Like Father, like Child: Early Life Family Adversity and Children’s Bullying Behaviors in Elementary School

Parental hostility poses a significant threat to child development (Richmond and Stocker 2008). Hostile behavior of parents (e.g., irritability, uncontrollable outbursts of temper) is directly associated with children’s externalizing and aggression problems (Carrère and Bowie 2012; Rijlaarsdam et al. 2014; Stover et al. 2016). However, to our knowledge, little research has been conducted that specifically examined the effect of having hostile parents in relation to bullying behavior of the child. Several studies examined hostile parenting as a risk factor of child bullying behavior (e.g., Schwartz et al. 1997), but this concerns parenting practices as opposed to the effect of hostility symptoms of parents. One might argue that hostile symptoms are manifested in the same way as hostile parenting practices. However, hostile symptoms of parents (including both state and trait hostility) could affect children through other mechanisms than only parent-child interactions. Parental hostile interpretations and hostile behaviors across different situations may serve as a model for children’s developing ideas about the world and result in more general hostile expectations and beliefs (Bandura 1978, 1986; Perry et al. 2001). Therefore, we focused on the broader concept of hostile symptoms of parents. Because parental hostility has been related to externalizing problems, we posit that parental hostility predisposes children to bullying behavior.

The partner relationship is often affected by parental psychopathology, such as hostility (Gordis et al. 2001). Children who bully are often exposed to family conflict, and their families often lack cohesion (Bowes et al. 2009; van Hoof et al. 2008). Typically, bullies also experience more family distress (Copeland et al. 2013; Burk et al. 2008). Therefore, children in these families are not only at an increased risk of developing behavior problems because they have a parent with psychological problems, but possibly also due to exposure to family distress. The studies just mentioned, however, can be prone to reverse causality as most of these studies did not correct for baseline levels of child behavior problems and did not examine family distress before the child was born.

Parents are more likely to endorse harsh disciplinary practices when they experience family distress or symptoms of hostility by spilling over this negativity into parent-child interactions (Frias-Armenta and McCloskey 1998; Jansen et al. 2012a; Stover et al. 2016). Importantly, harsh disciplinary practices like yelling and threatening have repeatedly been related to externalizing and aggressive behavior in children (e.g., Chang et al. 2003; Lereya et al. 2013; Pinquart 2017). Children who bully report that support from their parents is lacking (Demaray and Malecki 2003). Conversely, their parents report more use of punishment (Stevens et al. 2002). In a single-informant study, adolescents’ report of being more harshly disciplined by their parents mediated the association between fathers’ and mothers’ parenting styles (such as experiencing less support and understanding from their parents) and their own bullying behaviors (Gómez-Ortiz et al. 2016). In the current multi-informant study, we examined the association between fathers’ and mothers’ harsh disciplinary practices and children’s bullying behaviors. Moreover, we explored whether prenatal family distress or prenatal hostility is spilled over into parent-child interactions by the use of harsh disciplinary practices, eventually leading to child bullying behavior.

The studies summarized above indicate that family adversity in general is associated with bullying behavior of the child. For several reasons, we hypothesized that adverse family characteristics of the father predict child bullying behavior independently of maternal family adversity. First, the manifestation of fathers’ hostility may have a unique quality accounting for the link between fathers’ hostility and children’s bullying behaviors. It is possible that fathers and mothers experience equal levels of hostile feelings, but mothers may internalize these feelings, whereas fathers may act them out. A possible explanation for this gender difference is that women, consistent with traditional gender roles, are socialized to internalize hostile feelings (Eagly et al. 2000). Given that children can develop behavioral and social-cognitive problems from observing (Bandura 1978, 1986; Perry et al. 2001), children may learn disruptive behavioral patterns mostly by observing the father. Moreover, when father’s hostile feelings are directed outwards, this may increase negativity in the family and further predisposes a child to bullying behavior.

Second, differences between the content of mother-child interactions and father-child interactions may explain the strong associations of father’s hostility and child behavior. Mothers are most often the primary caregivers for children regardless of employment status (Schoppe-Sullivan et al. 2013). Mothers are more involved than fathers in socialization (e.g., taking the child to social events), didactic (e.g., reading with the child), and caregiving (e.g., assisting the child with eating); whereas fathers are more involved in physical play (e.g., playing outside with the child) (Schoppe-Sullivan et al. 2013). Given that mothers are more involved in the child’s life, mother’s hostility or harshness might be compensated by the caring aspects of her parenting. Father’s involvement in physical play encourages obedience and the ability to deal with conflict situations with peers in a socialized manner (Paquette 2004). When the father is hostile and harsh, this hostility and harshness might be more visible and displayed more directly towards the child during the typical father-child interactions. For the child, this might result in learning inappropriate (yet perceived as effective) aggressive ways of resolving a conflict (Bandura 1978, 1986). Furthermore, consistent with the fathering vulnerability hypothesis (Cummings et al. 2004), there is evidence that fathers are more likely to spillover negativity from the marital relationship to the father-child relationship than mothers (Schofield et al. 2009; Stover et al. 2016). Hence, in this way, the child may be more exposed to hostile and harsh behaviors on behalf of the father rather than the mother.

Third, compared to mothers, fathers may respond differently to children’s temperament. For example, in response to their child’s display of emotions, fathers are often more punitive or dismissive, whereas mothers are more likely to encourage the child to express his or her feelings (Cassano et al. 2007). The harsh or hostile responses from the father, in turn, increase the likelihood that the child will react aggressively and bully other children (Lereya et al. 2013) through the aforementioned learning process of observing adverse behavior (Bandura 1978, 1986).

This study will focus specifically on father-reported family adversity because of its presumed salience to the development of children’s bullying behaviors. In addition, we also examine four important aspects that have not been tested yet. First, the direction of the association between family adversity and children’s bullying behaviors is unclear. In our study, parental hostility and family distress were assessed twice, including once before the birth of the child, which allowed us to examine whether these family adversities are precursors of children’s bullying behaviors. Additionally, studying exposure to harsh disciplinary practices at preschool age, prior to school bullying, can reveal important information about the role of (harsh) parenting in later bullying. However, the association between postnatal family adversity and children’s bullying behaviors may be bidirectional as children’s behavior problems can influence disciplinary practices or increase the distress within families (Pardini 2008; Patterson 1982). In order to address alternative directions of association, we adjusted our analyses for child externalizing behavior as reported by the mother when the child was 18 months old.

Second, when the same informant reports on both the risk factors and on bullying involvement, the associations between family adversity and children’s bullying behaviors might be inflated by common method variance. To prevent this problem, we used both parents and children’s classmates as informants. Moreover, by using a peer nomination method that relies on ratings by classmates to assess bullying involvement, we avoided the problem that self-reported bullying involvement data are often biased by social desirability (Griffin and Gross 2004).

Third, adversity of the family environment involves a range of other background family risk factors that often co-occur, such as stressful life events, financial or housing problems (Appleyard et al. 2005). Without consideration of these background family risk factors, alternative explanations remain plausible for associations between family adversity and children’s disruptive behaviors (Davies et al. 2006). Background family risk factors can strain the family system and interfere with the parents’ ability to sensitively care for their child (e.g., less time to spend with the child, or aversive parenting behaviors) (Ramchandani and Psychogiou 2009; Ostberg and Hagekull 2000). Therefore, we examined whether background family risk factors are associated with later bullying behaviors of children. Subsequently, we examined whether the association between family adversity and children’s bullying behaviors changed when accounting for background family risk factors.

Fourth, in addition to parental gender differences, child gender differences should also be taken into account as fathers and mothers may differentially impact bullying behaviors in their sons and daughters. Boys are at higher risk for being involved in bullying than girls (Jansen et al. 2012b), which could be a result of differential socialization processes by parents (Keenan and Shaw 1997; Eagly et al. 2000). We therefore examined whether child gender moderated the association between family adversity and children’s bullying behaviors.

In sum, the objective of our study was to examine the association of separate father-reported family adversity factors, i.e., hostility, family distress, and harsh disciplinary practices, assessed pre- and postnatally, in relation to children’s bullying behaviors in early elementary school. These associations were contrasted with those between mother-reported family adversity and children’s bullying behaviors. We studied these objectives in a large prospective population-based birth cohort while accounting for possible influences of various socio-demographic factors, background family risk factors, and child externalizing problems. The first hypothesis we tested was that the contribution of father-reported family adversity to children’s bullying behaviors is more pronounced than mother-reported family adversity. The second hypothesis we tested was that the association between prenatal family adversity and children’s bullying behaviors is partly mediated by parental harsh disciplinary practices. Importantly, the current study assessed family adversity before and after children were born. This enabled us to examine the timing of the contribution of family adversity to children’s bullying behaviors. Moreover, by using a peer nomination method to assess school bullying, we minimized potential biases of common method variance and social desirability.

This study is embedded in the Generation R Study (Jaddoe et al. 2012), a large population-based cohort of children followed from fetal life onwards. The initial cohort comprised 9778 pregnant women with an expected delivery date between April 2002 and January 2006, living in Rotterdam, the Netherlands. The response at baseline was 61%, and general follow-up rates until the age of 6 years exceed 80%. Regular extensive assessments have been carried out among children and parents (see for detailed description Tiemeier et al. 2012). Written consent was obtained from the parents. The study has been approved by the Medical Ethics Committee of the Erasmus Medical Center.

Bullying was assessed in a subsample of Generation R Study participants. An assessment of bullying was carried out in a study of peer relations across 37 elementary schools (190 school classes) in Rotterdam and suburbs, at the time when the oldest Generation R participants attended grades 1–2 of elementary school (Verlinden et al. 2014). In total, 4017 children from the participating schools in Rotterdam completed assessment. Out of these 4017 children, 1664 children were Generation R Study participants. Parents of 1590 Generation R participants provided consent for data linkage. Subsequently, peer reports of bullying was available for 1590 Generation R Study participants. The analyses for the present study were conducted in 1298 children with data on family adversity.

Bullying behaviors were assessed in elementary school children in grades 1–2 (mean age 7.53 years, SD = 9.24 months) using the PEERS Measure (Verlinden et al. 2014). This interactive computerized instrument is a reliable and age-appropriate method of using peer nominations with young children.

All children in a class completed a computerized peer nomination assessment through which they nominated those classmates, who bullied them. Bullying was explained to children as intentional, repeated and continuous actions of peer aggression and that the victim finds it difficult to defend him or herself (Olweus 1993). Questions were accompanied by audio instructions and visual illustrations (Verlinden et al. 2014). Four different forms of bullying were assessed by a yes/no question: (1) physical bullying (e.g., kicking or pushing); (2) verbal bullying (e.g., calling names); (3) relational bullying (e.g., excluding); and (4) material bullying (e.g., taking away or hiding belongings). If a question was answered affirmatively, children were asked to nominate those classmates who bullied them. To perform the nominations, they could click on the photos of the classmates on the screen.

The nominations a child received from classmates were used to calculate individual bullying scores. The received nominations were weighted by the number of classmates performing the evaluation. Thus, the individual bullying scores were based on the ratings by about 20 classmates with regards to each bullying question, and represent the extent to which a child was perceived as a bully by his or her classmates. The scores on the four different forms of bullying were averaged, with higher scores representing more bullying nominations by classmates.

Following a method described elsewhere (Cecil et al. 2014; Rijlaarsdam et al. 2016), we created scores for three domains that can be considered background family risk factors. We used father’s and mother’s prenatal reports to gather information about these three domains: (i) life stress (e.g., death in family, work problems), (ii) contextual factors (e.g., financial difficulties, housing problems), and (iii) other background factors (e.g., educational level, criminal involvement, substance abuse). For each domain, items were summed and divided by the number of completed items. Higher scores correspond to the presence of more background family risk factors in the respective domain. See Supplementary Table 1 for full item descriptions.

Based on previous studies of bullying, we adjusted our analyses for the following socio-demographic and psychosocial covariates as they were considered possible confounding factors in the association between family adversity and school bullying: child age, gender, national origin, monthly household income and parity (Jansen et al. 2012b; Shetgiri et al. 2014), and child pre-existing externalizing problems (Bowes et al. 2009; Cecil et al. 2014). We considered financial difficulties as a background family risk factor (falling under the contextual factors domain) and household income as a possible confounding factor. In the Netherlands, an extended social security system provides financial security when ones income is insufficient due to, for example, unemployment. As a result, experiencing financial difficulties is a better indicator of stress, whereas a low income in the Netherlands does not necessarily imply financial difficulties but is a broad occupational, social and educational indicator.

Information about child gender and date of birth was obtained from midwife and hospital registries. All other socio-demographic covariates were assessed by prenatal questionnaires. National origin was defined by country of birth of the parents and categorized as Dutch, other Western and non-Western. Birth order of the child (i.e., parity) ranged from “No older siblings” to “Three or more older siblings”. Net monthly household income was categorized into “Less than €1,200” (below social security level), “€1,200 to €2,000” (modal income), and “More than €2,000” (above modal income). Marital status was assessed by questionnaire and defined as single during pregnancy or when the child was 3 years old.

Children’s externalizing problems at 18 months were assessed by mother report on the 24-item externalizing scale of the Child Behavior Checklist for toddlers, CBCL/1½–5 (Achenbach and Rescorla 2000). The psychometric properties of the CBCL are well established (Achenbach and Rescorla 2000). The mother evaluated child behavior (e.g., “Cannot sit still, is restless or hyperactive”, “Fights a lot”) over the past two months using answer categories ranging from “not true” (0) to “very true or often true” (2). Higher scores represent more externalizing problems. The internal consistency for the externalizing scale was α = .88.

Baseline characteristics of the data were explored and Spearman’s correlation coefficients between all family adversity variables and children’s bullying behaviors were calculated. To investigate possible selection bias (non-response analysis), we compared differences in baseline characteristics of participants with (n = 1298) and without (n = 292) data on family adversity with the Chi-square statistic, one-way ANOVAs, and the Mann-Whitney U test. Likewise, we compared baseline characteristics between boys (n = 631) and girls (n = 667).

For our main analyses, to approximate a normal distribution, the background family risk factors, hostility, bullying and externalizing scores were square root transformed; the family distress and harsh parenting scores were logarithmic (Log 10) transformed. Transformed variables were standardized to allow comparability.

To examine associations between father- and mother-reported family adversity (i.e., hostility, family distress, and harsh parenting) and children’s bullying behaviors (hypothesis 1), three multilevel linear regression models were conducted for each predictor separately. In order to examine the effect of the background family risk factors (i.e., life stress, contextual factors, and other background risks factors) and early childhood externalizing problems on children’s bullying behaviors, we also included these as predictors in the first two models. In model 1, unadjusted multilevel linear regression analyses were performed for each predictor separately. Next, to assess whether the associations were similar for boys and girls, we explored interactions between each predictor and child gender. Multiplicative interaction terms that were added to the model to test gender effect modification were non-significant (p > .05). In model 2, we adjusted for socio-demographic characteristics (i.e., child gender, age, ethnicity, parity, and monthly household income) by adding these variables as covariates. Since variables can be confounders without empirical evidence for it in the current dataset (Lee 2014), we based our choice of confounders on previous studies and added them as covariates without backward or forward selection methods. In order to address alternative explanations and directions of the associations between family adversity and children’s bullying behaviors, in model 3 we additionally adjusted for the background family risk factors (i.e., life stress, contextual factors, and other background risks factors) as well as for early childhood externalizing problems. To test for differences between the effect estimates (i.e., standardized regression coefficients) of each specific father-reported family adversity factor and the corresponding mother-reported family adversity factor, we produced 84% confidence intervals around the effect estimates and examined the overlap. As reported by Julious (2004), the level of statistical significance between the two effect estimates would be .05 or lower if the 84% confidence intervals do not overlap.

In further analyses, in order to get an indication of whether father-reported family adversity predicts children’s bullying behaviors over and above that of the background family risk factors, early childhood behavior problems and mother-reported family adversity, we tested a hierarchical multilevel model in which the following predictors were entered cumulatively in four blocks: (1) background family risk factors, (2) early childhood externalizing problems, (3) mother-reported family adversity, and (4) father-reported family adversity. The cumulative blocks were compared using the chi-square likelihood ratio test.

Lastly, to examine whether the association between prenatal family adversity and children’s bullying behaviors is partly mediated by parental harsh disciplinary practices (hypothesis 2), we performed mediation analyses. We accounted for all socio-demographic covariates and background family risk factors. Additionally, when considering fathers’ prenatal hostility or family distress as a predictor, we adjusted the analysis for mothers’ prenatal hostility or family distress respectively, and vice versa. Likewise, the analyses regarding fathers’ harsh disciplinary practices as a mediator were adjusted for mothers’ harsh disciplinary practices, and vice versa. The bootstrapping method, a nonparametric resampling procedure, was used (Hayes 2013). With this method, a confidence interval around each estimate of the indirect (i.e., mediating) effect was computed. If zero is not included within the confidence interval, an indirect effect is present. As recommended by Wen and Fan (2015), the proportion of the indirect effect relative to the total effect was calculated as an effect size measure (P_M). Using the SPSS script for the Indirect procedure (Hayes 2013), bootstrapping for five imputed datasets was performed separately; 5000 samples were requested; and 95% bias-corrected confidence intervals were calculated. Results from the five bootstrap analyses were averaged together.

Missing values of the covariates and risk factors were estimated using multiple imputation. The reported effect estimates are the pooled results of five imputed datasets. Results for our complete case sensitivity analysis were essentially unchanged. We accounted for the clustered data structure (i.e., children from the same school classes were tested) by performing multilevel analyses using school class as a grouping variable. For each predictor, the best fitting model (i.e., fixed effects model, random intercepts model, or random intercepts and slopes model) was chosen using the chi-square likelihood ratio test. For all predictors, the chi-square test of the difference in −2 log likelihood showed a significant increase in explained variance between the fixed effects model and the random intercepts model (p < .05), but not between the random intercepts model and random intercepts and slopes model (p > .05). Therefore, only results from the random intercepts models are reported. Statistical analyses were performed using SPSS Statistics 21.0 (IBM Corporation, Somers, NY, USA).

Data were more often missing for children with non-Western national origins (64% vs. 25%; χ ² (2) = 145.65, p < .001), families with a lower income (44% vs. 12%; χ ² (2) = 90.04, p < .001), and higher parity (11% vs. 2%; χ ² (3) = 82.98, p < .001). Moreover, higher background family risk factors scores (mean rank 793 vs. 599; U = 21,736, z = −4.00, p < .001), higher scores of mothers’ prenatal hostility (mean rank 702 vs. 560; U = 23,027, z = −3.33, p = .001), higher scores of mother-reported family distress 6 years after birth (mean rank 746 vs. 611; U = 45,919, z = −3.68, p < .001), and higher scores of mothers’ harsh parenting (mean rank 681 vs. 528; U = 10,318, z = −2.66, p = .008) were found in participants with missing data.

The results of the multilevel regression analyses predicting children’s bullying behaviors in early elementary school from background family risk factors, early childhood externalizing problems, and several components of family adversity (i.e., parental hostility, family distress, and harsh disciplinary practices) are presented in Table 3. In unadjusted model 1, the background family risk factor domains of contextual factors and other background factors, as well as children’s externalizing behavior problems, were associated with children’s bullying behaviors (all p < .05). Moreover, almost all family adversities were associated with children’s bullying behaviors, with the exception of mothers’ prenatal and postnatal hostility (p > .05).

After adjusting for socio-demographic covariates in model 2, the other background factors domain and children’s externalizing behavior problems remained associated with children’s bullying behaviors (p < .05). Furthermore, several family adversities remained associated with children’s bullying behaviors. First, the analyses of family distress showed that family distress as experienced prenatally by fathers (B = 0.10, 95% CI = [0.05, 0.15], p < .001) and as experienced 6 years after birth by mothers (B = 0.09, 95% [CI = 0.04, 0.14], p < .001) were associated with children’s bullying behaviors. Second, hostility symptoms of fathers 3 years after birth of their child (B = 0.06, 95% [CI = 0.01, 0.11], p = .02) were associated with children’s bullying behaviors. Third, results indicated that fathers’ but not mothers’ harsh disciplinary practices when their child was 3 years old were associated with children’s bullying behaviors (fathers B = 0.06, 95% CI = [0.01, 0.11], p = .02; mothers B = 0.02, 95% CI = [−0.03, 0.08], p = .44).

The results from the hierarchical multilevel model, in which we tested four blocks of predictors cumulatively, are presented in Supplemental Table 2. The comparison of block 3 to block 4 shows that father-reported family adversity predicted children’s bullying behaviors over and above that of the background family risk factors, early childhood externalizing problems and mother-reported family adversity (χ ² (4) = 14.93, p < .01).¹

Several mediation models tested whether the association between prenatal family adversity and children’s bullying behaviors is partly mediated by harsh disciplinary practices. The association between prenatal hostility symptoms of fathers and children’s bullying behaviors was mediated through fathers’ harsh disciplinary practices (Fig. 1). The point estimate of this indirect effect was 0.012, with a 95% bias-corrected confidence interval from 0.004 to 0.023. Relative to the total effect (0.048), this gives an effect size of P_M = .25. In other words, 25% of the effect of fathers’ prenatal hostility on children’s bullying behaviors occurred indirectly through fathers’ harsh disciplinary practices. Likewise, the association between prenatal family distress as reported by fathers and children’s bullying behaviors was mediated through fathers’ harsh disciplinary practices. The point estimate of this indirect effect was 0.007, with a 95% bias-corrected confidence interval from 0.001 to 0.017. Relative to the total effect (0.075), this gives an effect size of P_M = .10, meaning that only a small proportion (10%) of the effect of father-reported prenatal family distress on children’s bullying behaviors occurred indirectly through fathers’ harsh disciplinary practices. The indirect effects of the models containing mothers’ prenatal hostility or family distress as predictors, however, were non-significant (results not shown).

Analyses presented in Table 3 were repeated in a subsample excluding single-mother families (i.e., single during pregnancy or when the child was 3 years old; Supplementary Table 3). The results gave a similar picture as in the full sample, although the association between fathers’ hostility 3 years after birth of their child and children’s bullying behaviors was slightly attenuated.

Prior research on family adversity and children’s bullying behaviors has often failed to specifically include characteristics of both the father and the mother (e.g., Copeland et al. 2013; Schwartz et al. 1997; Burk et al. 2008). One of the major strengths of our study is that we utilized multiple informants (i.e., father and mother) to assess the risk factors, and multiple peer reports to assess the outcome. This strengthened the validity of the findings, and enabled us to study father’s and mother’s effects on child development separately. Another major strength of our study is its prospective longitudinal design, which is rare in studies of family adversity effects. Measuring parental hostility and family distress before the birth of the child and harsh parenting at preschool age, allowed us to examine the prospective development of the association. Finally, embedding our study in a large longitudinal cohort allowed us to adjust our analyses for numerous confounders.

Despite these strengths, our study has some limitations, which could be addressed in future studies. One limitation is that the non-response analyses indicated that some selection occurred toward families with a higher socioeconomic status. However, prospective cohorts (such as the Generation R Study) do not need to be representative of a population to be generalizable as long as the study population is sufficiently large and captures a diversity of exposures and backgrounds (Manolio and Collins 2010). Second, observational measurements of family adversity were not feasible in our large study. The self-reports that were used to assess family adversity could have led to socially desirable answers and to underestimation of the degree of family adversity. However, we used only questionnaires that were reliable and valid. Third, we excluded three items from the harsh disciplining questionnaire as they encompass illegal practices. This may reduce the comparability of the measure with other settings. Fourth, although the questionnaire measuring externalizing behaviors (CBCL/1½–5) is validated for 18-month-old children, some behaviors (e.g., “Cannot sit still”) might be normative for this age. However, we did not use this measure as an outcome, but only to assess whether the observed associations under study were independent of pre-existing externalizing problems. Fifth, in the current population-based sample, it was not possible to focus on families without a father.

Previously, little attention has been given to the influence of fathers on the development of children’s bullying behaviors. Our findings highlight that the influence of fathers cannot be ignored. Future studies should collect information on fathers to further examine fathers’ risk factors that predispose children to engage in bullying behaviors.

Next to implications for future research, our findings have practical implications for anti-bullying interventions. School bullying prevention is an important public health goal (Srabstein and Leventhal 2010). Anti-bullying interventions can prevent serious outcomes that are directly related to school bullying, such as internalizing and externalizing problems (Ttofi and Farrington 2012). Early interventions should target children who display bullying behavior to prevent them from developing more serious forms of aggression and violence later in life (Sourander et al. 2007). However, anti-bullying interventions that focus solely on schools may not adequately address the source of child bullying behavior. As family adversity is an important predictor of children’s bullying behaviors, it is imperative to involve parents, in particular fathers, in anti-bullying interventions. For example, anti-bullying interventions could include parental training sessions to teach positive disciplinary practices, and to teach strategies on how to cope with negative or hostile feelings, and how to handle conflicts or stress within the family. Parent involvement in school-based anti-bullying interventions has indeed been linked to a reduction in bullying (Axford et al. 2015). If children who bully their peers are targeted as early as possible using interventions that involve the entire family, then this may ensure that the negativity emerging from family adversity is not spilled over to the child, and eventually result in a strong reduction of school bullying.

From previous research, it is known that family adversity is associated with children’s bullying behaviors at school. The present study adds to this literature by demonstrating that father-reported family adversity is a risk factor for children’s bullying behaviors. In particular, although the effect sizes were relatively small, fathers’ hostile and harsh behaviors were related to children’s bullying behavior in elementary school, whereas the effect of the mothers’ behavior was less pronounced. An accurate understanding of how an adverse family environment predisposes children to engage in school bullying is key for future interventions.