The Effect of Having Aggressive Friends on Aggressive Behavior in Childhood: Using Propensity Scores to Strengthen Causal Inference.

This study uses propensity scores to statistically approximate the causal effect of having aggressive friends on aggressive behavior in childhood. Participants were 1,355 children (53% girls; 31% minority) in 97 third and fifth grade classrooms enrolled in the Classroom Peer Ecologies Project. Propensity scores were calculated to control for the impact of 21 relevant confounder variables related to having aggressive friendships and aggressive behavior. The 21 variables included demographic, social, and behavioral characteristics measured at the beginning of the school year. Presence/absence of aggressive friends was measured in the middle of the school year, and aggressive behavior was measured at the end of the school year. Results indicate a significant effect of having one or more aggressive friends on children's aggressive behavior above and beyond the effects of the 21 demographic, social, and behavioral variables. The propensity score model is compared to two other models of peer influence. The strengths and practical challenges of using propensity score analysis to study peer influence are discussed.