinquency such as shop lifting and drug selling.

Rick Weiss of the Washington Post, does not mention whether this within twin pair correlation is statistically significant. Also even within twin pairs correlation is not causation. Teenage boys are generally eager to have sex (putting it mildly). Thus the girls decide who gets it. Maybe they don't go for proto-delinquents. Causal explanations would include that delinquency is the result of boredom frustration and too much time on one's hands and sex uhm relieves these problems and that girls who are less delinquent on average than boys have more influence on their boyfriends than vice versa.

It is interesting that the clearly invalid causal argument based on correlation in the population is causation argument was accepted in this case. I think this is clearly true because it eased cognitive dissonance in people who disapprove of early (or all pre-marital) sex and because it corresponds to prejudices based on correlation. Weiss notes that, legislating that pi = 22/7, the US government
has decided the question by fiat "U.S. educators wanting a piece of the nation's $200 million "abstinence only" budget must adhere to a curriculum that links sex to delinquency and explicitly precludes discussion of contraception."

Another example, as I noted recently, is teenage pregnancy and infant mortality. Weiss notes that Arlene Geronimus is still on the case.

No comments: