Economic predictors of child maltreatment in an Australian population-based birth cohort

https://doi.org/10.1016/j.childyouth.2016.10.012Get rights and content

Highlights

  • Economic factors independently predicted all forms of child maltreatment.

  • Poverty and parental unemployment were the strongest determinants of maltreatment.

  • An estimated 27.3% of child maltreatment was attributable to economic factors.

  • Physical abuse, sexual abuse, and exposure to domestic violence were most sensitive.

Abstract

A correlation between socioeconomic disadvantage and child maltreatment has long been observed, but the drivers of this association are poorly understood. We sought to estimate the effects of economic factors on risk of child maltreatment after adjusting for other known influences using the Australian Temperament Project, a population-based birth cohort of 2443 individuals and their parents. We used logistic regression to estimate associations of childhood economic factors (parental education, occupation, and unemployment; type of housing; and retrospective perception of poverty) with retrospective reports of perceived child maltreatment (physical abuse, sexual abuse, emotional abuse, neglect, and witnessing of domestic violence), controlling for demographic factors, parental mental health and substance use, and child health. We then used these estimates to approximate the proportions of child maltreatment—population attributable fractions—that are theoretically preventable by addressing childhood economic disadvantage. Economic factors were associated with all types of child maltreatment. For the most part, these associations diminished only partially when controlling for noneconomic confounders, supporting hypotheses of causal relationships. Jointly, economic factors were significant predictors of physical abuse, sexual abuse, and witnessing of domestic violence but not of emotional abuse or neglect. Retrospective perceptions of childhood poverty were, in particular, strongly associated with most forms of child maltreatment but not with sexual abuse after accounting for other economic factors. We estimated that 27% of all child maltreatment was jointly attributable to economic factors. These findings suggest that strategies that reduce economic disadvantage are likely to hold significant potential to reduce the prevalence of child maltreatment.

Introduction

The association between child maltreatment and poverty has been well-established in the literature, despite initial concerns of detection bias and victim-blaming (Pelton, 1978, Zellman, 1992). The reasons for the association, however, are not well-understood (Berger & Waldfogel, 2011). In their recent review of the literature on poverty and child maltreatment, Drake and Jonson-Reid (2014) identified many likely contributors to the relationship between poverty and maltreatment but noted that the causality of the relationship has a particularly poor underpinning in both theory and empirical research. By modelling the temporal contribution of potential causal influences, the present study aims to provide empirical estimates of the causal contribution of parent- and family-level economic factors to the risk of child maltreatment.

In reviews of the risk and protective factors for child maltreatment, researchers have identified a range of social and environmental factors, many of which center on socioeconomic disadvantage (Brown et al., 1998, Stith et al., 2009, U.S. Department of Health and Human Services, 2015). Factors identified as being important include parent-level factors (e.g., low education and income) and broader social or structural factors such as income inadequacy, unemployment levels, social isolation, inadequate housing and homelessness, and poor access to resources (child care, welfare services, schools, etc.), exposure to racism or discrimination, and stressful life events (Lamont & Price-Robertson, 2013). The largest group of risk factors associated with the occurrence of abuse and neglect relate to parental characteristics that prevent or interfere with good parenting skills, appropriate monitoring, and affective responses to children and their changing developmental needs.

Such social and economic factors are, however, highly interrelated conceptually and empirically. One distinction that can be drawn is between factors that are fixed (age, race/ethnicity, and gender, etc.—factors that might be considered “demographic”), and factors that may be more amenable to change (education, unemployment, poverty, etc.). In the subsequent analysis, parental education, occupation, unemployment, housing, and poverty are considered jointly as potentially modifiable factors contributing to economic disadvantage. A further distinction is offered, however, between poverty—perhaps a purer measure of economic disadvantage—and the remaining indicators, which are more closely related to social status and arguably more intrinsic characteristics of the parents than are household wealth or income.

Poverty has been postulated to affect child maltreatment through a range of mechanisms, including limiting parental capacity to provide for the needs of their children (food, shelter, medical care, etc.), increasing parental stress, reducing incentives for parents to invest their time and money in child-rearing, and reducing alternatives for discipline (Berger & Waldfogel, 2011). Whether and how much parents work may directly influence the amount of time that parents spend with children and thus the opportunity for exposure to poor parenting of any type, and may greatly increase the psychological stress that parents are exposed to. Conger and Donnellan (2007) offer a theoretical framework through which socioeconomic disadvantage may influence parenting behavior and child well-being. Family economic pressures act as stressors that increase parental conflict and inhibit nurturing and involved parenting and increases the propensity for harsh parenting behavior in their model. If economic factors contribute to destabilization of parental relationships or increased rates of parenthood outside of stable relationships, this would also increase the opportunity for children to be maltreated by their parents' associates (boyfriends, step-parents, etc.).

Estimating the causal effect of economic factors on child maltreatment is important because it provides guidance regarding how best to intervene. If the relationship between economic factors and child maltreatment is not causal, then addressing economic disadvantage will have little effect in terms of preventing child maltreatment. On the other hand, if the relationship between economic factors and child maltreatment is causal, then intervention to address economic disadvantage is likely to reduce the prevalence of child maltreatment and policies that increase economic disparities may exacerbate the problem.

In this analysis, we adopt an epidemiological approach to causal inference that is rooted in the potential outcomes framework (Glass, Goodman, Hernán, & Samet, 2013). This approach focuses on the differences in potential outcomes (in this case, child maltreatment) that would occur under scenarios that differ only with respect to the distributions of certain risk factors (economic disadvantage). If changing only the risk factor will change the outcome, then the relationship can be said to be causal. However, with outcomes such as child maltreatment and risk factors like economic disadvantage, true experiments can be difficult to construct and we must often rely on observational data to test the causality of these relationships or estimate their strength.

The main limitation of observational data as compared with experimental data is that there are often differences between people that are associated with both the exposure and outcome in question. Such differences confound the observed association, making it appear weaker or stronger than would result from a causal effect alone. Addressing confounding is therefore central to causal inference in observational data (Glass et al., 2013).

For something to confound the relationship between economic disadvantage and child maltreatment, it must either cause or have a common cause with both. As such, confounders are mostly limited to demographic factors (family size and structure, ethnicity, parental age, etc.) and parental characteristics (particularly mental health, substance use, and parental history of child maltreatment). Child health problems may also affect the risk of maltreatment while placing additional economic pressures on parents (Font & Berger, 2015) but the relationship between child health and maltreatment is likely to be bidirectional. Domestic violence is a well-established risk factor for child maltreatment that is associated with economic disadvantage, but it can itself be considered a form of psychological or emotional abuse (James, 1994, Kitzmann et al., 2003) and this is how we conceptualize it here.

While experimental studies of the effects of economic factors on risk of child maltreatment are rare, occasional opportunities arise in the course of changes to things like income support programs. New programs or changes may be rolled out incrementally, producing experimental conditions in which direct comparison can be made between groups receiving the new and old services. Sometimes, programs can even be rolled out in a randomized manner to ensure comparability of the treatment groups and facilitate evaluation of the program. This was the case in the study by Cancian, Yang, and Slack (2013), which found that an exogenous increase in the proportion of child support payments that was distributed to resident parents, thus increasing their income, was associated with decreased screened-in child protection notifications regarding their children.

A related form of ‘natural’ or ‘historical’ experiment can occur when exogenous factors (factors external to the parent-child relationship) are suddenly changed or interrupted. Population-level economic factors have few theoretical pathways through which their relationship to child maltreatment can be confounded, reducing the need to collect or model data on large numbers of variables simultaneously. For example, Wildeman and Fallesen (2016—this issue) found that a substantial reduction in a specific type of Danish welfare payment increased risk of out-of-home placement by 25%. Similarly, Schneider, Waldfogel, and Brooks-Gunn (2016—this issue) linked macroeconomic indicators of the Great Recession (the American experience of the Global Financial Crisis) to measures of behavioral approximations of physical abuse and neglect taken over the corresponding period in a population-based birth cohort. They found that there were direct effects of the Great Recession on risk of behaviorally approximated physical abuse but no effects or weak protective effects on risk of behaviorally approximated neglect. At the same time, using state-level child protective services data from the U.S., Raissian and Bullinger (2016—this issue) found that increases to the state minimum wage reduced reports of child neglect.

Outside of true and ‘natural’ experiments, observational investigation of the causes of child maltreatment is generally restricted by the limited availability of large-scale epidemiological data sets that contain a sufficient range of postulated risk factors to be modelled simultaneously (Munro, Taylor, & Bradbury-Jones, 2013). As most of the causes of child maltreatment operate at or through the level of the parents, data collection must span multiple generations and long periods of time. There are few prospective cohort studies that collect such broad information over these periods. In the Mater–University of Queensland Cohort Study, Martin et al. (2011) reported an analysis of the effects of maternal economic and noneconomic risk factors (measured early in the life of their offspring) on sexual abuse as reported by offspring in early adulthood. Maternal education recorded during pregnancy, and family income (recorded when the child was 6 months old), were associated with penetrative but not non-penetrative sexual abuse. Other risk factors included in multivariate analyses included maternal age, marital status, smoking and alcohol consumption, mental health, and attitudes towards the baby. They found that most of the correlation between family income and penetrative sexual abuse was accounted for by other risk factors, while the association with maternal education changed little, which suggested some causal role for this factor. Using data from the National Longitudinal Survey of Youth, Berger, 2004, Berger, 2007 found that higher levels of income were independently associated with improved parenting practices, particularly in families that had experienced parental separation. Using data from the National Longitudinal Study of Adolescent to Adult Health (Add Health), Hussey, Chang, and Kotch (2006) observed associations between family income and neglect and sexual abuse, and between parental education and physical neglect and physical abuse, that were robust to adjustment for other sociodemographic variables. In the Christchurch Health and Development Study, family income and family living standards were strongly associated with risk of physical abuse, but were dropped from automated variable selection models in multivariate analyses, while associations with maternal education and parental occupation were not statistically significant (Woodward & Fergusson, 2002). Most other large cohorts with data on child maltreatment have either not collected detailed information on economic risk factors (e.g., the Adverse Childhood Experiences Study) or have not explored the relationship using multivariate analysis (e.g., Brown et al., 1998).

Trials and evaluations of family support programs can also sometimes provide data that are useful for analyzing the causes of child maltreatment. Using child maltreatment data collected by nurse home visitors in a trial of a support program for families with preterm, low birth weight infants, Berger and Brooks-Gunn (2005) found that socioeconomic factors predicted risk of child maltreatment after controlling for a rich set of potential confounders, including parental knowledge and behaviors that were themselves closely related to child maltreatment.

Studies using linked administrative data can provide some insights, but usually only limited information will be available about risk factors. O'Donnell et al. (2010) reported one of the best examples of administrative data research, which involved linking child protection data to a birth registry and perinatal data collection. The study obtained information about family demographics and child health, linked to hospital and mental health services, to obtain information about parental mental health, substance use, and assault-related injuries. However, information about economic factors in their study was limited to a single measure of the relative socioeconomic disadvantage in the area of residence at the time of the child's birth. They found that half of the correlation between area-level relative socioeconomic disadvantage and substantiated child maltreatment in non-Indigenous Australians was due to confounding by other factors, leaving half to be due to causal effects or residual (unmeasured) confounding. After adjustment for confounding, the most disadvantaged 8% (according to the area-based measure of socioeconomic disadvantage) were 5.4 times as likely to have substantiated child maltreatment compared with the least disadvantaged 8%. In a similar analysis of data from the U.S., Lee and Goerge (1999) found that the association between community level of poverty and alleged child maltreatment was essentially unaffected by adjustment for parental age, ethnicity, birth order, child gender, or region of birth, although these estimates may be biased by a larger number of omitted variables. Berger et al. (2015) linked administrative data on home foreclosures to child protection data, and observed increased risks of child protection involvement in the periods immediately before and after home foreclosure.

Berger (2005) reported an analysis of cross-sectional data from the (American) 1985 National Family Violence Study, which included an extensive range of parent- and state-level risk factors for child maltreatment and parental report of violence towards children. He found that economic risk factors (family income, low parental education, state unemployment rate, and state urbanization) had significant positive effects on rates of physical abuse, after adjusting for other factors (including demographics, parental mental health and substance use, parental exposure to domestic violence as adults or children, and parental experience of physical abuse). These effects applied in single-parent families but not in two-parent families. However, one of the main limitations of such cross-sectional studies is the lack of temporality in the measures of risk and effect.

This study attempts to estimate the effects of economic disadvantage on different forms of child abuse and neglect in a population-based birth cohort, the Australian Temperament Project (ATP). The ATP has prospectively collected data on parent- or family-level economic factors, social factors, mental health and substance use, and child health and temperament and retrospectively self-reported exposure to child maltreatment. The robustness of economic factors to confounding by other risk factors is assessed using multivariate logistic regressions. The contribution of economic factors to the prevalence of child maltreatment is assessed using population attributable fractions (PAF). Attributable risk analysis compares different scenarios: the present distributions of risk factors, and hypothetical alternatives in which certain risk factors have been removed or had their distributions altered. In the present context, a PAF provides an estimate of the proportion of child maltreatment that would be avoided if policy or intervention were to reduce economic disadvantage, assuming that the estimated coefficients of regression models are causal in nature, an admittedly tenuous assumption.

Section snippets

Participants

The ATP has surveyed the families of 2443 children, over 15 waves of data collection since enrollment at the age of 4–8 months in 1983 (baseline). Since baseline, the cohort has been expanded to include some additional twins who were excluded from this analysis. Methods pertaining to the collection of data and sample characteristics have been previously reported (Prior et al., 2000, Sanson and Oberklaid, 1985, Vassallo and Sanson, 2013). Sampling was stratified at the local government area to

Analysis

Substantial amounts of missing data are ubiquitous in cohort studies with extended follow-up such as the ATP. Of particular concern in this instance was the relationship between child maltreatment, loss to follow-up, and nonresponse. When questions about child maltreatment were asked in early adulthood, about two-fifths of the cohort had been lost to follow-up and another fifth did not return the questionnaire. Exploratory analyses indicated associations between child maltreatment and the

Descriptive statistics

Participant characteristics are presented in Table 2. After multiple imputation, reported exposure to child maltreatment ranged from 6.2% (emotional abuse = “very true”) to 18.9% (emotional abuse = “somewhat true”) and was 37.2% for any maltreatment. Of those who reported maltreatment, 44.4% reported multiple types. Among those who reported witnessing domestic violence, 82.1% reported at least one other form of maltreatment. The prevalence of economic risk factors varied greatly (from 9.3% to

Summary of findings and relation to previous studies

In summary, we found evidence that parent- and family-level economic disadvantage increased young adults' retrospective reports of child maltreatment, even after controlling for a rich set of confounders. These findings support hypotheses of causal effects of economic factors on risk of child maltreatment—particularly effects of poverty and parental unemployment—although we cannot be sure they do not reflect omitted variable bias. Poverty remained a strong predictor of most types of

Acknowledgements

The ATP study is located at the Royal Children's Hospital in Melbourne and is a collaboration between Deakin University, the University of Melbourne, the Australian Institute of Family Studies, the University of New South Wales, the University of Otago (NZ), and the Royal Children's Hospital; further information available at www.aifs.gov.au/atp. Funding for this analysis was supported by a PhD scholarship from the University of South Australia, and the South Australian Health Economics

References (55)

  • J. Amos et al.

    Understanding maltreating mothers: A synthesis of relational trauma, attachment disorganization, structural dissociation of the personality, and experiential avoidance

    Journal of Trauma & Dissociation

    (2011)
  • Australian Institute of Health and Welfare

    Child protection Australia: 2013–14

    (2015)
  • L.M. Berger

    Socioeconomic factors and substandard parenting

    Social Service Review

    (2007)
  • L.M. Berger et al.

    Socioeconomic status, parenting knowledge and behaviors, and perceived maltreatment of young low-birth-weight children

    Social Service Review

    (2005)
  • L.M. Berger et al.

    Economic determinants and consequences of child maltreatment

  • L.M. Berger et al.

    Home foreclosure and child protective services involvement

    Pediatrics

    (2015)
  • S. Berzenski et al.

    A multidimensional view of continuity in intergenerational transmission of child maltreatment

  • L. Broom et al.

    Opportunity and attainment in Australia

    (1976)
  • P.L. Brotherton et al.

    Development of an australian index of social class

    Australian Psychologist

    (1979)
  • M. Cancian et al.

    The effect of additional child support income on the risk of child maltreatment

    Social Service Review

    (2013)
  • R.D. Conger et al.

    An interactionist perspective on the socioeconomic context of human development

    Annual Review of Psychology

    (2007)
  • P.M. Crittenden

    Child neglect: Causes and contributors

  • R. Crutchfield et al.

    Poverty and violence

  • P. Delfabbro et al.

    Longitudinal trends in child protection statistics in south Australia: A study of unit record data

    Children Australia

    (2010)
  • J.C. Doidge

    Responsiveness-informed multiple imputation and inverse probability-weighting in cohort studies with missing data that are non-monotone or not missing at random

  • J.C. Doidge et al.

    Adverse childhood experiences, non-response and loss to follow-up: Findings from a prospective birth cohort and recommendations for addressing missing data

    (2015)
  • B. Drake et al.

    Poverty and child maltreatment

  • Cited by (53)

    • Reconsidering the best interests of the child construct

      2024, Children and Youth Services Review
    • How does COVID-19 impact intrafamilial child sexual abuse? Comparison analysis of reports by practitioners in Israel and the US

      2021, Child Abuse and Neglect
      Citation Excerpt :

      Economic stress and unemployment are central stressors in family lives and often have negative consequences on parents’ abilities to meet their children’s basic needs (Conger & Conger, 2002; Conrad-Hiebner & Byram, 2020). The link between poverty and child maltreatment was discussed by various researchers (Doidge et al., 2017; Slack et al., 2011), wherein food and housing insecurity are observed to contribute to child abuse and neglect (Marcal, 2018; Slack et al., 2011; Wood, Pecker, Russo, Henretig, & Christian, 2012). COVID-19 generated a worldwide economic crisis, with millions of people around the globe suddenly unemployed when restaurants, entertainment, and commercial venues were closed with little notice.

    View all citing articles on Scopus
    View full text