Eckenrode, J., Smith, E.G., McCarthy, M.E., & Dineen, M. (2014). Income inequality and child maltreatment in the United States. Pediatrics, 133 (3), 454-461
Recent studies have shown the importance of community factors—especially poverty levels—as predictors of child abuse and neglect; this current national study appears to be the first attempt to examine the added effect of income inequality. The unit of analysis here was county (N = 2877); data were annual rates, averaged over five years (2005–2009), for the following variables:
• Child maltreatment rate (substantiated reports per 1000 children): Data from the US Children’s Bureau, National Child Abuse and Neglect Data System (NCANDS), collected annually from state child welfare systems.
• Poverty level (percentage of children living at or below federal poverty line): Data from the US Census Bureau, American Community Survey (ACS).
• Income inequality (Gini coefficient, where 0 = perfect equality, and I = perfect inequality; a single individual has all the income): Data also from ACS.
Bivariate correlations between these three variables were all positive and highly significant; the correlation between poverty and child maltreatment was consistent with earlier studies. Nonparametric regression modelling was used to predict child maltreatment rate from poverty level and income inequality, controlling for possible confounding variables (ethnicity, education, public assistance income & infant death rate). Income inequality was found to be significantly associated with child maltreatment rates at all poverty levels, but the effect was greater at higher levels of poverty.
The authors state that these findings contribute to the growing literature linking greater income inequality to a range of poor health and well-being outcomes in infants and children. They cite evidence that inequality effects are generally larger for larger units of analysis (e.g., nations, states) and smaller for smaller units such as neighborhoods, where income is a stronger predictor.
The research appeared to be generally well-conducted, and the methods appropriate; a number of strategies were used to deal with various data problems. Because of differences in reporting practices in some states, 259 actual counties were combined into regions, or “statistically equivalent entities,” to give the total N of 2877. State child maltreatment rates varied widely, from 0.2% to 3.1%; state was included as a fixed effect in the analysis to control for this difference. A natural log transformation was used in the regression analysis to deal with a skewed distribution of county child maltreatment rates. Analysis was performed using the generalized additive model (GAM) in SAS—a flexible nonparametric regression-modelling technique.
Further research is clearly needed to determine how income inequality actually works. At a given poverty level, why should the mere presence of rich people make it more likely that children are maltreated? If this effect is really more common in larger geographic units, perhaps it reflects geographic subdivisions with different social climates (e.g., levels of social order or social capital). This might suggest avenues for possible future preventive work in communities.