What determines the out-of-home placement of children in the USA?

https://doi.org/10.1016/j.childyouth.2012.05.004Get rights and content

Abstract

Using NCANDS data of US child maltreatment reports for 2009, logistic regression, probit analysis, discriminant analysis and an artificial neural network are used to determine the factors which explain the decision to place a child in out-of-home care. As well as developing a new model for 2009, a previous study using 2005 data is replicated. While there are many small differences, the four estimation techniques give broadly the same results, demonstrating the robustness of the results. Similarly, apart from age and sexual abuse, the 2005 and 2009 results are roughly similar. For 2009, child characteristics (particularly child emotional problems) are more important than the nature of the abuse and the situation of the household; while caregiver characteristics are the least important. All these models have low explanatory power.

Highlights

► Determinants of the out-of-home placement of children in the USA are modelled. ► Four different estimation techniques give broadly similar results. ► The determinants are largely stable over the 2005–2009 period. ► characteristics are the most important determinants.

Introduction

The decision to place a child in out-of-home care is one of the most important made by the child care system. There have been many previous studies of this decision (Bhatti-Sinclair & Sutcliffe, 2012), but after fifty years of research a consensus on the most important variables determining this decision has yet to emerge. This paper contributes to this question in a number of ways. Most previous studies of the placement decision have used either logistic regression (logit analysis) or discriminant analysis; with probit analysis and artificial neural networks (ANNs) each used only once. Just one study has used more than one of these four techniques. Since none of these techniques is clearly superior, this study uses all four estimation techniques; permitting an examination of the robustness of the conclusions to the technique used. It is also the first placement study to replicate the results of a previous analysis. Knott and Donovan (2010) (K&D) analysed the National Child Abuse and Neglect Data System (NCANDS) data for 2005, and the current study replicates their analysis using 2009 data, enabling an investigation of the temporal stability of the factors associated with placement. As well as odds ratios, this paper also computes the percentage change in the probability of placement for each independent variable, and the cumulative probability of placement for a child with a specified set of risk factors.

This research is positive rather than normative and seeks to discover what factors are associated with actual placement decisions, not what factors should determine placement decisions. It investigates placements made by the decision-making process as a whole, in an attempt to estimate the average importance attached to various factors. Differences in the processes used by local child care systems and individuals are averaged out in the statistical analysis. The result is a model which tries to capture the average weights attached to the various independent variables by the average decision-making process.

Section 2 reviews the literature on the placement decision in the US, and Section 3 outlines the four estimation techniques used in this paper. Section 4 describes the data, while Section 5 has the results of re-estimating the K&D model using 2009 data and the four estimation techniques. In Section 6 data for 2009 is used to fit a new model, which is again estimated using the four techniques, and Section 7 has the conclusions.

Section snippets

Literature review

There have been eight previous national studies for the USA of the decision to place children at risk in out-of-home care. These studies use a wide range of independent variables, with little common ground between studies. Lindsey, 1991, April, Lindsey, 1992, September, Knott and Donovan (2010), Horowitz, Hurlburt, Cohen, and Zhang (2011), Barth, Wildfire, and Green (2006) and Rossi, Schuerman, and Budde (1999) identify inadequate financial resources as an important factor in the placement

Methodology

The aim is to quantify the association between the selected variables and the probability of a child being placed, where the dependent variable is either one or zero, i.e. placed or not placed. The type of placement that is appropriate for a particular child is outside the scope of this research, as is the factors determining the number of placements experienced by a particular child (see Eggertsen, 2008). A few authors have used multiple linear regression to estimate the factors associated

Data

The data is the NCANDS child file for 2009, which is described in NCANDS (2011).2 This contains all investigated reports of

Knott and Donovan model

K&D used logistic regression to analyse the NCANDS data for 2005. This offers the opportunity to replicate their study using 2009 data to investigate the stability and robustness of their results. Since K&D analysed only substantiated cases, for the purposes of this comparison, all unsubstantiated observations were removed. This reduced the data set to 259,264 observations. The results of a logistic regression using the same independent variables as K&D appear in Table 1.

In their study using

Analysis of the 2009 data

After including unsubstantiated cases, and adding substantiation as an additional independent variable, the development of a new 2009 model began with the K&D model. This was then modified by removing the insignificant variables, and adding some new variables which were thought to be of importance. So child alcohol abuse, child drug abuse and child learning problems were removed. The five separate child age variables were replaced with an under one year variable, and caregiver learning problems

Conclusions

The application of four different estimation techniques to the same data has found that, while there are many differences, there is a broad similarity in results across the four estimation techniques — logistic regression, probit analysis, discriminant analysis and ANN. This demonstrates that the results are robust to the estimation technique used, increasing confidence in the conclusions. The K&D analysis for 2005 was replicated using data for 2009 and, with the exceptions of age and sexual

References (46)

  • R.P. Barth et al.

    Placement into foster care and the interplay of urbanicity, child behaviour problems and poverty

    The American Journal of Orthopsychiatry

    (2006, July)
  • L.M. Berger et al.

    Out-of-home placement of children and economic factors: An empirical analysis

    Review of Economics of the Household

    (2004, December)
  • Bhatti-Sinclair, K. and Sutcliffe, C. (2012) Challenges in Identifying Factors Which Determine the Placement of...
  • C. Brooks

    Introductory econometrics for finance

    (2008)
  • B. Duncan et al.

    Economic incentives and foster care placement

    Southern Economic Journal

    (2007, July)
  • L. Eggertsen

    Primary factors related to multiple placements for children in out-of-home care

    Child Welfare

    (2008)
  • M. Ejrnæs et al.

    Risk factors of entry in out-of-home care: An empirical study of Danish birth cohorts, 1981–2003

    Child Indicators Research

    (2011, January)
  • E.M.Z. Farmer et al.

    Use and predictors of out-of-home placements within systems of care

    Journal of Emotional and Behavioural Disorders

    (2008, March)
  • E. Franzen et al.

    The epidemiology of out-of-home care for children and youth: A national cohort study

    British Journal of Social Work

    (2008, December)
  • C. Glisson et al.

    The role of specialty mental health care in predicting child welfare and juvenile justice out-of-home placements

    Research on Social Work Practice

    (2006, September)
  • R.M. Goerge et al.

    The entry of children from the welfare system into foster care: Differences by race

  • A. Grogan-Kaylor

    Who goes into kinship care? The relationship of child and family characteristics to placement into kinship foster care

    Social Work Research

    (2000, September)
  • I. Harpaz-Rotem et al.

    Out-of-home placement of children exposed to violence

    Children and Society

    (2008, January)
  • Cited by (31)

    • Decision-making on out-of-home care: The case of Flanders

      2023, Children and Youth Services Review
    • The relationship between foster care entries and high-dose opioid prescribing in California

      2019, Addictive Behaviors
      Citation Excerpt :

      To account for county-level trends in removal rates, we include county-specific linear time trends. As economic factors have been shown to affect removal risk (Bhatti-Sinclair & Sutcliffe, 2012; Rivaux, James, Wittenstrom, et al., 2008), the average weekly wage as reported in the Quarterly Census of Employment and Wages by the U.S. Bureau of Labor Statistics is included as a control. Parental criminal justice activity increases the probability of child removal, either due to increased risk of neglect or the need for children to be placed in out-of-home care while their parents are incarcerated.

    • The effects of economic factors on the out-of-home placement for maltreated children

      2017, Children and Youth Services Review
      Citation Excerpt :

      The characteristics of children and caregivers and the quality of caregiving in the family, which are micro-systems, have also been reported to be related to OOH placement. Prior research has documented that a child's age (Bhatti-Sinclair & Sutcliffe, 2012; Horwitz, Hurlburt, Cohen, Zhang, & Landsverk, 2011; Knott & Donovan, 2010), race (Bhatti-Sinclair & Sutcliffe, 2012; Knott & Donovan, 2010; Lindsey, 1992), and difficulties such as emotional, behavioral, and medical problems or drug abuse (Barth et al., 2006; Bhatti-Sinclair & Sutcliffe, 2012; Knott & Donovan, 2010) all predicted OOH placement. A caregiver's mental health difficulties or developmental limitations, medical problems, and substance abuse also have been shown to contribute to an increased probability of placement (Barth et al., 2006; Bhatti-Sinclair & Sutcliffe, 2012; Carter, 2009; Knott & Donovan, 2010; Zuravin & DePanfilis, 1997).

    View all citing articles on Scopus

    We wish to thank the referees of this journal for their helpful comments on a previous draft.

    1

    Tel.: + 44 118 378 6117.

    View full text