Importance:Although school environments are thought to influence health behaviors, experimental data assessing causality are lacking, and which aspects of school environments may be most important for adolescent health are unknown. Objective:To test whether exposure to high-performing schools is associated with risky adolescent health behaviors. Design, Setting, and Participants:This natural experiment used admission lotteries, which mimic random assignment, to estimate the association of school environments and adolescent health. A survey of 1270 students who applied to at least 1 of 5 high-performing public charter schools in low-income minority communities in Los Angeles, California. Schools had an academic performance ranked in the top tertile of Los Angeles County public high schools, applicants outnumbered available seats by at least 50, and an admissions lottery was used. Participants included lottery winners (intervention group [n = 694]) and lottery losers (control group [n = 576]) from the end of 8th grade and beginning of 9th grade through the end of 11th grade. Intention-to-treat (ITT) and instrumental variable techniques estimated the association of winning the lottery and attending high-performing schools with health behaviors and whether the association varied by sex. Data were collected from March 11, 2013, through February 22, 2017, and analyzed from October 1, 2017, through July 1, 2018. Exposures:Schools were considered high performing if they placed in the top tercile of public high schools in LA County on 2012 state standardized tests. Most students attended that same school for 3 years (9th-11th grades). Main Outcomes and Measures:Primary self-reported outcomes were 30-day and high-risk self-reported marijuana use. Additional health outcomes included 30-day alcohol use, alcohol misuse, ever being in a fight, ever having sex, and past-year delinquency. Potential intermediate factors (time studying, truancy, school mobility, school culture, school order, teacher support for college, and proportion of substance-using peers in students' social networks) were also examined. Results:Among the 1270 participating students (52.6% female; mean [SD] age at enrollment, 14.3 [0.5] years), ITT analysis showed that the intervention group reported less marijuana misuse than the control group (mean marijuana misuse score, 0.46 vs 0.71), as well as fewer substance-using peers (9.6% vs 12.7%), more time studying (mean, 2.63 vs 2.49 hours), less truancy (84.3% vs 77.3% with no truancy), greater teacher support for college (mean scores, 7.20 vs 7.02), more orderly schools (mean order score, 7.06 vs 6.83), and less school mobility (21.4% vs 28.4%) (all P < .05). Stratified analyses suggest that among boys, intervention participants had significantly lower marijuana use (mean misuse score, 0.43 vs 0.88; difference, -0.45; 95% CI, -0.78 to -0.13) and alcohol misuse (mean misuse score, 0.52 vs 0.97; difference, -0.44; 95% CI, -0.80 to -0.09) scores compared with control participants, whereas no significant health outcomes were noted for girls. Conclusions and Relevance:This natural experiment provides evidence that school environments can improve risky behaviors for low-income minority adolescents.