Ethnic Residential Segregation and Child Mortality: A Spatial Analysis of 1880 Newark, NJ

Hongwei Xu, Brown University
John R. Logan, Brown University
Susan E. Short, Brown University

Using administrative geographic units and ignoring spatial features and processes, previous research on residential segregation and mortality is susceptible to measurement errors, scale-sensitive results, and improper statistical inference. Using household-level geocoded data of 1880 Newark (NJ), we adopt a spatial perspective for neighborhood representation, segregation measures, and modeling strategy to examine the association between ethnic segregation in egocentric neighborhood and child mortality. We found that child mortality exhibited a strong spatial pattern. Controlling for neighborhood socioeconomic conditions, interaction with other ethnic populations on a local scale was positively related to child mortality, possibly due to an expanded social network that facilitated the transmission of infectious diseases. Spatial autocorrelation remained notable at 500-900 meters, a scale that exceeded the distance between adjacent administrative neighborhoods in the inner city. Variations in mortality risk were accounted for more by spatial effects, compared to non-spatial neighborhood effects. Our findings suggest that spatial thinking should be employed in future research.