The study sample was obtained from the Illinois Families Study (IFS), a panel study of 1998 welfare recipients in Illinois. The IFS survey was linked to other administrative data, such as employment, cash welfare benefits, and child protection events. The final sample includes 949 families who participated in all four waves of interviews (1999 through 2003) yielding 3,796 person-wave observations. Economic hardship indicators include housing hardship, a dichotomous measure of eviction, being behind on rent/mortgage, or having one or more utilities turned off within the last year, and food hardship, a 4-item index adapted from the USDA food security scale. The key outcome is child maltreatment, operationalized as investigated CPS reports. Logistic regression, event history and fixed-effects models are estimated.
Housing hardship is significantly associated with CPS involvement using all three analytic approaches, even after controlling for income to poverty ratio, sources of income, and a set of observed time-varying variables. However, the effect is moderated by marital/cohabitation status: only single, non-cohabiting caregivers have an elevated risk of CPS reports associated with housing difficulties. Food hardship is not associated with CPS involvement regardless of marital/cohabitation status. Other findings from logistic regression and event history analyses indicate that the number of children in the household and domestic violence are predictors of a CPS report, while perceive social support reduces the likelihood of this outcome. However, most of these effects become statistically insignificant in the fixed-effects model.
This study helps further our understanding of how poverty experiences within low-income populations affect the risk of child maltreatment. Within this sample of low-income women, housing problems are predictive of involvement with CPS; food hardship (at least as it is measured) is not. These findings suggest that specific forms of economic hardship may be more salient for CPS involvement than others, at least among low-income, single caregivers. Furthermore, the comparative analytic results suggest that more conventional approaches to modeling CPS events (logistic and event history techniques) are heavily influenced by unobserved characteristics. Fixed-effects models address this concern, at least as it relates to time invariant factors.