Decades of research suggest that neighborhood socioeconomic disadvantage increases children's health risk. This proposed project seeks to address two major weaknesses in conventional neighborhood effects research and interventions: a) the assumption that residential neighborhoods function independently of each other - ignoring that risk factors in areas where people work, learn, and play away from home may interact with residential factors; and b) as importantly, insufficient understanding of neighborhood effects mechanisms and heterogeneity in effects. To systematically address these critical barriers in the field, I propose a research and training program that will enable me to learn, use, and adapt recent advancements in Big Data analytics. I plan to model hidden interdependencies among individuals and neighborhoods and operationalize mechanisms of neighborhood effects by drawing on multiple large datasets (demographic, geospatial, networks, population flows), with several hundred million observations across multiple states, cities, and years, and match them to locally and nationally representative restricted survey data. The massive volume, great variety, and unique complexity of such data, such as relational data on inter-neighborhood dependencies and interactions, pose a challenge to the standard capabilities of hardware, algorithms, and analytical methods and models of social and population science. The proposed training program in Big Data analytics and machine learning will enable me to overcome computational and conceptual challenges and uniquely position me to: a) examine the ecological inter-neighborhood networks (econetworks) to which population groups are differentially exposed to across space and time; and b) test new contextual mechanisms underlying children's exposures to health risks. Specifically, I propose to: a) develop computational models of dynamic large scale econetworks to assess population differences in exposures to health risk factors, as they commute daily between home and workplaces; b) examine heterogeneity in econetwork effects on child health using a hybrid design that links Big Data to local and national surveys; and c) model child health risk mechanisms and causal effects using natural experiments on Big Data. The proposed training program will enable me to learn and adapt Big Data analytics, draw on its strengths, but also address some of its key limitations. With the support of a unique team of distinguished mentors and advisors, established experts in Big Data analytics, spatial demography, network analysis, child development and health risk, neighborhood change, and population heterogeneity, I will embark on a training program that will uniquely enable me to address these research goals and position me to become an independent scholar and a leader in the field.
This project contributes to advancing public health scholarship by leveraging and adapting Big Data analytical tools to overcome critical barriers in the field related to conventional assumptions about effects of neighborhood risk exposures on child health. It addresses the need to understand and model inter- neighborhood interdependencies and underlying multidimensional social capital mechanisms in order to better understand population heterogeneity in neighborhood effects on child health across space and time. The project also advances public health through an integration of rigorous social and data science methodology to create new measures and test causal hypotheses that advance our understanding of social capital forces that foster resilience to adversity and improve child health behavior and outcomes.
