Research

Neighborhood environments play a critical role in shaping health and well-being. Yet, existing indices of neighborhood disadvantage, such as the Area Deprivation Index, rely solely on multi-year, survey-based data that may overlook physical, environmental, and infrastructural dimensions of place. 

This project aims to develop a multidimensional index of neighborhood disadvantage by integrating three data sources: (1) street-level imagery from Google Street View, (2) environmental indicators from remote sensing, and (3) standardized socioeconomic measures from the U.S. Census and American Community Survey (ACS). We will first extract features of the built environment (e.g., housing condition, sidewalk quality), environmental context (e.g., vegetation cover, impervious surfaces), and safety (e.g., visible signs of disorder such as graffiti, broken windows, or abandoned buildings) using computer vision and geospatial tools. These will be combined with ACS-based measures such as education and employment to construct three composite indices using complementary approaches: theory-driven domain weighting, Bayesian latent factor modeling, and machine learning–based dimensionality reduction. We will then link these indices to individual-level longitudinal data from the Precipitating Events Project, a cohort study of older adults in New Haven, CT, to evaluate how well each index predicts four categories of aging outcomes: mental health, cognitive functioning and impairment, physical functioning and disability, and major clinical events including hospitalization and mortality. 

This project will produce a policyrelevant index of neighborhood disadvantage that captures social, economic, environmental, and physical dimensions of place. By advancing measurement and prediction of spatial inequality, this work will lay the foundation for future research on place-based health disparities and inform local and national efforts to allocate resources, design interventions, and promote equity. 

This project advances the study of neighborhood effects and educational inequality by improving upon two challenges: incomplete understanding of mechanisms and limits to causal inference. Our high-resolution approach captures with greater precision the visible, environmental, and institutional features that influence children’s outcomes. In turn, this would, reduce measurement error and clarify which neighborhood features matter most, for whom, and through what pathways. Although predictive performance alone does not establish causality, our index represents a necessary precursor for more precise causal identification. By providing subdomain scores that illuminate potential mechanisms and by capturing previously unobserved neighborhood dimensions, the index would provide a publicly-available tool that would strengthen future quasi-experimental and longitudinal studies that seek to establish causal effects through additional research designs (e.g., natural experiments). By creating and openly sharing a high-resolution, multidimensional measure of neighborhood disadvantage, this project directly advances the Russell Sage Foundation’s mission to understand and reduce inequality. 

Our visual index will give researchers and policymakers a more precise, timely, and actionable tool to identify which neighborhood conditions most affect children’s educational opportunities. The open-access dataset, code, and documentation will lower barriers to studying place-based inequality, support replication and methodological transparency, and inform evidence-based interventions, such as targeted school funding, early childhood initiatives, and neighborhood investment, to promote educational equity across communities.