Understanding the effects of group composition by risk for violence. Experimental evidence from El Salvador

Abstract

Globally, 150 million adolescents report experiencing or engaging in peer-to-peer violence in and around schools. After-school programs (ASPs) have emerged as a strategy to mitigate this risk, but how does the peer group composition, based on the risk of violence, influence an ASP’s effectiveness? To address this question, I conducted a randomized experiment in El Salvador, assigning youths to a control group, homogeneous peer groups, or heterogeneous peer groups within an ASP. I find that there are limited average differences between the homogeneous and heterogeneous groups. This lack of differences is explained by the differential impacts of group composition based on students’ baseline propensity for violence: the heterogeneous treatment benefits the most vulnerable students and does not harm the least vulnerable. Additionally, the opposing effects of group composition components play a role: a higher mean of peers’ risk for violence worsens outcomes, while a higher variance improves them. These findings have important implications for public policy, offering insights into optimal group composition for violence reduction programs.