Adolescent delinquency has impactful and long-lasting negative effects on development and society (Roberson & Azaola, Reference Roberson and Azaola2021;Thornberry & Krohn, Reference Thornberry and Krohn2000). Adolescents who engage in delinquent behaviors (i.e., behaviors that are illegal, carry risk of arrest, or if committed by an adult would violate criminal law) are more likely to perform poorly in school (Robison et al., Reference Robison, Jaggers, Rhodes, Blackmon and Church2017), abuse substances (Rocca et al., Reference Rocca, Verde and Gatti2019), and engage with delinquent peers (Walters, Reference Walters2020). As adults, they are more likely to be unemployed, arrested, and incarcerated, have poor physical and mental health, and engage in antisocial behavior (Carter, Reference Carter2019; Drury et al., Reference Drury, DeLisi and Elbert2020; Kim et al., Reference Kim, Kim, Oh, Lippert and Subramanian2020).
Researchers have found that exposure to parent-, peer-, school-, community-, and family-level risk factors predict adolescent delinquency (Roberson & Azaola, Reference Roberson and Azaola2021). Some of the most significant risk and protective factors for delinquency reside in the family system (Labella & Masten, Reference Labella and Masten2018; Salmanian et al., Reference Salmanian, Mohammadi, Hooshyari, Ahmadi, Khaleghi and Alavi2021). Studies have examined family risk as a set of characteristics of the home environment (e.g., chaotic home environment, Jacobsen & Zaatut, Reference Jacobsen and Zaatut2022) and of the mother (e.g., maternal distress, Choe et al., Reference Choe, Olson and Sameroff2013). In recent years, researchers have focused on children’s exposure to adverse childhood experiences (ACEs; e.g., child abuse, domestic violence; Jones & Pierce, Reference Jones and Pierce2021; Perez et al., Reference Perez, Jennings and Baglivio2018). The ACEs research has made important contributions to our understanding of risk for delinquency. However, it has not always distinguished the source of the risk — mother, father, another family member, or overall family that contributes to the occurrence of the risk. In particular, few studies of adolescent delinquency include fathers’ risk (Simmons et al., Reference Simmons, Steinberg, Frick and Cauffman2018). This is a significant gap considering that fathers’ and mothers’ risk in heterosexual parental relationships independently contribute to children’s outcomes (for review, see Cabrera et al., Reference Cabrera, Fagan, Wight and Schadler2011). Research that focuses on low-income father-level risk (in addition to mother-, child-, and overall family-level risk) is needed because low-income men tend to experience more risk factors and reduced father involvement with children than higher income fathers (Kotila & Kamp Dush, Reference Kotila and Kamp Dush2013).
Framed within a risk and resilience perspective (Masten, Reference Masten2018) and Sameroff’s (Reference Sameroff and Sameroff2009) transactional model of development, this study addresses these gaps by including father-level risk in addition to mother-, child-, and family-level risks, as predictors of adolescent delinquent behavior. In addition, the study examines how fathers’ and mothers’ relationships with children, parental monitoring, and children’s self-control may explain the association between all sources of risk and youth behaviors. Using data from the Future of Families and Child Wellbeing Study (FFCW), we address the following research questions: (1) are low-income fathers’, mothers’, children’s, and family-level risk factors during early childhood associated with adolescent delinquent behavior, (2) does child self-control in middle childhood mediate the associations between fathers’, mothers’, and family risk in early childhood and adolescent delinquent behavior, and do quality of parent’s relationships with children and parental monitoring in middle childhood mediate the association between child risk in early childhood and delinquent behavior, and (3) do parents’ engagement with children in early childhood, quality of parent’s relationships with children and parental monitoring in middle childhood, and child sex at birth moderate the associations among fathers’, mothers’, children’s, and family risk during early childhood and adolescent delinquent behavior? The focus of this study is on risk factors early in children’s lives because early risk may be especially predictive of delinquent behavior in adolescence (see Murray et al., Reference Murray, Irving, Farrington, Colman and Bloxsam2010; Trentacosta et al., Reference Trentacosta, Hyde, Goodlett and Shaw2013).
Theoretical foundation
The current study is guided by Masten’s (Reference Masten2018) risk and resilience perspective and transactional theory (Sameroff, Reference Sameroff and Sameroff2009). The risk perspective maintains that children’s development is compromised by exposure to adverse psychological or social factors because they increase the likelihood of poor outcomes (Masten, Reference Masten and Zelazo2013). In contrast, the resilience perspective posits that the negative impact of risk on development can be mitigated by protective factors, which can increase a child’s adaptive functioning (Masten, Reference Masten and Zelazo2013). Risk studies have sometimes examined one or two risk factors (e.g., incarceration) and at other times a constellation of risk factors (e.g., cumulative risk) in relation to child outcomes. Current thinking about risk suggests that the higher the number of risk factors present in a child’s life, the greater the probability of delinquent acts (LoBraico et al., Reference LoBraico, Bray, Feinberg and Fosco2020; Reingle et al., Reference Reingle, Jennings and Maldonado-Molina2012). Sameroff and MacKenzie (Reference Sameroff and Mackenzie2003) argue that assessing a constellation of risk factors best captures the contextual complexities in which children develop. Researchers have also debated whether it is best to examine this constellation of risk factors separately or as additive composites in relation to outcomes (Evans et al., Reference Evans, Li and Whipple2013). In the additive model, each risk factor is given equal weight and then summed to construct a composite of risk. Although the additive model does not distinguish among the effects of individual risk factors, it is advantageous because type of risk may be less important than the total number of risk factors (Evans et al., Reference Evans, Li and Whipple2013; Trentacosta et al., Reference Trentacosta, Hyde, Goodlett and Shaw2013). Moreover, children’s exposure to various forms of risk tend to co-occur rather than exist in isolation (Ayoub et al., Reference Ayoub, O’Connor, Rappolt-Schlictmann, Vallotton, Raikes and Chazan-Cohen2009).
Current theory also points to the importance of examining multiple sources of cumulative risk in relation to child outcomes (Masten, Reference Masten and Zelazo2013; Sameroff & Mackenzie, Reference Sameroff and Mackenzie2003). Risk can occur at the individual, family, community, and societal levels (Masten, Reference Masten and Zelazo2013). The current study focuses on cumulative individual and family risk. Individual risk refers to characteristics of the person (i.e., fathers, mothers, children) that may increase the likelihood of poor child outcomes (Luthar, Reference Luthar1991). Family risk refers to characteristics and properties of the family as a whole that may be associated with poor child outcomes (e.g., household poverty, see LoBraico et al., Reference LoBraico, Bray, Feinberg and Fosco2020). A number of studies have examined multiple sources of cumulative risk in relation to adolescent delinquent behavior. For example, Trentacosta et al. (Reference Trentacosta, Hyde, Goodlett and Shaw2013) found that multiple cumulative risk domains (early child sociodemographic risk, caregiving risk, and child attributes risk) significantly predicted a disruptive behavior diagnosis in adolescence. Using FFCW data, Miller et al. (Reference Miller, Machlin, McLaughlin and Sheridan2021) found that multiple sources of cumulative risk, including caregiving deprivation and various threats such as interpersonal violence, from birth to age 3 significantly predicted higher levels of internalizing and externalizing behavior at age 15. In a study of 466 European American and 100 African American boys and girls from a broad range of socioeconomic levels, Deater-Deckard et al. (Reference Deater-Deckard, Dodge, Bates and Pettit1998) found that cumulative risk in multiple domains (child, sociocultural, parenting, peer-related) at age 5 predicted 19%–32% of the variance in children’s externalizing behaviors in middle childhood. There is growing recognition that fathers’ risk factors contribute to adolescent delinquency, yet, these studies examine father risk in isolation from other sources of risk (e.g., Simmons et al., Reference Simmons, Steinberg, Frick and Cauffman2018; Swisher & Roettger, Reference Swisher and Roettger2012; Yan et al., Reference Yan, Schoppe-Sullivan and Beauchaine2021). We are not aware of any studies that have included cumulative father-level risk in addition to mother-, child-, and family-level risk in studies of adolescent delinquency.
Sources of risk
Fathers’ and mothers’ risk
Research conducted with low-income families indicates that key parental risk factors are similar for fathers and mothers and include criminal justice involvement, depressive symptoms, substance and alcohol abuse, poor health, unemployment, and low education (Fagan, & Wildfeuer, Reference Fagan and Wildfeuer2022). Researchers using FFCW data found that mothers of adolescents who self-report delinquent behaviors are more likely to have been charged with breaking the law than mothers of adolescents who have not (Marçal & Maguire-Jack, Reference Marçal and Maguire-Jack2021). Studies of low-income families have shown positive associations among maternal depression, parents’ drug and alcohol abuse, and adolescents’ delinquent behavior (Jackson, Reference Jackson2013; Marçal, Reference Marçal2021; Wickham et al., Reference Wickham, Senthilselvan, Cameron Wild, Hoglund and Colman2015). Parents’ poor physical health has been identified as a risk factor for youth conduct disorder because it increases stress within low-income families (BeLue et al., Reference BeLue, Halgunseth, Abiero and Bediako2015). Parental unemployment was also a significant correlate of working-class fathers’ and mothers’ stress in parenting (Nomaguchi & Johnson, Reference Nomaguchi and Johnson2016) as well as adolescent delinquency (Fleisher, Reference Fleisher1963). Finally, in a longitudinal study of 856 third graders in a semi-rural county in New York State, Dubow et al. (Reference Dubow, Boxer and Huesmann2009) found significant associations between parents’ low education levels and long-term child outcomes including youth aggression and low educational motivation.
Children’s risk
Transactional theory suggests that children are active agents of their own development and, as such, are not only influenced by their parents, but they also influence their parents (Sameroff, Reference Sameroff and Sameroff2009). Research reviews indicate that key child-level risk factors for delinquency include speech and language problems, attention problems, aggression, depression, anxiety, and poor health (Shader, Reference Shader2003). Researchers have reported high rates of language impairments in conduct disordered and incarcerated youth (Brownlie et al., Reference Brownlie, Young, Beitchman, Young, Atkinson, Wilson, Douglas, Escobar and Johnson2004). Meta-analyses indicated significantly increased rates of attention-deficit/hyperactivity disorder in juvenile offenders (Baggio et al., Reference Baggio, Fructuoso, Guimaraes, Fois, Gloay, Heller, Perroud, Aubry, Young, Delessert, Getaz, Tran and Wolff2018). Childhood aggressive behavior during the early years (Koops & de Castro, Reference Koops, de Castro, Jackson and Goossens2006) and childhood depression and anxiety (Fontaine et al., Reference Fontaine, Brendgen, Vitaro, Boivin, Tremblay and Côté2019) have been found also to be longitudinally associated with delinquent behavior. Twelfth grade students with poor health lifestyles (e.g., lack of vigorous exercise, lack of sleep) were significantly more likely to engage in delinquent behaviors, compared with those with healthier lifestyles (Semenza, Reference Semenza2018).
Family risk
Key family-level risk factors include household poverty, family instability (i.e., repeated changes in family structure especially during early childhood), residential mobility, biological father not residing with the child, and intimate partner violence (IPV) (Murray & Farrington, Reference Murray and Farrington2010). Household poverty (i.e., total household income below the poverty line) is positively associated with youth delinquent behavior (Rekker et al., Reference Rekker, Pardini, Keijsers, Branje, Loeber and Meeus2015; Roberson & Azaola, Reference Roberson and Azaola2021). Using US Panel Study of Income Dynamics data, Bosick and Fomby (Reference Bosick and Fomby2018) found that family instability during childhood predicted 18–26-year-old males’ higher likelihood of arrest or incarceration. Decades of research have documented an association between residential mobility and youth delinquent behavior (Gasper et al., Reference Gasper, DeLuca and Estacion2010). Researchers have conceptualized households with no biological father present as a contextual source of stress on children (Kotila & Kamp Dush, Reference Kotila and Kamp Dush2013); children raised in a household with no biological father present are more likely to engage in youth delinquent behavior than children living with two biological parents (Leiber et al., Reference Leiber, Mack and Featherstone2009). The US Office of Juvenile Justice and Delinquency Prevention includes IPV as a family risk factor because it takes place between romantic partners who may or may not live together (Development Services Group, Inc. 2015). Comprehensive reviews of research have shown that children are at greater risk for youth delinquency when their parents are victims of IPV (Artz et al., Reference Artz, Jackson, Rossiter, Nijdam-Jones, Géczy and Porteous2014).
Moderators
Protective factors may buffer children from the negative effects of adversity and risk. Masten (Reference Masten2018) suggests that parenting and caregiver–child relationships are promotive factors that can increase child resilience and help children to develop the skills they need in the context of risk. The quality of fathers’ and mothers’ affective relationships (e.g., attachments) with children and parents’ positive engagement with children profoundly influence children’s development (Ainsworth, Reference Ainsworth1989; Cassidy & Shaver, Reference Cassidy and Shaver2018; Sameroff, Reference Sameroff and Sameroff2009). Driscoll and Pianta (Reference Driscoll and Pianta2011) suggest that individuals’ representations of attachment are best described as closeness (warmth, affection, and open communication) and conflict (behavioral opposition or overt disagreement). Children with close parental relationships and low to moderate levels of relationship conflict during childhood should be protected against the negative effects of father-, mother-, child- and family-level risk across time on their development (Gerlach et al., Reference Gerlach, Fößel, Vierhaus, Sann, Eickhorst, Zimmermann and Spangler2022; Hoeve et al., Reference Hoeve, Dubas, Eichelsheim, van der Laan, Smeenk and Gerris2009; Yoder et al., Reference Yoder, Brisson and Lopez2016). Children with close parental relationships draw parents to them when experiencing distress, have better problem-solving skills and relationships with peers and adults, and are generally happier than children who are not as close and have more conflict with their parents (Moretti & Peled, Reference Moretti and Peled2004).
Researchers have found that close and supportive parental monitoring is an important protective factor for delinquency and is also associated with less increase in delinquent behaviors over time (Hoeve et al., Reference Hoeve, Dubas, Eichelsheim, van der Laan, Smeenk and Gerris2009; Murphy et al., Reference Murphy, Marelich, Herbeck and Payne2009; Tilton-Weaver et al., Reference Tilton-Weaver, Burk, Kerr and Stattin2013). Longitudinal evidence based on a large cohort of adolescents in 5th–7th grades showed that parental monitoring buffered (i.e., protected against) the negative effects of community violence exposure on youth delinquent behavior (Low & Espelage, Reference Low and Espelage2014).
Child sex at birth has been found to protect children against the negative effects of risk factors on youth delinquent behavior. For example, boys exposed to child sexual abuse are more likely to display delinquent behavior than girls (Gauthier-Duchesne et al., Reference Gauthier-Duchesne, Hébert and Blais2022). Boys display more externalizing behaviors than girls when fathers are absent from the household (Mott et al., Reference Mott, Kowaleski-Jones and Menaghan1997). Using data from the Longitudinal Studies of Child Abuse and Neglect, Leban (Reference Leban2021) found that boys who experienced higher ACEs were at greater risk of externalizing trajectories than girls. These studies as well as research showing that boys are twice as likely as girls to be arrested for violent crimes (Cauffman, Reference Cauffman2008) appear to indicate that being a girl protects children against delinquency. In this study we test whether child sex at birth moderates the association among father, mother, child, family risk, and delinquent behavior.
Mediators
Sameroff’s (Reference Sameroff and Sameroff2009) transactional model focuses on the complex influence that child behaviors have on parental behaviors and vice versa. The influence that children have on parents and parents have on children has been examined extensively (e.g., Lansford et al., Reference Lansford, Rothenberg, Jensen, Lippold, Bacchini, Bornstein, Chang, Deater-Deckard, Di Giunta and Dodge2018). Essentially, the transactional model suggests that child behavior (e.g., early child risk) triggers parental responses (Fiese & Sameroff, Reference Fiese and Sameroff1989) and influences parent–child relationships, and parent and family risk influence child behaviors. These transactional processes can be examined with tests of mediation. Researchers suggest that children’s self-control is an important mediator of the association between early childhood parental and family risk and delinquent behavior (Meldrum et al., Reference Meldrum, Campion Young, Soor, Hay, Copp, Trace, Smith-Darden and Kernsmith2020). Residing in a home with high levels of parental and family risk may expose children to highly stressful environments whereby parents are less emotionally available, provide less guidance about how to improve self-control, and do not model effective self-control strategies (Jackson et al., Reference Jackson, Jones, Semenza and Testa2023). Jackson et al. (Reference Jackson, Jones, Semenza and Testa2023) found children’s lack of self-control in school partially mediated the relationship between exposure to ACEs (i.e., ACEs include both family and parental risk factors) and later youth delinquency (Jackson et al., Reference Jackson, Jones, Semenza and Testa2023). Similarly, Fava et al. (Reference Fava, Meldrum, Villar, Zucker and Trucco2022) found that self-control (and sleep problems) in early adolescence mediated the association between ACEs exposure before age 11 and delinquent behaviors in late adolescence. In a study of Portuguese boys, children’s self-control partially mediated the association between exposure to trauma (e.g., parental abuse) and youth antisociality/criminality (Pechorro et al., Reference Pechorro, DeLisi, Abrunhosa Gonçalves and Oliveira2021). Researchers have also examined the mediational influence of parenting behavior on the association between early child risk and youth delinquency. Nelson et al. (Reference Nelson, East, Delva, Lozoff and Gahagan2019) found that severe hyperactivity at age 5 was associated with youth delinquency because it increased maternal hostility toward 10-year-olds. To the best of our knowledge, researchers have not examined the simultaneous transactional processes in which child self-control mediates the association between early father, mother, and family cumulative risk and youth delinquent behavior, and parental monitoring and parent–child relationships mediate the association between early child risk and youth delinquent behavior.
Current study
The current study examined the associations among father-, mother-, child-, and family-level risk during early childhood and adolescent delinquent behavior. On the basis of studies showing that risk at each of these levels predicts delinquency, we constructed separate cumulative risk composites by adding together the risk variables for each individual (e.g., fathers) and for the overall family. The main effects of each composite are examined in a multivariate, structural equation model in relation to delinquency. We focus on cumulative risk at ages 3–5 because research evidence and theory suggest early childhood is a sensitive developmental period for risk exposure (Feng et al., Reference Feng, Shaw and Silk2008). Although delinquent behavior peaks at around ages 16-18 (Farrington, Reference Farrington, Lerner and Steinberg2004), we assessed delinquent behavior at age 15 because FFCW only collected adolescent data at this age. Drawing from the rich FFCW longitudinal data, we were able to support temporal sequencing assumptions by assessing mediators (father–child closeness, mother–child closeness, parental monitoring, child self-control) at 9 years. In addition, middle childhood is an important transition point for development of self-control (Pener-Tessler et al., Reference Pener-Tessler, Markovitch and Knafo-Noam2022).
Based on the risk and resilience perspective and the plethora of studies that have found direct associations between fathers’, mothers’, children’s, and family risk and youth delinquent behavior (e.g., Baggio et al., Reference Baggio, Fructuoso, Guimaraes, Fois, Gloay, Heller, Perroud, Aubry, Young, Delessert, Getaz, Tran and Wolff2018, Bosick & Fomby, Reference Bosick and Fomby2018), the current study tests the following hypothesis:
H1: Father-, mother-, child-, and family-level cumulative risk at ages 3–5 will be positively associated with adolescent delinquent behavior (main effects hypotheses).
Based on transactional theory and evidence that parenting mediates the association between child risk and delinquent behavior (e.g., Nelson et al., Reference Nelson, East, Delva, Lozoff and Gahagan2019), and child self-control mediates the associations among parent and family risk and delinquent behavior (e.g., Pechorro et al., Reference Pechorro, DeLisi, Abrunhosa Gonçalves and Oliveira2021) the current study hypothesizes:
H2: Father, mother, and family cumulative risk at ages 3–5 will be indirectly associated with adolescent delinquent behavior via lower levels of child self-control at age 9, and child risk at age 5 will be indirectly associated with delinquent behavior via its association with lower parental monitoring at age 9 and lower levels of parent–child relationship closeness at age 9 (mediation hypotheses).
Based on the risk and resilience perspective (Masten, Reference Masten2018) and evidence that higher levels of father and mother engagement with children during early childhood and closer parent–child relationships protect children from the negative influences of cumulative risk (Gerlach et al., Reference Gerlach, Fößel, Vierhaus, Sann, Eickhorst, Zimmermann and Spangler2022; Hoeve et al., Reference Hoeve, Dubas, Eichelsheim, van der Laan, Smeenk and Gerris2009; Yoder et al., Reference Yoder, Brisson and Lopez2016), the current study hypothesizes:
H3: Higher levels of parental engagement with children at age 5, closer relationships with fathers and mothers at age 9, and higher parental monitoring at age 9 will protect children from the negative effects of father-, mother-, child-, and family-level cumulative risk at ages 3–5 on adolescent delinquent behaviors (moderation hypothesis).
Child sex at birth is also expected to influence the association between risk and delinquent behaviors.
H4: The associations among father-, mother-, child-, and family-level cumulative risk at ages 3–5 and delinquent behavior will be stronger for boys than girls (moderation hypothesis).
Method
This study used the FFCW study, which followed a cohort of nearly 5,000 children born in large United States cities between 1998 and 2000. The sampling approach resulted in a large number of Black, Hispanic, and low-income families and oversampled births to unmarried couples (McLanahan & Garfinkel, Reference McLanahan and Garfinkel2000). Fathers and mothers were interviewed shortly after the birth of their focal child (baseline) and again when children were 1 (Y1), 3 (Y3), 5 (Y5), 9 (Y9), and 15 (Y15) years of age. Children were interviewed at Y9 and Y15. The study sample was comprised of 3,710 unwed couples and 1,187 married couples at baseline. The analytic sample in the present study (n = 4,255) was limited to cases in which the child resided with the biological mother and did not have autism, mental retardation, or down’s syndrome (n = 416 omitted). Cases were omitted also when data on all study variables (e.g., fathers’ cumulative risk, mother–child closeness) were missing (n = 226). The first author of this paper received institutional review board approval for this study. The study analysis code is available from the first author. This study was not preregistered.
Sample characteristics
As shown in Table 1 slightly less than half of the mothers in the sample were Black (47.9%). Less than one-third were Hispanic (26.9%), non-Hispanic White (21.3%), and Other race/ethnicity (3.7%). Approximately one-third of mothers (33.9%) and fathers (32.8%) completed less than high school at baseline. The median household income at Y5 was $24,000. On average, mothers and fathers were approximately 25 and 28 years old, respectively, at baseline. About one-third (29.7%) of couples were married at Y5. Less than half of the sample included co-residing mother–father couples (42.7% at Y5 and 30.0% at Y9).
Participant characteristics
Measures
Adolescent delinquent behavior
Delinquent behavior at Y15 was measured using items from the National Longitudinal Study of Adolescent Health (Add Health) Wave I and Wave II Home Visit interviews with teenagers. The adapted FFCW survey included 13 of the items from the original Add Health survey (e.g., deliberately damaged property, taken something from a store without paying for it, gotten into a serious physical fight). Adolescents indicated whether they engaged in each delinquent behavior never (0), 1 or 2 times (1), 3 or 4 times (2), or 5 or more times (3). The items are then summed to construct a measure of self-reported delinquent behavior (range: 0-39; α = .74).
Risk variables
Four risk composite variables were constructed by adding together risk items for fathers, mothers, children, and families.
Fathers’ and mothers’ cumulative risk
Six indicators of fathers’ and mothers’ risk at Y3 and/or Y5 were assessed: drug and/or alcohol abuse, unemployment, being charged with breaking the law, depressive symptoms, low education, and poor health status. All indicators were dichotomous variables coded so that 0 = no risk and 1 = presence of risk. The items were identical at Y3 and Y5 except education, a relatively static construct asked only at Y3. Parents who indicated that they experienced a risk factor at either Y3 or Y5, or at both times, were then coded as having experienced risk in that category (e.g., depressive symptoms at Y3 and/or Y5). The father risk indicators at Y3/Y5 were added together to construct a single measure of early childhood fathers’ cumulative risk; the same was done for mothers. Thus, each parent’s cumulative risk could range from 0-6.
To assess drug and/or alcohol abuse, each parent was asked two questions at Y3 and Y5: ‘‘In the past year, was there ever a time drinking interfered with your life?’’ and ‘‘In the past year, did your drug use interfere with life?’’ Those who answered yes to either question were coded as having problems with drugs/alcohol at Y3 or Y5.
Unemployment was measured with two items at Y3 and Y5 that asked parents whether they did any regular work for pay in the last week and whether they were looking for work. Parents who indicated that they did not work in the past week and were looking for work were coded 1 = unemployed. Otherwise, parents were coded 0 = employed or not looking for work.
Mothers and fathers were also asked at Y3 and Y5 whether the police or courts had charged them with breaking the law in the past two years. Parents who answered affirmatively were scored with 1; others received a score of 0.
The Composite International Diagnostic Interview-Short Form (CIDI-SF, Kessler et al., Reference Kessler, Andrews, Mroczek, Ustun and Wittchen1998) was used at Y3 and Y5 to classify parents according to criteria for a DSM-IV major depressive episode. Parents were asked screening questions about whether they have had feelings of depression (“Did you feel sad or blue for two more weeks in the past 12 months?”). If so, they were asked specific questions regarding: (1) losing interest, (2) feeling tired, (3) change in weight, (4) trouble sleeping, (5) trouble concentrating, (6) feeling worthless, and (7) thinking about death. The constructed dichotomous variable for a depressive episode is based on a liberal definition of depression, requiring an affirmative response to the screening question and parent report of experiencing three or more of the specific CIDI-SF items.
Mother’s and father’s educational risk were based on each parent’s report of the highest level of education completed at Y3 (1 = less than high school diploma, 2 = high school diploma or equivalent, 3 = some college or technical training, 4 = college graduate or graduate school). Parents who completed less than a high school diploma were coded as having educational risk = 1. Parents who completed high school or higher were coded as having no educational risk = 0.
Data regarding poor health status was obtained from the question, ‘‘In general, how is your health?’’ which was posed to fathers and mothers at Y3 and Y5. Responses to this question ranged from 1 = excellent to 5 = poor. Parents who indicated that their health was fair or poor were coded as having health risk (= 1). All other parents were coded as having no risk (= 0).
Child risk
Six indicators of child risk were assessed: poor health, speech and language problems, and four indicators of child behavioral risk. Data regarding child’s poor health status were obtained from the Y3 and Y5 primary caregiver questionnaire (mostly mothers) that asked, ‘‘In general, how would you describe your child’s health?’’. Responses to this question ranged from 1 = excellent to 5 = poor. Parents who indicated that the child’s health was fair or poor at Y3 and/or Y5 were coded as having health risk (= 1). All other children were coded as having no risk (= 0). Data regarding child’s speech and language problems were obtained from the Y5 primary caregiver questionnaire that asked, ‘‘Has a doctor told you that your child has a speech/language problem?’’ Responses to this question were 1 = yes and 0 = no.
The Child Behavior Checklist/2–3 and 4–18 (CBCL, Achenbach, Reference Achenbach1992; Achenbach & Rescorla, Reference Achenbach and Rescorla2001) were administered to parent caregivers at Y3 and Y5 to assess children’s problem behaviors. The FFCW survey forms at Y3 and Y5 included most of the CBCL items (56 of the original 100 behavior problem items at Y3, and 72 out of the original 113 items at Y5). Items from the Withdrawn/Depressed (8 items at Y3 and 9 items at Y5, α = .65, .56), Anxious/Depressed (8 items at Y3 and 13 items at Y5, α = .65, .61), Aggression (19 items at Y3 and 20 items at Y5, α = .84, .84), and Attention Problems (6 items at Y3 and 11 items at Y5, α = .71, .67) subscales were used in this study. Respondents rated each behavior on a scale from 0 = not true to 2 = very true or often true. Children who obtained a subscale score 2 standard deviations above the mean at Y3 and/or Y5 were then designated as risk (= 1) (e.g., attention problems at Y3 and/or Y5). The normative CBCL scores were not be used to establish clinical cutoffs because FFCW did not use all items in the full measure.
Family risk
Five indicators of overall family risk at Y3 and Y5 were assessed: IPV, residential mobility, the biological father not residing with the child, household poverty, and family instability. IPV at Y3 was assessed with one item that asked mothers whether they have been seriously hurt in a fight with the other parent since the child’s birth. At Y5, mothers were asked the same question about the last two years. Reports of being hurt in a fight at Y3 and/or Y5 were coded as risk (= 1). Absence of such reports were coded as no risk (= 0).
Residential mobility was assessed with two items where mothers reported how many times the family moved at Y3 (since birth) and Y5 (in the last two years). Mothers who indicated that the family moved two or more times at Y3 and/or Y5 were rated as having risk (= 1). Mothers who indicated that the family moved once or not at all at Y3 and Y5 were rated as having no risk (= 0).
Data regarding the biological father not residing with the child was obtained from the Y3 and Y5 constructed variables that indicated whether or not the biological parents lived together all or most of the time or 7 nights per week. Parents who did not reside together all or most of the time or 7 nights per week at Y3 and/or Y5 were rated as having risk (= 1). Those who lived together at Y3 and Y5 were rated as no risk (= 0).
Household poverty was based on FFCW constructed variables of poverty categories at Y3 and Y5: 1 = 0%–49% of the poverty line, 2 = 50%–99%, 3 = 100%–199%, 4 = 200%–299%, 5 = 300%+). These were determined by mothers’ reports of total household income and size for the past 12 months, and the official poverty thresholds for the corresponding years. Poverty status was recoded so that 1 = less than 100% of the poverty line at Y3 and/or Y5 and 0 = 100% of the poverty line or greater at Y3 and Y5.
Family instability was measured by maternal reports of the number of changes a child has experienced in coresident family structure from birth to Y5 that resulted from a mother’s union (marriage and cohabitation) dissolution or new formation (see Fomby & Osborne, Reference Fomby and Osborne2017). Transitions from cohabitation to marriage as a union status change were not included. The number of family transitions reported by mothers ranged from 0 to 3. Instability was defined as having 1 or more transitions based on evidence that on average, children nationwide experience .43 relationship transitions through age 5 (Brown et al., Reference Brown, Stykes and Manning2016). Sensitivity analyses further verified that having 1 or more transitions was a better predictor of adolescent delinquency than having 2 or more transitions (rs = .11, .07, respectively).
Moderating (protective) and mediating variables
Parent–child relationship closeness at Y9 was indicated by the focal child’s response to six parallel questions pertaining to their relationships with the mother and biological father. Two items were obtained from the National Survey of Children’s Health: “How close do you feel with your mom/dad?” (0 = not very close to 3 = extremely close) and “How well do you and your mom/dad share ideas or talk about things that matter” (0 = not very well to 3 = extremely well). These items have been shown to be strong markers of the quality of parent–child relationships (Bandy & Moore, Reference Bandy and Moore2008). Four items asked the child to indicate the degree to which the parent includes them in important decisions, listens to their side of an argument, spends enough time with them, and misses events or activities that are important to them. These items are measured on a scale from 0 = never to 3 = always. The six mother items (α = .68) and six father items (α = .75) were summed to construct an observed measure of children’s perceptions of mother–child closeness and father–child closeness, respectively.
Father and mother engagement with child (8 items for each parent) were included at Y5 and were reported by mothers and fathers. Response options ranged from 0 = no days per week to 7 = seven days per week. Sample self-report items included how often the parent plays games, sings songs or nursery rhymes, reads stories, plays inside with toys, and hugs or shows physical affection. Parents’ responses to the items were summed, with higher scores indicating higher levels of engagement (father α = .90, mother α = .69, range = 0–56).
Parental monitoring was assessed with five child reported items at Y9. Children were asked how often the primary caregiver knows what you are doing during free time, which friends you hang out with, things that happen when you are not with her/him, where you are going and with whom, and what you spend money on. Item scores ranged from 0 = never to 3 = always. A latent variable was derived from the five monitoring items because of low reliability (α = .46, range = 0–15).
FFCW used the Social Skills Rating System (SSRS-T[teacher report], Gresham & Elliott, Reference Gresham and Elliott1984) to assess child self-control at Y9 (10 items from the teacher version). Teacher perceptions of self-control in the classroom are highly predictive of child outcomes including delinquent behavior and student graduation from high school, even after accounting for students’ achievement, background characteristics, and school aspirations (Soland, Reference Soland2017). Teachers are asked to rate the child’s behavior from 1 = never to 4 = very often. Sample items include: controls temper in conflict with peers, receives criticism well, and accepts peers’ ideas for group activities. The SSRS-T is a reliable and valid teacher rating scale of children’s social behavior in terms of temporal stability, internal consistency, inter-rater reliability, and construct and criterion validity (Elliott et al., Reference Elliott, Gresham, Freeman and Mccloskey1988). Gresham et al. (Reference Gresham, Elliott, Vance and Cook2011) reported strong reliability and validity for the teacher report of classroom-related self-control, and Walthall et al. (Reference Walthall, Konold and Pianta2005) found good support for the factor structure of child self-control (teacher report) and consistency in the factor structure based on race and gender. The self-control items were summed (α = .94, range = 10 to 40).
Child sex assigned at birth was included as a moderator (1 = female, 0 = male). This variable was obtained from the baseline interview with mothers.
Controls
Youths of color are more likely than White youth to be arrested for violent behavior and to be incarcerated in juvenile delinquency centers (Abrams et al., Reference Abrams, Mizel and Barnert2021). Therefore, we controlled for mother’s race/ethnicity, which was measured at baseline based on mother’s report indicating whether she was Non-Hispanic White, Black, Hispanic, or Other (includes Asian American and American Indian). Given that the effects of father and mother risk may relate to father non-residence later in development, biological father not residing with the child at Y9 was controlled using a constructed dichotomous variable that indicated whether or not the biological parents lived together all or most of the time or 7 nights per week. Finally, mother–father marital status was controlled with one item at baseline (1 = married, 0 = unmarried).
Data analysis
We first calculated descriptive statistics for sample characteristics, individual risk items and risk composites using SPSS software. Next, we conducted bias analyses to determine if cases that were missing delinquency data differed from cases that were not missing data on demographic characteristics. Next, we calculated Pearson correlation coefficients for all continuous study variables. The SPSS multiple imputation function with 20 iterations was used to handle any missing data before estimating these bivariate associations.
Structural equation modeling (SEM) with AMOS software was used to test our model (see Figure 1). Three indices were used to assess fit: comparative fit index (CFI), root mean square error of approximation (RMSEA), and the Tucker–Lewis index (TLI). CFI, RMSEA, and TLI were selected because they are the least affected by estimation technique and sample size under conditions of multivariate normality, especially with large sample sizes (Cangur & Ercan, Reference Cangor and Ercan2015). Little (Reference Little2013) provides the following guidelines for these indices (CFI: >.90 is acceptable; RMSEA: < .08 is acceptable; TLI > .95 indicates goodness of fit).
Hypothesized SEM with mother, father, child and family cumulative risk at Y5; mediators at Y9; controls; and delinquent behavior at Y15.
Multiple imputation was carried out on the SEM in AMOS. If a participant was missing any items in the cumulative risk composites (e.g., substance abuse item in the father risk composite) or observed variables (e.g., delinquent behavior), then the whole composite or observed variable was considered missing. See Appendix A for details, as well as frequencies for item-level missingness in the risk composites. Based on recommendations from Mazza et al. (Reference Mazza, Enders and Ruehlman2015), we applied multiple imputation and incorporated auxiliary variables to handle missing data at the variable level. We incorporated marital status and race/ethnicity as auxiliary variables based on their correlations with risk item missingness. This approach mitigates the loss in power due to missing items in composite and scale scores. Auxiliary variables (which can also serve as confounding variables) help to reduce measurement error in mediation analyses that include observed and latent variables (Fritz et al., Reference Fritz, Kenny and MacKinnon2016).
Because the delinquency variable was skewed (3.22) and highly kurtotic (15.38), we conducted bootstrapping, a nonparametric resampling procedure to generate the estimates (Streukens & Leroi-Werelds, Reference Streukens and Leroi-Werelds2016). Bootstrapping was conducted with 5,000 samples and a bias-corrected confidence interval of 95 in the SEM. Direct and indirect effects were calculated. We used the estimands command to estimate specific indirect effects (e.g., child risk → mother–child closeness → delinquent behavior). Effect sizes were calculated using squared multiple correlations. For example, the effect size of fathers’ cumulative risk on delinquency was determined by subtracting the squared multiple correlation for delinquency when father risk was excluded from the analysis (R 2 excluded) from the squared multiple correlation with all risk variables (and other predictors) in the model (R 2 not excluded), divided by 1 – R 2 not excluded.
Moderation analyses were conducted by first mean centering the risk composites (e.g., fathers’ cumulative risk) and parenting variables (e.g., mother–child closeness) and then multiplying each risk composite times each parenting variable (e.g., mother risk X closeness). These interactions enabled us to examine whether the parenting variables protected youth from risk at ages 3 and 5. We examined five sets of possible protective effects. Each set included the four risk composites (father, mother, family, child). The first set examined their product with mother engagement at Y5. The next four sets examined father engagement at Y5, mother–child closeness at Y9, father–child closeness at Y9, and parental monitoring, respectively. Five SEMs were then fitted with each set of interaction terms. Finally, all significant interaction terms were then placed in a final model. Multi-group analysis was used to examine and compare the structural equation models for child sex at birth (boys and girls, moderation effects).
Because we conducted two sets of SEMs in this study (one set examining direct and indirect effects and one set examining moderation effects), we adjusted for familywise error using Bonferroni’s correction by establishing an alpha level of .05 divided by 2 = .025.
Results
Descriptive analyses
Descriptive statistics for the risk composites indicated mothers reported an average of 1.01 risk factors and fathers reported 1.40 risk factors at Y3/5 (see Table 2). Youths were exposed to an average of 1.67 family-level risk factors. This was largely because nearly 50 percent of mothers reported household incomes below the poverty line and 56 percent of mothers and fathers did not reside together at Y5. The average delinquent behavior score was 1.40 (range = 0-39). As noted above, the data were skewed, with approximately 54.1 percent of youth indicating that they did not engage in any delinquent acts. Twenty-four percent of FFCW cases were missing delinquent behavior data. Data were more likely to be missing when mothers reported they are Hispanic and Other race/ethnicity versus Black and White: for delinquent behavior (χ 2[3] = 39.66, p < .001), child risk (χ 2[3] = 64.30, p < .001), father risk (χ 2[3] = 35.56, p < .001), and family risk (χ 2[3] = 356.78, p < .001), but not for mother risk (χ 2[3] = 5.23, p = .264). These results indicate the data are not missing at random.
Correlation matrix, means, and standard deviations
M = mother; F = father; C = child.
Bivariate analyses
Pearson correlation coefficients for associations among mother, father, child, and family cumulative risk composites ranged from .10 to .19 (see Table 2). The correlation between mother–child closeness and father–child closeness at Y9 was .27, p < .001. All risk composites at Y3/Y5 were significantly associated with delinquent behavior at Y15, with correlations ranging from .10 to .15, ps < .001. Mother–child closeness and father–child closeness were inversely associated with delinquent behavior, r = -.06 and r = -.09, ps < .001, respectively.
Multivariate analyses
Model testing
We tested a SEM that regressed delinquent behaviors at Y15 on father, mother, child, and family cumulative risk at Y3/Y5, mother and father engagement with child at Y5, Y9 variables (children’s perceptions of their closeness with mothers and fathers, parental monitoring, and teacher perception of child self-control), and control variables (mother’s race/ethnicity, child sex at birth, marital status at birth, and biological father not residing with the child at Y9). Children’s perceptions of their closeness with mothers and fathers at Y9, parental monitoring at Y9, and teacher perception of child self-control at Y9 were also regressed on father, mother, child, and family cumulative risk at Y3/Y5 (see Figure 1). There were no interaction terms in this model. This model produced an acceptable fit to the data, χ2(df = 74) = 207.09, p < .001; CFI = .99, TLI = .98; RMSEA = .021 (90% CI = .017–.024).
We then tested a SEM that included the same variables as in model 1, plus interaction effects (protective factors analysis). This model produced an acceptable fit to the data, χ2(df = 151) = 747.75, p < .001; CFI = .98, TLI = .96; RMSEA = .030 (90% CI = .027–-.032).
Direct effects
Table 3 provides estimates (bias-corrected) for the model assessing direct linkages in the SEM with no interaction effects (test for H1). Y3/Y5 cumulative father risk (β = .06, p < .001; effect size = .01), Y3/5 mother risk (β = .06, p = .001; effect size = .01), and child risk (β = .05, p = .001; effect size = .011) were positively associated with Y15 delinquent behavior. Child self-control was the only Y9 variable associated with delinquency (β = − .15, p < .001; effect size = .02).
Parameter estimates (bias-corrected) in the path analysis
BCBI = Bias− corrected bootstrapped 95% confidence interval (b = 5,000); lower = lower boundary of the 95% confidence interval; upper = upper boundary of the 95% confidence interval. M = mother. Fam = family. F = father. C = child. Control variables include race/ethnicity (reference group for race/ethnicity is non− Hispanic Black), child sex at birth, marital status, and M − F coresidence at Y9. ***p < .001.
Indirect effects (mediation analyses)
Bias-corrected estimates for indirect effects are shown in Table 3 (test for H2). There was a significant total indirect effect of father cumulative risk on delinquent behavior. The estimands test showed that the positive association between father risk at Y3/Y5 and delinquent behavior at Y15 was significantly mediated by child self-control at Y9 (β = .03, p < .001). Specifically, father risk at Y3/Y5 was negatively associated with self-control, which in turn was negatively associated with delinquent behavior. There were no additional specific indirect effects in the model.
Protective factor analyses
Two moderation effects were found to be significant (test for H3). Children’s perceptions of closeness to fathers at Y9 significantly moderated the association between family cumulative risk at Y3/Y5 and delinquent behavior (b = − .065, p < .001); the complete table including moderation analyses is available from the authors). Youth who reported feeling close to their fathers reported significantly fewer delinquent behaviors when family cumulative risk was high compared with youth who were not close to fathers (see Fig. 2).
Graph showing moderation effect of father–child closeness on the association between family risk and delinquency. FC = father child.
Multiple group analysis (H4) revealed a significant child sex moderation effect on the association between child risk and delinquent behavior, χ 2 Δ = 4.85, p = .024. Boys with higher levels of cumulative child risk at Y3/Y5 were more susceptible to delinquent behaviors than girls with higher level of cumulative child risk. There were no sex X risk moderations for father risk, χ 2 Δ = .04, ns; mother risk, χ 2 Δ = .29, ns; or family risk, χ 2 Δ = .32, ns.
Discussion
National prevalence studies have shown that the majority of youth engage in at least 1 delinquent act (Snyder et al., Reference Snyder, Espiritu, Huizings, Loeber and Petechuk2003). Evidence from the FFCW data set are consistent with these findings. About 59% of 15-year-olds reported at least 1 delinquent behavior (range was from 0 to 27 delinquent behaviors). A growing body of research has examined multiple sources of cumulative risk (e.g., caregiver risk, child risk, Trentacosta et al., Reference Trentacosta, Hyde, Goodlett and Shaw2013) in relation to adolescent delinquent behavior, but few studies have included assessments of fathers’ cumulative risk in addition to mothers’, children’s and family risk (e.g., Simmons et al., Reference Simmons, Steinberg, Frick and Cauffman2018; Swisher & Roettger, Reference Swisher and Roettger2012; Yan et al., Reference Yan, Schoppe-Sullivan and Beauchaine2021). This is important because children’s behaviors are influenced by myriad factors, among the most critical are child, mother, father, and family. Our study suggests that excluding fathers is an important gap in delinquency research. The results of the SEM showed that father’s cumulative risk during early childhood, as well as mothers’ and children’s risk, were significantly and positively associated with youth reports of delinquent behavior (main effects). Our findings have important implications for researchers examining predictors of adolescent delinquency. Specifically, research that omits fathers’ risk may not obtain a full picture of predictors of delinquency.
Child and mother cumulative risk at ages 3–5 were significantly and directly associated with youth delinquent behavior. Many studies have found that higher levels of cumulative mothers’ risk predicts delinquent behavior during adolescence (e.g., Deater-Deckard et al., Reference Deater-Deckard, Dodge, Bates and Pettit1998; Miller et al., Reference Miller, Machlin, McLaughlin and Sheridan2021). It is also now well established that early child developmental risk factors predict youth delinquent behavior (e.g., Staff et al., Reference Staff, Whichard, Siennick and Maggs2015). These child risk factors may be manifestations of neuropsychological deficits in early childhood (Moffitt, Reference Moffitt2005). Studies indicate that children with these developmental deficits are also at increased risk for delinquency when they are exposed to high levels of disadvantage in the home environment (Staff et al., Reference Staff, Whichard, Siennick and Maggs2015). The current study extends the findings of previous researchers by showing the combination of cumulative child, father, and mother risk in early childhood increase the likelihood of youth delinquency. Early intervention with children, fathers, and mothers who show evidence of cumulative risk may be important for programs aiming to prevent and reduce the occurrence of youth delinquency.
Based on the transactional model, we hypothesized that child self-control at age 9 would mediate the association between fathers’, mothers’, and family cumulative risk and delinquent behavior, and parental monitoring at age 9 and lower levels of father–child and mother–child relationship closeness would mediate the association between child risk at ages 3–5 and delinquent behavior. Although we did not find support for parental monitoring or parent–child relationship closeness as mediators of the association between early child cumulative risk and adolescent delinquent behavior, child self-control at age 9 mediated the association fathers’ cumulative risk and delinquent behavior. This is a new finding showing that the pathways through which father risk influences children are through the child himself/herself. One alternate possible explanation, however, is that fathers and children are both at risk for delinquent behavior due to shared genetic influences (Azeredo et al., Reference Azeredo, Moreira, Figueiredo and Barbosa2019). We did not find support for father–child or mother–child closeness as mediators of the association between early child risk and delinquent behavior. It is noteworthy that child risk was significantly related to mother–child but not father–child closeness, but neither mother–child nor father–child closeness was related to delinquency. Researchers have suggested that father–child relationships have a special quality that is different from mother–child relationships (Grossmann et al., Reference Grossmann, Grossmann, Kindler, Zimmermann, Cassidy and Shaver2008). Specifically, fathers offer security to young children in situations that are challenging and stimulating (Grossman et al., Reference Grossmann, Grossmann, Kindler, Zimmermann, Cassidy and Shaver2008). Many fathers like to stimulate children through “rough and tumble” play. This type of play is very enjoyable to young children but can also become overwhelming when the child is overstimulated (Pellegrini, Reference Pellegrini, Smith and Hart2002). Conceivably, fathers’ engagement in supportive play might help children who are at high risk to inhibit their behaviors (i.e., fathers model social control), which may result in children engaging in less delinquent behavior as adolescents. Fathers’ engagement in play during early and middle childhood would be a fruitful area for future research on mediators of the association between child risk and delinquent behavior.
The finding that parental monitoring at age 9 did not mediate the associations between cumulative child risk and delinquent behavior may be due to the age when monitoring was assessed. Parents tend to take a more active role in monitoring when children approach adolescence (e.g., digital behavior, Anderson, Reference Anderson2016). Parental monitoring may become more important for predicting delinquent behavior during the adolescent years (Racz & McMahon, Reference Racz and McMahon2011). Many studies have found that consistent and supportive parental monitoring during adolescence is associated with lower delinquency (for review Racz & McMahon, Reference Racz and McMahon2011).
We found limited support for our hypothesis that child sex at birth would moderate the associations among cumulative risk variables and delinquent behavior: boys with higher levels of child-level risk at ages 3–5 were significantly more likely to report delinquent behavior than girls with higher levels of risk. This is consistent with literature pointing to boys’ particular vulnerability to delinquency (Leban, Reference Leban2021; Gauthier-Duchesne et al., Reference Gauthier-Duchesne, Hébert and Blais2022). Father–child relationship closeness at age 9 significantly moderated the association between early family cumulative risk and delinquent behavior. These findings are important because they highlight the importance of fathers’ positive parenting as protective against the negative effects that family risk can have on children’s social adaptation. The challenge is to be able to intervene so that cumulative family risk does not negatively impact fathering, as it typically tends to do.
The lack of significant moderation effect of mother–child closeness at age 9 on the association between any cumulative risk variables during early childhood and adolescent delinquent behavior may be explained by Lerner and Bornstein’s (Reference Lerner and Bornstein2021) specificity principle, which states that specific child outcomes involve “coaction of specific individuals at specific times in specific places through specific processes.” The father–child relationship might be particularly important during middle childhood in protecting children from the negative effects of early risk on later delinquency because of the uniqueness of father–child relationships compared with mother–child relationships (e.g., McDowell et al., Reference McDowell, Kim, O'Neil and Parke2002). Fathers may have an inhibitory effect on delinquency because they are more sensitive to early warning signs given their own exposures and experiences or because they provide a sense of safety and security.
Limitations
A limitation of the current study is that other potentially significant sources of risk (e.g., peer, school) are not included as predictors of delinquency. The measure of children’s perceptions of parent–child closeness has been widely used in research (Bandy & Moore, Reference Bandy and Moore2009), but it has not been validated with low-income children. The measure of parental monitoring was only administered to primary caregivers (mostly mothers). There were also limitations with the measures of child self-control. Specifically, teacher assessments like other self-report measures are susceptible to systematic bias. There was also potential race bias in the risk indices. For example, the father and mother risk item, “charged with breaking the law,” as an indicator of criminal justice involvement may be biased against families of color. Moreover, being charged with breaking the law does not mean being convicted of a crime or incarcerated (these variables were not available in all waves of data collection). Although convictions and incarcerations are also race-biased, they are stronger risk factors than being charged with breaking the law. The measures of father and mother engagement at age 5 are counts of the number of days parents are involved with children, which may be confounded by some of our risk items (e.g., biological father does not reside with child, unemployment).
There was substantial missing data especially for fathers. Bias analyses showed that the study variables (and items within the risk composites) were more likely to be missing among Hispanic, Other, and unmarried parents at baseline. The current study may therefore be more applicable to parents who were married at baseline and Black and non-Hispanic White families and children. Although our study has the strength of longitudinal design spanning early childhood to adolescence, it is limited in the inclusion of repeat measures. A period of 10 years separated the measures of cumulative risk and youth delinquent behavior, and mediating variables were measured only in middle childhood. More proximal assessments of risk and delinquent behavior may produce larger effects. It is nonetheless significant that father, mother, and child risk measured during early childhood (at ages 3–5) were longitudinally associated with adolescent delinquent behavior. Another limitation of the study is that it assumes a family structure with a mother and father and thus, is not inclusive of same sex co-parents.
Conclusions
A significant contribution of the current study is the inclusion of father-level risk, not just mother, family, and child risk, in the association between risk and delinquency, as is typically done in the literature. Our findings indicate that it is important to assess father, as well as mother, family, and child risk, during early childhood because they each significantly predict adolescent delinquent behavior in multivariate analyses. Importantly, fathers who have closer relationships with children during middle childhood protect them against the negative effects of family risk on delinquent behavior. Another significant finding was that the pathway through which higher father risk was associated with delinquency was through lower levels of child self-control at age 9.
It is important for policy makers and programs serving children and families to identify which levels of risk in the family system are associated with delinquent behaviors. Our findings indicate that low-income fathers (as well as mothers and children) should be included in interventions aimed at preventing risk for delinquency. These interventions should acknowledge that many low-income fathers have positive relationships with their children and that these need to be encouraged and supported so that they can protect children from the negative effects of risk on their behaviors. Programs that target mothers at the exclusion of fathers, for example, may be less effective at preventing juvenile delinquency than programs that include both parents as well as children. Also, programs that ignore child risk may be less effective than those that consider it when formulating parenting programs and curricula. These findings highlight the importance of addressing risk as early as possible to reduce the odds that children will engage in delinquent behaviors and potentially become involved in the justice system.
Acknowledgements
The authors wish to thank the Future of Families and Child Wellbeing staff for their support of this project. We also wish to thanks the parents and children who participated in this study.
Funding statement
Research reported in this publication was supported by the Eunice Kennedy Shriver National Institute of Child Health & Human Development of the National Institutes of Health under Award Number R01HD078547 to the second author.
Competing interests
None.
Appendix A. Risk item descriptives, missing data and associations with auxiliary variables
*p < .05. **p < .01. ***p < .001.
Notes. F = father. M = mother. Fam = family. C = child. Bio = biological. Depress = depressed.
Open access
