Should I stay or should I leave? The relationship between duration in poverty-related programs and number of child maltreatment reports

Date Published

Kim, H., & Drake, B. (2016). Duration in poverty-related programs and number of child maltreatment reports: A multilevel negative binomial study. Child Maltreatment, 22(1), 14-23. 

Reviewed by
Biru Zhou

Poverty at the family level and at the neighborhood level have been identified as risk factors for child maltreatment. This study set out to examine the relationship between long term exposure to poverty and the cumulative maltreatment histories among children (birth to age 15) from a large Midwestern metropolitan area in the US. 

The authors used two samples to examine the effect of years in poverty-related programs (i.e., Aid to Families with Dependent Children, Temporary Assistance for Needy Families, or Medicaid) on the number of maltreatment reports. The first sample included children who had reports of being abused or neglected with the Missouri Child Protection Services in 1993 and 1994 (the CAN sample). This sample had 3343 families nested in 250 neighbourhoods (i.e., census tracts). The other sample included children who had a record with the Aid to Families with Dependent Children (AFDC) program without a known maltreatment report prior to the sampling period (1993-1994). This sample contained 2805 families nested within 223 tracts. These two samples of children were born in between 1989 and 1994, and were followed from birth to age 15 in different government records and data sets.

Due to the inverse-U shape distribution in both samples in the study, the authors included a quadratic term in their multilevel negative binomial model. Duration in poverty-related programs (measured in years) for families was entered as the level 1 predictor with a quadratic term. Child race (White versus non-White) was also entered as a Level 1 predictor in the multilevel analysis. The percentage of children under the federal poverty line within neighbourhoods (tract) in 1990 census data was used as the Level 2 predictor for the analysis. The authors conducted multilevel negative binomial model (i.e., a random intercept model) separately on each of the sample, controlling for child mental health, child special education, parental mental health/substance abuse, parental criminal histories, maternal foster care history, child gender and child foster care history.

Results of the analyses in both samples showed that families who stayed in poverty-related programs for a longer duration of time had greater numbers of reported maltreatment. Neighbourhood poverty did not relate to the number of child maltreatment reports in either of the sample. There was a significant child race effect in both of the analyses. In the AFDC sample, there were 30% less maltreatment reports for non-White children than for White children, and 15% less in the CAN sample. This finding suggests that low-income White families may be at higher risk of reported child maltreatment compared to non-White families. Furthermore, the positive slope coefficients (γ10 = .204  for AFDC sample and γ10 = .116 for CAN sample) for the Level 1 predictor – Family’s duration in poverty-related programs – and its negative coefficients for the quadratic terms (γ20 = -.021  for AFDC sample and γ20 = -.014 for CAN sample) in the two samples indicated that with every additional year in the poverty program, the positive association between the duration in poverty-related program and the number of maltreatment reports reduced significantly over time.

Methodological notes

The use of multilevel negative binomial models to examine nested data with two separate samples being followed up for 15 years is a strong merit of this study. The authors emphasized in their study that the non-White population in this study consisted of over 98% Black families. Therefore, the child race effect in this study represented a proxy for White versus Black family comparisons. Furthermore, the authors did not interpret or discuss the γ10 and γ20 coefficients together as the reviewer did above. Discussions and potential explanations for the diminishing effects of the duration in poverty program will be most fruitful in informing future studies and policy making. Having said that, the reviewer must stress that the interpretation and discussion in this study was interesting and very informative theoretically and analytically. The interpretation issue the reviewer brought up in this review is more a matter of methodoligical style than a disagreement.