Wednesday, March 15, 2006

Estimating Peer/Social Effects without Random Assignment

In the previous posts, I highlighted the use of random assignment of neighborhoods or peers to estimate how individuals are affected by social influences. Social effects can also be estimated using events or policies which randomly affect some members of a particular social network. If social effects are relevant to the behavior of interest and my friends behavior receives a shock, then I should be more likely to engage in the same behavior than an otherwise similar person who's friends did not receive the shock.

One of my 985 students last year, Julia Gluck, wrote a thesis (winner of the 2005 Allyn Young Prize for outstanding economics thesis) which provides a nice example of this methodology. Julie wanted to estimate the effect of peers on adolescent risk behaviors. A large literature documents the fact that girls who reach menarche earlier have sex earlier and are more likely to become pregnant. Julie shows, however, that no relationship exists between the age of menarche among friends. Thus having friends who develop relatively early (or late) is an exogenous shock to the behavior in girls' social networks. Ultimately, Julie shows that girls whose friends develop early are more likely to engage in sexual activity than otherwise similar girls whose friends develop later.

Julie's thesis provides fairly strong evidence for peer effects in adolescent sexual behavior. However, it does not specify what specifically generates these effects. Does increased sexual activity by friends increase desires for sex? Does it increase pressure to conform? Does it reduce fears about the risks involved? Does it change the typical social situation (e.g., do boys become more prevalent and parties become more common )? These questions require further research.

Comments: Post a Comment

Subscribe to Post Comments [Atom]

<< Home

This page is powered by Blogger. Isn't yours?

Subscribe to Posts [Atom]