Files
Abstract
This dissertation investigates how living in disadvantaged neighborhoods influences children’s educational outcomes and how these effects vary across subgroups defined by demographic, family, health, and school characteristics. Prior research has documented negative effects of neighborhood disadvantages on outcomes such as cognitive skills and high school graduation but has often been limited by a focus on a few moderators and by theoretical frameworks that offer conflicting or imprecise predictions. Using a causal forest approach with data from ECLS-K:2011 and HSLS:09, this study inductively identifies patterns of treatment effect heterogeneity without relying solely on prespecified subgroups, enabling a broader and more nuanced understanding of neighborhood effect heterogeneity. Across three chapters, the dissertation examines children’s cognitive skills in elementary school, social-emotional skills in elementary school, and college enrollment after high school. Findings reveal substantial variability in how neighborhood disadvantage affects different groups. For cognitive skills, children living in single-parent households and urban disadvantaged areas are most negatively impacted, suggesting that family structure and locality type are key moderators beyond conventional socioeconomic measures like family income. For social-emotional skills, family structure, caregiver age, household SES, and home language shape vulnerability, highlighting the role of both family resources and cultural background. In contrast, the college enrollment results tentatively suggest that youths from more advantaged families—those with higher SES and stronger educational aspirations—may be more sensitive to neighborhood disadvantage, a pattern that aligns with the relative deprivation theory. Theoretically, the dissertation extends the compound disadvantage theory by emphasizing dimensions such as family structure, residential locality type, and cultural factors, while also suggesting that different theories may be more or less relevant for explaining neighborhood effects depending on the specific educational outcome under consideration. Methodologically, it illustrates the value of machine learning techniques like causal forests for uncovering complex, intersectional patterns of heterogeneity that traditional models might miss. Policy implications include the need to tailor housing and neighborhood-based interventions to the most vulnerable subgroups, integrate family-level supports, and invest in community resources that facilitate college transitions. Overall, this dissertation underscores the importance of looking beyond average effects to capture the diverse ways neighborhood disadvantage shapes educational trajectories, offering new directions for theory, research, and practice.