Cross Country Estimates of Peer Effects in Adolescent Smoking Using IV and School Fixed Effects
Melbourne Institute Working Paper No. 07/12
Date: March 2012
This paper presents micro-econometric evidence on peer effects in adolescent smoking between classmates aged 15/16 years across 13 European countries. Both instrumental variables and school fixed effects are used for identification. Omitting school fixed effects, as in some existing IV studies of peer effects, is shown to lead to substantial overestimates consistent with endogenous sorting into schools. When fixed effects are included, estimated peer effects range from 0.04 to 0.34 depending on the instrument set. The preferred estimate uses the smoking behaviour of peers’ older siblings to instrument for peer smoking behaviour and suggests a statistically insignificant peer effect of 0.16. This estimate is robust to restricting the sample by dropping schools that non-randomly sort pupils into classes. Ultimately, we cannot rule out zero peer effects in smoking between adolescent classmates in Europe.