Abstract
In this article, we investigate the factors that lead to changes in the socioeconomic complexion of urban neighborhoods along four critical dimensions: crime, youth social distress, home ownership, and economic conditions. We argue that the dynamics of these dimensions are better apprehended simultaneously considering their potential mutual relationships and we propose a cross-lagged panel model approach within a structural equation modeling framework. Neighborhoods in Charlotte, North Carolina, are used as a case study, and change is modeled at several time lags throughout the 2000–2010 decade. Findings indicate that the model performs well and that it offers a very promising avenue for modeling the socioeconomic changes of neighborhoods that accounts for complex longitudinal effects as well as spatial dependencies. Specifically, it shows that lower human capital manifested by a decline in youth indicators is significant in explaining subsequent increases in crime and declines in economic indicators. Also, the predominance of housing stock constructed in the 1950s and 1960s is a significant trigger of declines across all neighborhood indicators. Finally, spatial spillover effects between neighborhoods are found to be short-lived and dissipate after a few years.
Notes
1. Specified as the six nearest neighbors within 5 miles, after testing different alternative specifications. This neighborhood definition considers the varying size of neighborhoods within the study area.
2. While we model race as an exogenous, predictor variable in this study, it is certainly plausible to argue that race is an endogenous part of neighborhood socioeconomic dynamics. Unfortunately, data constraints prohibit its inclusion as a dynamic variable.
3. Since only the 2000 data are used to measure transit access, walking distance to light rail is not incorporated as the rail line opened only in 2007. While research has indicated that this line has had a significant, positive effect on real estate values in surrounding areas (Yan, Delmelle, & Duncan, Citation2012), and therefore possibly also on other socioeconomic dimensions of neighborhoods, the effect of this new transportation infrastructure is not studied here.
4. NSA boundaries largely follow Census blocks, so the Census variables were aggregated from block-level counts.