The Add Health survey that was conducted among US adolescents has been widely cited in the media recently. It was this data set that got evolutionary biologist Satoshi Kanazawa into so much trouble a few weeks ago when he misrepresented the data as proving that black women were less attractive than white women. That blunder aside, the data itself is extremely useful to researchers attempting to unravel the mysteries of adolescent behavior. Economists have recently tapped into an unusual feature of this data, the identification of peer groups within high schools, to directly measure effects of peer pressure on risky behavior. Specifically they ask the question: If a student’s best friend engages in a risky behavior (sex, smoking, marijuana use or truancy) then what is the probability that the friend will engage it the same behavior? Best friends tend to come from similar family situations, be the same race and age and have similar educational goals as well as similar attitudes toward risk. The authors are able to control for these factors and find that if a student’s best friend has had intercourse then the probability that she or he will have had intercourse in the following year is 4.5 percentage points higher than the base line of 14%. The probability of “intimate contact” (essentially making it to third base) increases by 4 percentage points on a base line of 22%. The best friend effect is as large as living in a single parent household or of having parents who did not finish high school – both of which increase the probability that a student has had been sexually active.
What I find interesting about these results is that unlike other risky behaviors that best friends might do together (like smoking or skipping school), this is risky behavior is that they do apart. One explanation is that teens are smoking pot together and that pot smoking leads to a higher rate of sexual activity or that they are drinking together and that this leads to higher rates of sexual activity. It turns out that pot use doesn't increase entrance to sexual activity in teens, so even though having a best friend who smokes pot increases (to a small degree) the probability that the student also smokes this can`t explain the increase in sexual activity. Alternatively while alcohol use increases sexual activity (significantly), surprisingly having a best friend who drinks alcohol does not increase the probably that a student drinks above the baseline. So that isn`t the effect either. My final thought is this. Everyone wants to argue that sexual intercourse among teens is psychologically damaging (we have talked about this before) and increases depression in adolescents. If this is true then why would teens not learn this from their best friends? Other risky behavior that is done together I understand will not lead to learning, but sexual intercourse is sequential – someone has to be the first to do it. If it made teens unhappy, then would that not lead to a decrease in a friend’s probability of having sex in the subsequent year?
Finally, on a somewhat related note, I was sent this fascinating infographic this week on teen pregnancy that I wanted to share with you. Enjoy!
Card, David and Laura Giuliano (2011). “Peer Effects and Multiple Equilibria in the Risky Behavior of Friends.” NBER Working Paper 17088.