In previous research, community violence exposure has been found to be directly associated with poor academic functioning. Further, internalizing and externalizing problems have been identified as significant mediators of poor academic functioning in children who have witnessed or experienced community violence. This study reports an investigation of the direct and mediated links between community violence exposure and academic functioning in adolescents. A sample of 1,552 adolescents was selected from a public-use dataset from the National Longitudinal Study of Adolescent Health (AddHealth). Measures of socioeconomic status, cognitive ability, community violence exposure, internalizing behavior problems, externalizing behavior problems, and academic functioning were constructed from variables in the AddHealth dataset and analyzed using structural equation modeling. Multi-group comparisons were conducted to compare the direct and mediating models by gender and level of urbanicity. Analyses suggest that community violence exposure is directly associated with poor academic outcomes, even after controlling for cognitive ability and socioeconomic status. The direct model applied equally well to adolescents across level of urbanicity, but did not apply equally well to males and females. Support was no! found for !he meditation model.

This dissertation consists of three essays in the field of environmental economics. The first chapter provides the first causal evidence that hostile activities online lead to physical violence. Given the recently documented relationship between pollution and social media, I exploit exogenous variation in local air quality as the first step to instrument for online aggression. In an event study setting, I find volatile organic compounds (VOCs) increase by 7% when refineries experience unexpected production outages. Together with higher air pollution, I find more aggressive behaviors both online and offline, as well as worse health outcomes near refineries. A one standard deviation increase in surrounding VOCs leads to 0.16 more hate crimes against Black people and 0.23 more hospital visits per thousand people each day. Second, I consider how emotional contagion spreads through social networks. On days with pollution spikes, surrounding areas see 30% more offensive and racist tweets and 12% more crimes; those geographically distant but socially networked regions also see offensive and racist tweets increase by 3% and more crimes by 4.5%. Nationally, overlooking spillovers would underestimate crime effects of pollution by 24%. My findings highlight the consequences of social media hostility and contribute to the public debate on cyberspace regulation. The second chapter, which is coauthored with Andrew Wilson, analyzes the relationship between weather and railway accidents. Rail thermal expansion and contraction are key considerations in rail design and construction; rail operators and rolling stock may likewise exhibit vulnerability to temperature changes. We quantify the sizes of these effects by leveraging a comprehensive dataset of railway malfunctions in the United States spanning 1997-2019. We find that both heat and cold cause elevated rates of railway malfunctions, with relatively larger increases in the number of incidents leading to a casualty as well as the number of injuries and deaths resulting from these incidents. We find that exposure to daily temperatures averaging over 30°C (86°F) leads to a 16% increase in the number of rail malfunctions, a 13% increase in the number of incidents leading to a casualty, and 18% and 36% increases injuries and deaths-effects net of any operational adjustments made to mitigate these effects. Further, while we also find that warmer locations exhibit a weaker relationship between heat and railway malfunctions, we find no evidence that companies are learning, year-over-year, how to reduce accidents. Finally, we note that effects of heat are strongest for derailments (versus other types of malfunctions) and freight trains (versus passenger trains). Our findings highlight the vulnerability of the railway system to the climate. The number of injuries and deaths associated with weather exposure-especially in comparison to operators' reported private costs of equipment failure-suggests a role for enhanced rail safety regulations and adaptation funding to protect critical heat-exposed infrastructure. The third chapter, which is a joint work with Douglas Almond and Muye Ru, explores the impact of federal policy rollback on methane leakage. Improvements in satellite measurement enable independent assessment of regulatory and climate policy. In August 2020, the Trump Administration lifted Obama-era requirements that oil and gas firms detect and repair methane leaks. We merge geo-identified data from the European TROPOMI (satellite instrument) to the specific locations of the US oil and gas infrastructure. Using a difference-in-differences design, we find a prompt increase in US methane emissions following the summer 2020 rollback. The number of high-methane emission events from the oil and gas sector more than doubled after the rollback relative to the coal sector, which did not experience the same regulatory rollback. While the oil and gas industry claims it faces a persistent, profit-making incentive to stem natural gas leaks and emissions, we find a large and nimble response by industry to changes in federal policy. Public policies that reduce methane externalities are critical given that global methane concentrations are rising at an increasing rate.