U.S. fuel-economy and greenhouse-gas standards index each vehicle's target to its footprint, the area between its wheels, so larger vehicles face less stringent targets and manufacturers are rewarded for increasing vehicle size. Using differences across vehicles in the regulatory reward for enlargement, I show that manufacturers increased the footprint of more-exposed vehicles and that road deaths rose as these vehicles entered local fleets. To evaluate alternative standards, I estimate an equilibrium model of vehicle demand, design, fuel-saving technology, and credit trading. Removing the reward for size while maintaining the regulation's average fuel-economy target would avert about 2,600 road deaths per year on average from 2013 to 2024. Over the affected vehicles' lifetimes, tailpipe emissions rise by 1.4 percent and the net economic cost, including climate damages, is about $120,000 per road death averted. Under footprint indexing, tighter standards raise credit prices and increase the reward for enlargement. Tightening the indexed standard increases road deaths, while the same tightening with a one-quarter reduction in the footprint slope preserves 98 percent of its emissions reduction and reduces deaths instead. The safety consequences of tighter standards depend on how strongly their targets reward vehicle size.

How do local churches shape economic and social life? We study this question in the Southern Baptist Convention (SBC), the nation's largest Protestant denomination, exploiting two shocks to congregational life: pastor deaths and pastor convictions for sexual abuse. Using matched-controls event study designs, we show that pastor deaths lead to church closures, while convictions generate large and persistent reductions in Baptist attendance and engagement. Both shocks reduce labor-force participation, raise unemployment, and lower nearby housing prices for several years. The shocks diverge on outcomes plausibly affected by religious beliefs rather than networks: convictions shift two-party Democratic vote share leftward and increase opioid mortality, while deaths produce no detectable changes in either. This pattern supports a framework in which deaths primarily disrupt access to congregational networks, while convictions additionally erode religious norms that shape political and health behaviors.

I study how the deregulation of radio content in the U.S. transformed local politics and public health outcomes. The 1987 repeal of the FCC’s Fairness Doctrine removed requirements for broadcasters to air contrasting viewpoints on political issues and triggered the expansion of conservative talk radio. I find that quasi-exogenous exposure to growth in conservative talk radio raised Republican vote shares across presidential, Senate, and House elections, and also increased “deaths of despair” (alcohol-, overdose-, and suicide-related mortality). These effects emerge in the early 1990s and persist decades later, underscoring how partisan media ecosystems can reshape both politics and public health.

School-based health centers (SBHCs) aim to improve child and adolescent health by providing care at school, but evidence on their causal and long-run effects remains limited. We estimate these effects using difference-in-differences designs that exploit staggered SBHC openings relative to matched control schools. Linking thirteen years of near-universe psychotropic prescription data to more than 1,800 SBHC openings, we find persistent increases in youth antidepressant and antipsychotic use. Preliminary analyses using nationwide Medicaid claims also indicate that SBHC openings increase health-care utilization. We also use Texas administrative education records linked to 54 SBHC openings to estimate effects on attendance, disciplinary outcomes, high school completion, college enrollment and graduation, and longer-run economic outcomes. Together, the project examines how school-based access to care affects both health-care use and students’ broader trajectories.

Using thirteen years of U.S. outpatient prescription claims covering more than 100 million psychotropic prescriptions, we distinguish within-physician learning from differences across medical-school cohorts in guideline-deviant prescribing. Although older physicians prescribe “red-flag” doses more often in the cross-section, following fixed cohorts over time reveals that red-flag prescribing declines and becomes less dispersed, consistent with learning-by-doing and selection. The apparent age gradient is instead driven largely by cohort differences: more recently trained physicians begin with safer prescribing patterns, while earlier cohorts retain higher baseline propensities to deviate from guidelines. An event study around the 2011 FDA revision discouraging high-dose citalopram shows that physicians of all ages respond quickly, but persistent cohort gaps remain. The results suggest that experience improves prescribing within physicians even as differences in training generate lasting disparities across cohorts.

This project studies how sorting across online communities and changes in individual speech contribute to online polarization. Using the universe of comments and posts made on the social media platform Reddit, I track users across communities and examine how their language evolves before and after they begin or end participation, including their speech in other communities. The goal is to distinguish changes in community composition from changes in the behavior of existing participants.

This paper estimates the effects of fatal school shootings on youth psychotropic drug use over the subsequent 5.5 years. Using an event study approach that compares trends in drugs prescribed by providers practicing close to a school shooting relative to those practicing slightly farther away, we find that fatal school shootings lead to large increases in youth psychotropic drug use that persist for years after the event. These effects are driven primarily by increases in prescriptions for antidepressants and antipsychotics and are observed both among youth who were previously taking psychotropic medication and those who were not.