This study investigates whether peer obesity is a driver of individual weight changes in public school children and whether the impact of peer effects changes as children age. Quantifying peer effects is important for understanding the social determinants of obesity and for planning effective school wellness policies. However, the extant empirical research on peer effects is limited due to difficulties in separating causal influences from confounding factors. This study overcomes some of these difficulties by using a within-school, across-cohort empirical design to separate confounding factors at the individual, school and school-grade level for over one million public school children. The results show that increases a one standard deviation increase in average classmate body mass index (BMI) leads to a modest but meaningful increase of 0.395 standard deviation increase in a child's own BMI. Peer-effects are highest (0.813) for children in Kindergarten and decline with age. These findings suggest that the critical time for school-grade level intervention may be in the earliest ages of childhood development.