Violence in schools is recognized as a serious public problem, but both its causes and its effects are difficult to study with scientific rigor. Existing research links school violence exposure to various negative outcomes – including socio-emotional problems like depression and suicidality, and academic problems like lower rates of post-secondary attainment – but because of violence’s non-random distribution process, it is very difficult to determine whether these constitute causal effects. And theories about crime etiology predict that heterogeneous pathways lead to victimization for different reasons – for example, some students may be victimized because they place themselves at risk by voluntarily participating in violent situations, whereas a different type of students may be victimized because social isolation makes them “easy targets” – but endogeneity and confounding make these pathways difficult to examine. To complicate matters, violence exposure may have different effects on different kinds of students, and different types of students may vary in how much they benefit from schools’ protection. In this dissertation, I address these and related problems using data from the National Longitudinal Study of Adolescent Health. Leveraging the school-based longitudinal structure and using techniques to draw causal inferences from observational data, I test several hypotheses about the effects of violence exposure on different types of students and the role of the school in violence prevention. I find that violence increases depression for both males and females; that students who experience violence early in life fare worse than those who do not; that students who experience isolated exposures suffer acute shocks to their well-being; that students who suffer repeat victimizations fare worse overall but are less affected by each specific instance; that students who perpetrate violence and students with few friends are both more likely to be victimized; and that positive school climate protects socially isolated students from victimization, but has no such protective effect on students who are perpetrators. Besides confirming predictions about violence derived from life course theory, routine activities theory, and others, this dissertation contributes to a growing body of methodologically rigorous evidence that violence exposure itself exerts negative causal effects on students’ lives.