Samantha Cervantes

This study evaluates the fiscal and academic consequences of New York City’s hold harmless policy during COVID-19, which aimed to stabilize school expenditures amid unexpected enrollment declines by restoring schools’ funding up to initial levels. We examine how school racial composition predicts whether or when schools receive hold harmless “treatment” and assess the impact of hold harmless on financial resources, staffing, and student outcomes, exploring heterogeneity by timing of policy announcement. Although schools with higher White student shares were no more likely than those with higher Hispanic or Black shares to receive hold harmless funds, schools with higher Black shares that did receive them saw larger per-pupil allocations due to deeper enrollment losses. Overall, hold harmless schools experienced significant increases in per-pupil spending, and reduced pupil-teacher ratio and class size, while maintaining the size of the teaching workforce. We find hold harmless had no effect on attendance or chronic absenteeism in 2021 or 2022, but improved both in 2023, when it was announced earlier. Although funds often rolled over to later years, we find no corresponding gains in student outcomes. Overall, the policy effectively preserved school-level spending and staffing – as intended – with some improvements in student outcomes when announced early.

Despite broad consensus that good schools matter, evidence on the causal effect of proximity to good schools on student performance is limited. This paper explores the causal impacts of living near a good school, leveraging plausibly random variation in proximity to good schools among public housing children. We find that living closer to a good school increases reading and math scores, improves attendance, and reduces chronic absenteeism. The probabilities of attending the nearest good school and any good school increase with proximity, suggesting both direct and indirect effects. As a policy lever, pupil transportation appears to ameliorate the negative effects of distance.

Place-based crime prevention policies are typically evaluated based on their effects on crime. Yet these interventions may also shape the behavior of residents in targeted areas in ways that are less well understood. This paper asks whether a crime-reduction intervention that increases visible police presence affects residents’ political behavior and residential mobility. We study a randomized experiment conducted in New York City public housing developments, in which high-intensity police streetlights were installed to deter crime. These lights dramatically increased nighttime illumination and introduced visible police infrastructure into residential environments. The intervention may have affected residents through multiple channels. First, a safety channel: prior work shows that the lights significantly reduced nighttime outdoor crime. Second, an intrusion channel: the lights generated substantial brightness (approximately 600,000 lumens) and noise from generators, which residents reported as disruptive. Third, a signaling channel: the lights served as highly visible markers of police presence, potentially altering perceptions of surveillance and state control. These mechanisms generate distinct empirical predictions. If responses are driven by safety improvements or physical intrusion, effects should scale with treatment intensity and be strongest among residents closest to the lights. If instead the lights operate primarily as a signal of policing, effects should appear at the extensive margin of treatment and not vary systematically with dosage or proximity. We link voter registration and turnout records, police data, and public housing administrative files with precise residential addresses to construct a panel of residents in treated and control developments over the study period (2012-2024). Exploiting development-level random assignment of streetlights within matched pairs, we estimate intention-to-treat effects of exposure to any lights and examine heterogeneity by dosage and vertical proximity to light towers, which we proxy with the residential floor in which voters live: those living in lower floors are more exposed to the brightness of the lights than those in higher floors. We find no meaningful effects on residential mobility or attrition from public housing. However, the intervention led to significant changes in political behavior: treated residents were less likely to register as Democrats and more likely to register as Republican or non-partisan, and were less likely to vote across election types. These effects do not scale with treatment intensity and do not vary by residential proximity to the lights, consistent with a signaling mechanism rather than physical nuisance. These findings suggest that visible policing infrastructure can reshape political engagement even when it reduces crime, highlighting the importance of evaluating place-based interventions beyond their intended outcomes.