Smoking behavior among adolescents is influenced by their social networks. Using data from two high schools with different network structures in Add Health, Wang et al. examine how peer influence spreads and weakens as it moves through friendship networks. By simulating interventions aimed at reducing smoking by targeting different groups of students, the study measures both direct effects on targeted students and indirect spillover effects across the wider network, focusing on how influence decays across multiple degrees of separation and how this depends on network structure and intervention strategy.
Findings reveal how targeting a small share of highly connected students produces the strongest overall impact, with in-degree and eigenvector targeting consistently outperforming random selection, especially at 10-30% coverage. However, benefits level off at higher coverage levels (around 40–50%) due to network saturation, where most individuals have already been reached directly or indirectly. This study highlights the importance that context matters and policies should cater towards the density structure of social networks.
To read the full article, click the link below. For more important findings using Add Health data, visit the Add Health publications page.
Wang C, Butts C, Hipp J, Lakon C. Peer influence decay and behavioral diffusion in adolescent networks: A simulation approach. Science. Published online April 30, 2026. https://doi.org/10.1126/science.aea9297