Relationship Transitions and the Risk for Child Maltreatment

Abstract

Family structure as a risk for child maltreatment has long been viewed as a static state in the child maltreatment literature. Drawing on data from the Fragile Families and Child Wellbeing Study, the author uses a series of individual fixed-effects models to investigate whether particular types of relationship transitions over children’s first decade of life are associated with increased risk for maternal and paternal child abuse and maternal neglect. Findings question and confirm a number of long-standing theoretical and empirical findings from the child maltreatment literature. Results indicate that transitions to being single are associated with increased risk for maternal child abuse and neglect. In addition, the frequency and severity of paternal harsh parenting may be closely linked with the nature of fathers’ relationship transitions. Last, results largely do not provide support for the theory that the presence of social (nonbiological) fathers increases mothers’ risk for engaging in child abuse or neglect.

Introduction

Child maltreatment remains a widespread and pernicious problem in the United States. In 2013, the most recent year for which data is available, state Child Protective Services (CPS) agencies received an estimated 3.5 million referrals for child maltreatment, involving approximately 6.4 million children or 47.1 referrals per 1,000 children in the United States. Of those cases referred to CPS agencies, approximately 2.1 million were determined to be possible cases of maltreatment and received a disposition (U.S. Department of Health and Human Services 2015).

Research on the ontology of child maltreatment has drawn on social-ecological models focusing on the multiple forces of individual, family, community, and societal factors that interact to contribute to the risk for child maltreatment (Belsky 1984). Other theories have also focused on the possible causal effect of poverty and differences in the determinants of the risk for child abuse and child neglect (Paxson and Waldfogel 2003; Schneider et al. forthcoming).

Although limited, a secondary literature has begun to explore the role of men in the risk for child maltreatment (Guterman and Lee 2005). Indeed, common across much of the extant literature is the prevailing notion that (1) children growing up with two biological parents are less likely to experience maltreatment than children growing up in a single-parent or stepparent household (Berger 2004); (2) child abuse and child neglect may have different antecedents, with neglect more highly correlated with poverty (Drake and Pandey 1996); and (3) compared with mothers, fathers are disproportionally responsible for child abuse given the amount of time they spend with children (Margolin 1992).

A growing literature indicates that relationship transitions may negatively affect parenting and children (Amato 2005; Fomby and Cherlin 2007; Lee and McLanahan 2015; Osborne and McLanahan 2007). Relationship transitions are a measure of relationship change, as opposed to marital status, which is a static measure of current relationship status. It is unclear whether or how relationship transitions might affect the risk for child maltreatment. In this article, I use data from the Fragile Families and Child Wellbeing Study (FFCWS) to test three related hypotheses (Reichman et al. 2001): (1) the nature of relationship transitions, as opposed to marital status, may be an important predictor of the risk for child maltreatment; (2) relationship transitions may play differing roles in men and women’s risk of child maltreatment; and (3) the risk for child abuse and child neglect may be differentially affected by relationship transitions.

In addition to testing the main hypotheses, I carry out two additional sets of analyses. First, I investigate mediating processes because if relationship transitions are predictive, it is important to understand whether the risk is the result of individual psychopathology or changes in economic resources. Second, I analyze the potential moderating role of child age, drawing on evidence indicating that younger children may be at greater risk of experiencing child maltreatment.

Background

Child Maltreatment

No single universally accepted definition of child maltreatment exists. “Child maltreatment”—the overall term for child abuse and neglect—is defined and measured differently by the federal government, state governments and agencies, researchers, and policymakers (Slack et al. 2004). The Child Abuse Prevention and Treatment Act of 1974 (CAPTA) was the first federal statute to recognize child maltreatment, defining it as “any recent act or failure to act on the part of a parent or caretaker, which results in death, serious physical or emotional harm, sexual abuse or exploitation, or an act or failure to act which present an imminent risk of serious harm.” Child abuse, then, is the serious physical or emotional harm to a child, and child neglect is a failure to act that puts children at serious risk of harm.

Although often linked, child abuse and child neglect are thought to be distinct from each other in a number of ways. Although child abuse has been the subject of the bulk of research in the literature, child neglect is much more prevalent. In 2013, 79.5 % of victims of child maltreatment were neglected, 18 % were physically abused, 9 % were sexually abused, and 8.7 % were psychologically maltreated (children may appear in multiple categories) (U.S. Department of Health and Human Services 2015). A range of literature has indicated that neglect may be particularly strongly associated with low-income status, insofar as it is often characterized by the insufficient provision of needed resources for children (Dubowitz et al. 2005; Martin and Walters 1982; Pelton 1978).

In the mid-1960s, child maltreatment began to be widely recognized as a social problem (Kempe et al. 1962). Early theories of the determinants of child maltreatment were largely drawn from psychiatric models in which individual psychopathology was thought to be the sole cause of child maltreatment. This psychiatric model focused on mothers’ use of drugs and alcohol, depression, and various personality disorders as the casual determinants of the risk for child maltreatment (Gil 1971). In the late 1970s, researchers began to recognize that psychopathology alone could not fully explain child maltreatment. Researchers such as Garbarino (1977), Bronfenbrenner (1979), Belsky (1984, 1993), Cichetti and Lynch (1993), and Sameroff (1998), proposed developmental-ecologically based theories that incorporated notions about the ways in which individual, family, community, and societal factors interact to increase or decrease the risk for child maltreatment.

Family Structure and Relationship Transitions

The last 50 years have been marked by a dramatic increase in the number of children who experience parental divorce, spend time in a single-parent household, live with a stepparent, or live with nonmarried cohabiting parents (Bumpass and Lu 2000). A growing share of children are born to unmarried parents (Bumpass and Lu 2000); approximately 50 % of children born to coresidential parents as well as two-thirds of children born to unmarried nonresident parents will experience the entrance or exit of a biological or social (nonbiological) father during childhood (Bzostek et al. 2012). The increasing likelihood that a child will experience one or more parental relationship transitions has spurred work investigating the associations among family instability, parenting quality, and child well-being.

Two existing sociological literatures focus on (1) the ways in which divorce is linked to negative changes in parenting and child well-being, such as increased child behavior problems, early delinquency, and harsh parenting (Amato 2005; Beck et al. 2010; McLanahan et al. 2013); and (2) the role of social (nonbiological) father entrances in families. This latter literature has largely found that social fathers’ presence does not provide overall benefits for children (Heatherington et al. 1992; Magnuson and Berger 2009).

A third literature contends that the type of transition (married to single, or single to married, for example) is less important than the number of transitions that a child experiences. Fomby and Cherlin (2007) outlined two hypotheses about the ways in which children may be affected by parental relationship transitions that captures the literature well. Their instability hypothesis contends that children are equally, or more, affected by the number of parental relationship transitions that they experience than by the type of transition. The selection hypothesis proposes that the negative effects of relationship transitions may simply be a by-product of the same antecedents that make some mothers more likely to experience multiple transitions. The instability hypothesis stems from theories indicating that transitions of any kind result in disruptions of the family system.

Child Maltreatment and Marital Status

In contrast to the broader sociological literature about the association between relationship transitions and parent and child well-being, the child maltreatment literature has focused almost exclusively on the role of marital status, both in research and in policy and practice. A number of studies have shown that children in single-parent families are more likely to experience child maltreatment. Sedlak and Broadhurst (1996), using data from the Third National Incidence Study of Child Abuse and Neglect (NIS-3), reported that children living in single-parent families were 77 % more likely to experience physical abuse and 87 % more likely to experience physical neglect than were children in two-parent families. In particular, low-income children living with a single mother are more likely to experience child abuse and neglect than those children living with both biological parents (Dubowitz et al. 1987; Gelles 1989; Gil 1971; Paxson and Waldfogel 1999, 2002, 2003; Waldfogel 1998).

Given prior evidence that child abuse and neglect are associated with single parenthood, one would expect that transitions to marriage or cohabitation with a child’s biological father would be associated with decreased risk of child abuse as mothers gain resources and help in parenting, and stress is reduced. Transitions to marriage or cohabitation with a social father might be associated with increased risk of child abuse because the addition of a new parent to the family system increases stress. Transitions to being single may be associated with increased risk of abuse as economic and parenting resources are diminished.

In a recent study examining the role of relationship transitions and harsh parenting, Beck and colleagues (Beck et al. 2010) found that coresidential and dating transitions are both associated with increased maternal stress and harsh parenting. However, theirs is among the few studies to investigate the role of relationship transitions in the risk for child abuse.

Theories and empirical evidence from the two literatures—relationship transitions and child maltreatment—are consistent in some ways and conflict in others. First, the relationship transitions literature would indicate that the exit of a biological father from the household would be associated with decreased child well-being (unless the parental relationship quality was particularly poor), and the entrance of a father would be associated with increases in child well-being (Osborne et al. 2012). However, little research has distinguished between entrances and exits of biological versus social fathers.

In contrast, the child maltreatment literature has indicated that marriage with biological fathers generally reduces the risk for maternal child maltreatment (with little distinction between marriage and cohabitation), but the presence of social fathers vastly increases the overall risk for child maltreatment (Berger 2004; Martin and Walters 1982). Consistent across the two literatures is that single motherhood increases the likelihood of poor child outcomes and child maltreatment alike.

Prior research about the determinants of child abuse and neglect has suggested that relationship transitions may influence the risk for each differently. Transitions to marriage or cohabitation may decrease the risk for child neglect because the presence of fathers may increase material resources and supervision of children. However, the presence of social fathers may increase the risk for neglect. For example, mothers may shift their attention away from the child and toward the new partner, or social fathers may withhold investments in nonbiologically related children (Coohey and Zhang 2006; Radhakrishna et al. 2001). To that end, I propose three related hypotheses:

• Hypothesis 1a: Mothers’ relationship transitions to marriage or cohabitation with the child’s biological father will decrease the risk for child abuse and neglect by mothers.

• Hypothesis 1b: Mothers’ relationship transitions to marriage or cohabitation with a social father will increase the risk for child abuse and neglect by mothers.

• Hypothesis 1c: Mothers’ relationship transitions to being single will increase the risk for child abuse and neglect by mothers.

The Role of Possible Maternal Mediators: Income, Material Hardship, Unemployment, Depression, Parenting Stress, and Social Support

A robust literature has demonstrated that low-income status is highly associated with the risk for child maltreatment (Steinberg et al. 1981). Prior empirical work has also indicated that maternal unemployment may be associated with a decrease in the risk for child maltreatment given that mothers are able to spend more time on childcare (Paxson and Waldfogel 2003).

Studies have found that parental depression is often closely linked to hostile and rejecting interactions with children (Belsky 1984), which may affect secure parent-child attachment (Crittenden and Ainsworth 1989) and hinder parental perspective-taking ability (Feshbach 1989), thus increasing the likelihood of both abuse and neglect (Carlson et al. 1989).

Parenting stress, or self-efficacy, has been similarly linked to the risk for child maltreatment (Slack et al. 2004). Stress may increase the likelihood of child maltreatment through parents’ feelings of ineffectual parenting (Brayden et al. 1992) or through difficulty in controlling emotions, thus resulting in harsher and more punitive punishment (Azar 2002).

Social isolation has been defined in a variety of ways. Theory and empirical research have found that families who are disconnected from peers, social organizations, and social and economic supports as well as families who move frequently, are more likely to maltreat than families that are highly integrated within a community (Coohey 1996). Lack of social support may contribute to child maltreatment through parents’ inability to access material and perceived support from family and peers (Gaudin et al. 1993).

In sum, the literature has indicated that a number of individual economic and psychological factors contribute to the risk for child maltreatment. Prior research has often suggested that these mechanisms may be particularly powerful in the association between single parenthood and the risk for child maltreatment (Paxson and Waldfogel 2003). However, to my knowledge, no empirical work to date has investigated the role of these primary mechanisms in the association between relationship transitions and the risk for child maltreatment.

If the association between relationship transitions and the risk for child maltreatment is explained by changes in economic resources, mental health, or social isolation, then perhaps these transitions themselves are less important in the risk for maltreatment than the accompanying changes in economic well-being and mental health. It may also be the case that parents who are more susceptible to relationship transitions are also more likely to be low-income or have mental health problems. In contrast, a finding of an association between different types of relationship transitions and the risk for child maltreatment that is robust to the inclusion of individual economic and psychological factors would provide further evidence that the nature of a relationship transition itself represents an important risk for maltreatment, above and beyond associated changes in contextual factors.

The preceding indicators might well be expected to have different directional influences, depending on the type of relationship transition experienced. Mothers’ transitions to marriage or cohabitation with the child’s biological father might be associated with increased household income as well as decreased material hardship and depression, resulting in a decrease in the risk for child maltreatment. In contrast, it may be that the potential mediators function differently when mothers’ transition to coresidential relationships with a social father. The presence of a social father in the household might lead to increased household income and decreased material hardship, but such presence may also be associated with increased parenting stress and potentially depression, resulting in an increase in the likelihood of child maltreatment. Last, mothers’ transitions to being single might be associated with increases in depression, parenting stress, social isolation, and economic hardship, culminating in an increase in the risk for child maltreatment. Thus, I propose the following hypothesis:

• Hypothesis 2: The association between relationship transitions and the risk for child maltreatment by mothers may be partly mediated by changes in economic and psychological well-being.

Biological Fathers

The role of fathers in the risk for child maltreatment, and relationship transitions in general, have received scant attention in the literature to date. A long-standing body of research stemming from social-biological theory has contended that the presence of stepfathers and nonbiologically related men in the household increases the risk for child abuse, potentially as a result of their decreased biological imperative to care for children who are not biologically related to them (Daly and Wilson 1980; Daly and Wilson 1996; Lightcap et al. 1982), or as a result of the incomplete institutionalization of the role of social fathers in the family (Cherlin 2009; Margolin 1992).

The vast majority of prior research investigating the role of fathers in the risk for child maltreatment has focused on the ways in which fathers’ presence, coparenting, or economic support influences the risk for maternal child maltreatment (Coohey and Zhang 2006; Guterman et al. 2009). However, a limited body of work has indicated that fathers are overrepresented as perpetrators of child maltreatment compared with mothers, relative to the amount of time that each spends in childcare (Lee et al. 2009).

Although research on fathers’ role in the risk for child maltreatment is limited and somewhat mixed, fathers’ probability of maltreatment is clearly correlated with access to the child. Fathers who coreside with their child’s biological mother (the most likely primary caregiver) are likely to spend more time with the child. Related research on father involvement in parenting indicates that fathers who are not coresidential are likely to form new partnerships and decrease their involvement with existing children (Edin and Nelson 2013). To that end, I offer three related hypotheses about how fathers’ relationship transitions might affect the risk of abuse of their biological children:¹

• Hypothesis 3a: Fathers’ relationship transitions to marriage or cohabitation with the child’s biological mother will increase the risk for paternal child abuse.

• Hypothesis 3b: Fathers’ relationship transitions to marriage or cohabitation with a social mother will decrease the risk for paternal child abuse.²

• Hypothesis 3c: Fathers’ relationship transitions to being single will decrease the risk for paternal child abuse.

The Role of Possible Paternal Mediators: Income, Material Hardship, Unemployment, Depression, Parenting Stress, and Social Support

Although the literature has identified a number of possible maternal characteristics that are likely to mediate the association between relationship transitions and the risk for child maltreatment by mothers, little research has identified characteristics that are unique to fathers that might be thought of as potential mediators (Guterman and Lee 2005). Therefore, I draw on the same set of possible maternal mediators outlined earlier in analyses of fathers’ risk for child abuse.

Fathers’ transitions to coresidence with the child’s biological mother might be associated with increased household income and decreased material hardship, depression, parenting stress, and social isolation, which might result in a decrease in the risk for paternal child abuse. In contrast, fathers’ transitions to being single might be expected to have the opposite effect, and it is perhaps unclear how fathers’ transitions to a relationship with a new partner might be affected. Thus, I propose the following:

• Hypothesis 4: The association between relationship transitions and the risk for paternal child abuse may be mediated by changes in economic and psychological well-being.

The Role of Child Age

In addition to the aforementioned indicators, evidence exists from the prior literature that the risk for maltreatment is highest before approximately age 5 and decreases thereafter (MacKenzie et al. 2014). Prior work has also found that children’s age at the time of parental relationship transitions may be important for child well-being (Amato and Keith 1991). This research suggests the following hypothesis:

• Hypothesis 5: The association between relationship transitions and the risk for child abuse and neglect may be moderated by child age.

Data

This study draws on data from the Fragile Families and Child Wellbeing Study (FFCWS), a rich longitudinal birth cohort study made up of nearly 5,000 families in 20 large U.S cities in 15 states between 1998 and 2001. FFCWS oversampled nonmarital births and, as a result, socioeconomically disadvantaged families (Reichman et al. 2001). Follow-up surveys were conducted when the focal child was approximately 1, 3, 5, and 9 years old.

Measures of Relationship Status

At each wave of the study, mothers and fathers were asked about their current relationship status with the child’s other biological parent. In addition, beginning when children were age 1, parents also reported about whether they were married to or cohabiting with a social parent (father or mother).

Measures of the Risk for Child Maltreatment

Parents in the FFCWS tend to use high levels of harsh parenting.³ Spanking and harsh parenting practices are often thought to be indicators of the risk for abuse when they occur frequently (Gershoff 2002). For this reason, I focus on changes in the prevalence of high-frequency harsh parenting.

Risk of Abuse

Beginning when children were approximately age 1, both mothers and fathers were asked how often in the past month they had spanked their child (every or nearly every day, a few times per week, a few times per month, never). I recode spanking to create two dichotomous indicators: (1) ever spanked, and (2) high-frequency spanking (a few times per week or more). Fathers were asked about spanking only if they had seen their child in the 30 days prior to the interview. For those fathers who had not seen their child, I assign a 0 for both indicators of spanking. Although generally not considered child abuse, the use of high frequency corporal punishment is associated with the risk for child abuse (Gershoff 2002).

Beginning when the focal child was age 3, mothers were also asked about their own physically and psychologically aggressive parenting behaviors drawn from the Conflict Tactics Scale for Parent and Child (CTSPC) (Strauss et al. 1998). This scale is frequently used as an indicator of the risk for child abuse. On this scale, physically aggressive parenting behaviors include: (1) hit child on the bottom with something like a belt, hairbrush, stick, or other hard object; (2) shook child; (3) spanked child on the bottom with your bare hand; (4) slapped child on the hand, arm, or leg; or (5) pinched child. Psychologically aggressive parenting behaviors include: (1) shouted, yelled, or screamed at child; (2) swore or cursed at child; (3) said you would send child away/kick out of the house; (4) threatened to spank or hit child but did not do it; or (5) called child “dumb,” “lazy,” or similar name. I recode these scales so that high-frequency physically and psychologically aggressive behavior is defined as aggressive behavior that occurred 11 or more times in the past year. Because mothers were asked to report on fathers’ aggressive parenting only when fathers were coresident, I am unable to draw on these questions in the analyses.

Risk of Physical Neglect

I follow Font and Berger (2015) in constructing a measure of physical neglect. Child neglect consists of many possible components. Physical neglect is most often defined as a caregiver’s act of omission that results in the failure to provide for the basic needs of a child, including nutrition, safe housing, medical care, or (sometimes) education. Drawing on questions asked of mothers at each wave (beginning at age 3), I construct a measure of physical neglect based on several distinct indicators, including whether (1) the child did not receive sufficient food, (2) the child did not receive needed medical care, (3) the family was homeless or doubled-up, (4) the household had utilities shut off or was physically unsafe according to an in-home observer, or (5) the child appeared to have poor physical hygiene according to an in-home observer. Because many of the indicators were zero-skewed, I dichotomize all scales by creating a cutoff point at the ninetieth percentile. I then sum the items and create an indicator of any evidence of physical neglect (Berger et al. 2013).

Risk of Supervisory/Exposure Neglect

I also follow Font and Berger (2015) in constructing a measure of supervisory/exposure neglect (beginning at age 3). Supervisory and exposure neglect is potentially more difficult to discern because the signs are less obvious than those of physical neglect or child abuse. Determination of risk for supervisory/exposure neglect is based on four questions asking whether (1) the child was left alone without an adult, (2) the child was exposed to parental substance abuse, (3) the child was exposed to domestic violence, or (4) the child was exposed to criminal activity. I again dichotomize all scales by creating a cutoff point at the ninetieth percentile. I then sum the items and create an indicator of any evidence of supervisory/exposure neglect (Berger et al. 2013).

Child Protective Services Involvement

In addition to the aforementioned indicators of the risk for child maltreatment, mothers were also asked at the 5- and 9-year follow-up surveys whether they had ever been contacted by Child Protective Services. Using this information, I construct wave-specific cumulative indicators of CPS involvement. This information is unfortunately likely quite limited because mothers likely underreported their own involvement with CPS (Berger et al. 2009).

Time-Varying Covariates and Mediators

In all models, I include measures of parent and child age at the time of the survey. I also seek to determine whether individual time-varying characteristics of parents mediate the main associations. The FFCWS contains a rich set of possible mediators measured at ages 1, 3, 5, and 9.

Material Hardship

An indicator of material hardship is based on parents’ self report of whether they could not pay their rent or mortgage, were evicted due to nonpayment, could not pay the full amount of their utilities bill, needed medical attention but could not afford it, had telephone service or gas or heating oil cut off because they could not afford it, or received free food or meals.

Unemployment

At each wave, mothers and fathers were asked whether they had worked a regular job for pay in the last week. Drawing on this information, I create a dichotomous indicator of current unemployment status.

Depression

I draw on 15 questions designed to assess Major Depressive Episodes (MDE) derived from the Composite International Diagnostic Interview – Short Form (CIDI-SF) (Kessler et al. 1998). Parents were asked about their feelings of unease and dissatisfaction, and the extent to which they no longer found pleasure in their usual hobbies and activities during the past year. Parents were coded for depression if they reported two weeks of symptoms that lasted one-half of the day, almost every day.

Parenting Stress

Parents were asked how strongly they agreed with four questions about feelings of being overwhelmed or discouraged by parenting responsibilities, scored on a four-point Likert scale (strongly disagree to strongly agree).

Social Support

I also test the role of maternal and paternal self-reported social support as measured by a series of questions about whether parents can rely on family or friends to loan them money, help with childcare, or provide a place to live.

Drug and alcohol use could also be thought of as possible mediators. However, because drug and alcohol use are distinct parts of child neglect, I do not include them here as mediators. I have included them in alternative specifications, and results are robust.

Analytic Strategy

I model the association between mothers’ and fathers’ relationship transitions and the risk for child abuse and neglect estimated with odds ratios from logistic regressions. The data are restructured as person-wave files, with approximately 15,122 possible observations for spanking, which draws on four waves of data. Rates of missingness were quite low. All variables had 6 % or less missing information. I use multiple imputation to impute missing information on covariates; however, I do not use imputed data for the dependent variables.⁴ Replication using listwise deletion produces very similar results.

To understand the association between relationship transitions and the risk for child abuse and neglect, I estimate a series of individual fixed-effects models. Individual fixed effects are particularly useful in this context for a number of reasons. First, using fixed effects removes time-invariant differences in families. Although some unmeasured factor that varies over time may influence selection into a relationship transition, the fixed-effects model limits the potential bias by controlling for time-invariant or fixed factors and may help to rule out what Fomby and Cherlin (2007) described as the “selection hypothesis” in relationship transitions. Second, the fixed-effects model is useful insofar as it relates within-parent change in relationships to changes in parenting behaviors that are a risk for child maltreatment. This specification relies only on within-parent variation to estimate the association between relationship transitions and the risk for child maltreatment. (As a result, Ns in tables reflect cases in which both the independent and dependent variable changed.)

I begin by estimating separate models of the association between mothers’ and fathers’ relationship transitions (married or cohabiting with the child’s other biological parent, and married or cohabiting with a social father/mother, with single as the reference category) and mothers’ and fathers’ spanking, controlling for parent and child age at each wave. Although I test both fathers’ self-reported spanking and mothers’ report of fathers’ spanking (with similar results), I present fathers’ self-reports because it seems plausible that mothers may not know the full extent of nonresident fathers’ parenting. In addition, although mothers provide some information about harsh parenting by social fathers, this information is quite limited, and I do not consider it here.

In consecutive models, I replace spanking with mothers’ and fathers’ high-frequency spanking, and then with mothers’ physical and psychological aggression. Last, I reestimate the preceding models for the association between maternal relationship transitions and physical and supervisory/exposure neglect. I do not estimate models for paternal neglect because the neglect measures rely heavily on information about the custodial parent and the child’s physical environment, and fathers are rarely the custodial parent. In addition, I estimate a second model estimating the association between transitions to being single and each of the outcomes, with all other relationship statuses as the reference group. Because mothers and fathers transition to being single from all other relationship statuses, it is informative to understand the association between transitions to being single and the risk for maltreatment in a separate model.

Next, I test the role of six possible parenting mediators: household income, material hardship, unemployment, depression, parenting stress, and social isolation. If the addition of these mental health and economic indicators reduces the association between relationship transitions and the risk for child maltreatment, it may indicate that changes in relationship status work through these various pathways rather than having a direct effect on the risk for maltreatment.

Last, in an effort to better understand the possible moderating role of child age, I interact relationship status and child age. The fixed-effects models assume that the effect of a relationship transition is the same for children, no matter the age at which it occurs. Interacting age and relationship status provides a check on this assumption.

Some of the prior literature has indicated that the number of transitions rather than type of transition that a child experiences may be more important for well-being. As a robustness check, I test an alternative measure of relationship transitions focusing on the number of relationships that parents report having been involved in. At the age 5 and 9 follow-up surveys, mothers were asked to report how many relationships and cohabitations lasting longer than one month had occurred since the last wave. I combine this information with parental reports about their current relationship status to test the number of transitions hypothesis. I do so by pooling this information across the waves and conducting similar individual fixed-effects models as described earlier.

Results

Descriptive Results

Table 1 shows the frequency with which mothers and fathers report using harsh parenting practices with their children, which may indicate a risk for child abuse. Both maternal and paternal spanking begins quite early: 26 % of mothers and 18 % of fathers reported having spanked their 1-year-old child, but that behavior declined after age 3. Notably, fathers were more likely than mothers to report using high-frequency spanking when children were 3 and 5 years old. In contrast, a larger percentage of mothers compared with fathers reported using high-frequency physically and psychologically aggressive parenting at all ages. A much larger percentage of mothers reported physical neglect of their child than supervisory/exposure neglect. In addition, the percentage of mothers and fathers reporting depression was fairly consistent across the waves (approximately 18 % for mothers and approximately 13 % for fathers). Mothers were younger, had lower incomes and more material hardships than fathers, and were more likely to be unemployed. Last, the share of mothers who were married or cohabiting with the child’s biological father was largely consistent across waves, while the percentage of mothers who were married or cohabiting with a social father increased as children aged.

Mothers’ Relationship Transitions and the Risk for Child Abuse: Transitions With Biological Fathers

Hypothesis 1a is that mothers’ relationship transitions to marriage or cohabitation with the child’s biological father will decrease the risk for child abuse by mothers. (I discuss Hypothesis 1a in terms of child neglect in the upcoming section, “Mothers’ Relationship Transitions and the Risk for Child Neglect: Transitions With Biological Fathers.”) I test this hypothesis by conducting a series of logistic regression models using individual fixed effects and a variety of possible indicators of the risk for child abuse.

Table 2 relates changes in marital status (in reference to being single) and the risk for child maltreatment. As shown in the table, mothers’ transitions to cohabitation with the biological father are associated with a 17 % decrease in the odds of spanking by mothers (odds ratio (OR) = 0.83) (Model 1). Mothers’ transitions to cohabitation with a social father are associated with 31 % increase in the odds of spanking by mothers (OR = 1.31). High-frequency spanking may be more related to the risk for child abuse than spanking overall. The results indicate that only transitions to cohabitation with a biological father are significantly associated with a decrease in the likelihood of high-frequency spanking by mothers (40 % decrease in the odds; OR = 0.60).

Next, I test two other possible indicators of the risk for child abuse by mothers: high-frequency physically and psychologically aggressive parenting. Table 2 shows that mothers’ transitions to cohabitation with the biological father are associated with a 42 % decrease in the odds of high-frequency physical aggression by mothers (OR = 0.58). Similarly, the table shows that mothers’ transitions to cohabitation with the biological father are associated with a 37 % decrease in the odds of high-frequency psychological aggression by mothers (OR = 0.63).

Mothers’ Relationship Transitions and the Risk For Child Abuse: Transitions With Social Fathers

Turning to Hypothesis 1b, I next investigate whether mothers’ relationship transitions to marriage or cohabitation with a social father increases the risk for child abuse by mothers. (I discuss Hypothesis 1b in terms of child neglect in the upcoming section, “Mothers’ Relationship Transitions and the Risk for Child Neglect: Transitions With Social Fathers.”) Table 2 shows that transitions to cohabitation with a social father are associated with a 31 % increase in the likelihood of spanking by mothers (OR = 1.31). However, I find no association between transitions to cohabiting with a social father and maternal high-frequency spanking, psychological aggression, or physical aggression.

Mothers’ Relationship Transitions and the Risk for Child Neglect: Transitions With Biological Fathers

Hypothesis 1a also asks whether mothers’ transitions to marriage or cohabitation with the child’s biological father will be associated with a decrease in the risk for child neglect by mothers. Table 2 shows results indicating that mothers’ transitions to marriage with a biological father are associated with a 34 % decrease in the odds of physical neglect and a 48 % decrease in the odds of supervisory/exposure neglect by mothers (OR = 0.66 and 0.52, respectively). I find similar results for mothers’ transitions to cohabitation—in particular, a 28 % decrease in the likelihood of physical neglect and a 33 % decrease in the odds of supervisory/exposure neglect by mothers (OR = 0.72 and 0.67, respectively).

Mothers’ Relationship Transitions and the Risk for Child Neglect: Transitions With Social Fathers

Hypothesis 1b also asks whether mothers’ transitions to marriage or cohabitation with a social father will increase the risk for child neglect by mothers. In contrast to this hypothesis, the results indicate that mothers’ transitions to marriage with a social father are associated with a 52 % decrease in the odds of physical neglect and a 55 % decrease in the odds of supervisory/exposure neglect by mothers. The results indicate no significant association between mothers’ transitions to cohabitation with a social father and maternal physical neglect, but a 21 % decrease in the odds of supervisory/exposure neglect by mothers (OR = 0.48, 0.45, and 0.79, respectively).

Mothers’ Relationship Transitions and the Risk for Child Maltreatment: Transitions to Being Single

In Table 3, I examine whether mothers’ transitions to being single are associated with increased risk for child abuse and neglect by mothers (Hypothesis 1c). In these models, the reference group is all other relationship categories given that parents transition from all marital statuses to being single. Although transitions to being single are not associated with increased risk for maternal spanking overall, they are associated with a 43 % increase in the odds of high-frequency spanking by mothers (OR = 1.43). However, mothers’ transitions to being single are not associated with high-frequency physical aggression by mothers and are only marginally associated with high-frequency psychological aggression by mothers.

Turning to child neglect, Hypothesis 1c also proposes that mothers’ transitions to being single will be associated with increased risk for child neglect by mothers. I find strong evidence in support of this hypothesis, with transitions to being single associated with a 36 % increase in the odds of physical neglect by mothers and 60 % increase in the odds of supervisory/exposure neglect by mothers (OR = 1.36 and 1.60, respectively).

Finally, albeit self-reported and thus likely to be an imperfect measure, I draw on mothers’ report of their own involvement with Child Protective Services (CPS). Overall, Table 3 provides strong evidence that transitions to being single are associated with increased risk of CPS involvement (OR = 1.95).

The Role of Possible Mediators: Mothers

In Model 2 of Tables 2 and 3, I test Hypothesis 2 by including potential mediators (mothers’ mental health, resources, or economic well-being). Although many of the mediators are frequently statistically significant, in no case does their inclusion in the model significantly mediate the association between relationship transitions and the indicators of the risk for child abuse or neglect by mothers.

Fathers’ Relationship Transitions and the Risk for Child Abuse: Transitions With Biological Mothers

In Hypothesis 3a, I propose that fathers’ relationship transitions to marriage or cohabitation with the child’s biological mother will increase the risk for paternal child abuse as fathers gain greater access to the child.⁵ Table 4 draws on fathers’ self-reported spanking. (In results not shown, I use mothers’ reports of fathers and find nearly identical results.) Fathers’ transitions to marriage with the focal child’s biological mother are associated with an approximately 2.2 times increase in the odds of paternal spanking compared with single fathers, and transitions to cohabitation are associated with a 79 % increase in the odds of spanking (OR = 2.17 and 1.79, respectively). Similarly, fathers’ transitions to marriage are associated with a 2.5 times increase in the odds of paternal high-frequency spanking, and transitions to cohabitation are associated with a 3 times increase in the likelihood of paternal high-frequency spanking (OR = 2.50 and 3.03, respectively).

Fathers’ Relationship Transitions and the Risk for Paternal Child Abuse: Transitions With Social Mothers

Next, I examine Hypothesis 3b—that fathers’ relationship transitions to marriage or cohabitation with a social mother will decrease the risk for paternal child abuse of the father’s biological child, as fathers become enmeshed in new relationships and families. However, I do not find any significant association between fathers’ relationship transition to marriage or cohabitation with a social mother and spanking or high-frequency spanking by fathers.

Fathers’ Relationship Transitions and the Risk for Child Abuse: Transitions to Being Single

Hypothesis 3c contends that fathers’ transitions to being single will decrease the risk for child abuse by fathers as they spend less time with and have less access to their child. Table 5 demonstrates support for this hypothesis, indicating that relative to all other marital statuses, fathers’ transitions to being single are associated with a 47 % decrease in the odds of paternal spanking. Similarly, transitions to being single are associated with approximately a 66 % decrease in the likelihood of high-frequency spanking by fathers (OR = 0.53 and 0.34, respectively).

The Role of Possible Mediators: Fathers

I test Hypothesis 4 by reestimating Model 1 in each of the prior tables, adding measures of paternal mental health, resources, and economic well-being. Although parenting stress and material hardship are frequently significant, the addition of these and other measures in no case mediates the association between relationship transitions and the indicators of the risk for paternal child abuse.

The Role of Possible Moderators: Child Age

I do not find evidence of significant interactions between child age and the risk for child abuse and neglect. Although the risk for maltreatment declines as children age, the risk associated with parental relationship transitions does not appear to vary depending on the age of the child.

Robustness Check: Number of Transitions

Table 6 in the appendix shows results from individual fixed-effects models testing the association between the number of cohabitations experienced by parents between Waves 3 and 9 and the risk for child maltreatment; Table 7, also in the appendix, shows similar results for the number of relationships. The results indicate that more cohabitations are associated with increased odds of supervisory/exposure neglect, but no other associations are significant.

Discussion

Relationship Transitions and the Risk for Maternal Child Abuse

The literature has largely not distinguished between married and cohabiting families in investigating the risk for child maltreatment. In fact, prior research has placed great emphasis on the protective qualities associated with marriage and coresidence (Berger 2004; Dubowitz et al. 1987). However, I find that while transitions to cohabitation decrease the risk for abuse by mothers (spanking and harsh parenting), transitions to marriage do not significantly affect the risk. One possible explanation may be that those who transition to cohabitation are moving from instability to greater stability, while those transitioning to marriage may be simply formalizing a relationship that already existed. This finding warrants further investigation.

In keeping with the extant literature, I find that mothers’ transitions to being single greatly increase the likelihood of maternal spanking and CPS involvement. However, I do not find evidence that transitions to being single significantly increase the risk for high-frequency physical or psychological aggression by mothers. A primary explanation for why being single may be associated with increased risk for child abuse has been that being single is often associated with economic hardship and life stressors, which in turn decrease parents’ ability for warm and responsive caregiving (MacKenzie et al. 2014); however, my findings are not mediated by similar indicators.

I find some partial support for the theory that the presence of unrelated men in the household may increase mothers’ risk for child abuse. However, this finding is limited to increases in spanking among mothers who transition to cohabiting with a social father. Overall, these results imply that the nature of the transition itself may play an important role in increasing or decreasing the risk for abuse.

Relationship Transitions and the Risk for Maternal Child Neglect

Although child neglect is much more common than child abuse (Dubowitz et al. 2005). the former has received much less attention in the research literature. As a result, much less is known about the determinants of child neglect beyond economic explanations.

I find evidence that transitions to marriage to the child’s biological father do in fact reduce the risk for both physical and supervisory/exposure neglect by mothers. Similarly, my results indicate that transitions to being single greatly increase the risk for neglect by mothers. The majority of CPS reports are likely to be related to child neglect; findings for CPS are therefore consistent with this result.

Importantly, these findings do not seem to be the result of the traditional economic explanation—perhaps indicating that the type of relationship transition itself may play an important role in the risk for maternal neglect. One might surmise, for instance, that transitions to marriage may decrease the risk for supervisory/exposure neglect as parents increase their coparenting abilities and supervision of children. The reverse might also easily be true insofar as transitions to being single may result in less ability to oversee children and manage daily parenting tasks.

In marked contrast to some aspects of social-biological theory, I find evidence that transitions to marriage with a social father greatly decrease the likelihood of maternal child neglect, and this finding is not mediated by measures of economic hardship. Perhaps marriage between a mother and a social father institutionalizes the social father’s role in the family to the degree that the risk for neglect is reduced, and a less-formal relationship would otherwise increase the risk for neglect.

Although maternal depression does not significantly mediate the association between relationship transitions and the risk for supervisory/exposure neglect by mothers, the large and significant odds ratios for this variable are compelling. The finding that depression is associated with maternal supervisory/exposure neglect but not with physical neglect may indicate that this type of neglect is less coupled to economic factors than outcomes such as physical neglect.

Relationship Transitions and the Risk for Paternal Child Abuse

No prior research has investigated the role of fathers’ relationship transitions and the risk for paternal child abuse. My study may prove particularly important given the recent focus on multipartnered fertility and father involvement in parenting after nonmarital births (Tach et al. 2010).

I find that transitions to coresidence (marriage or cohabitation) are associated with increased odds of spanking and high-frequency spanking by fathers. In contrast, transitions to being single are associated with decreased odds of spanking and high-frequency spanking by fathers. These results may indicate that the more access fathers have to children, the more likely they are to engage in riskier corporal punishment.

The Role of Child Age

Prior research in child maltreatment has indicated that younger children may be more likely to experience child maltreatment than older children, and may be more adversely affected by parental relationship dissolution than older children (Cherlin et al. 1998). I find no significant interactions between child age and relationship transitions in the risk for child maltreatment.

Robustness Check: Number Versus Type of Transition

Recent work by Lee and McLanahan (2015) investigating the effect of relationship transitions on child behavior suggested that the salience of number of transitions versus the type of transition may depend on the outcomes measured. Similarly, evidence from robustness checks in the present study indicates that models estimating the nature of relationship transitions may better capture influences on harsh and neglectful parenting than the number of transitions.

Limitations

This study has several limitations. First, the study does not capture substantiated cases of child maltreatment. Rather, it largely draws on parents’ self-reported parenting behaviors as part of the Conflict Tactics Scale. Second, the FFCWS measures paternal physical and psychological aggression only for resident fathers. As a result, I cannot use this information because this work is concerned with fathers’ relationship transitions. Third, although some of the measures of neglect exist for fathers, few nonresident fathers are the child’s primary caregiver, thus making it difficult to measure paternal neglect.

Conclusions

My findings both question and confirm a number of long-standing theoretical and empirical findings from the child maltreatment literature. The literature has long contended that being single is associated with the risk for maternal child abuse and neglect. First, although I find support for the contention that mothers’ transitions to being single are associated with increased risk of maternal neglect, I find limited support for the proposition that mothers’ transitions to being single are associated with increased risk for maternal child abuse. Second, I find that the frequency and severity of paternal harsh parenting may be closely linked with the nature of fathers’ relationship transitions. My results offer quite limited support for the theory that the presence of social fathers increases mothers’ risk for either child abuse or neglect. Last, this work reinforces the need for further investigations of fathers’ risk for child maltreatment and the ways in which it may differ from mothers’ risk.

Acknowledgments

This article has benefitted from very helpful advice and comments from the Demography reviewers, Jane Waldfogel, Jeanne Brooks-Gunn, and Ariel Kalil. I am grateful to the Doris Duke Foundation Fellowship for the Promotion of Child Wellbeing and Columbia University for generous support.

Appendix

Notes

References