Abstract
The number of crimes related to child-to-parent violence (CPV) has increased in recent years. Most cases are undetected. This is the first study to compare CPV-justice involved youth (CPV-JI/detected CPV) and CPV non-justice involved youth (CPV non-JI/undetected CPV), with the aim of examining differences in risk factors and pattern of violence. The sample included 306 Spanish youth (229 males, 77 females) aged 14 to 19 years: 83 CPV-JI, 105 CPV non-JI, and 115 non-CPV. The results show a worse individual, family, and social profile in CPV-JI youth compared with CPV non-JI youth, and in the latter compared with non-CPV youth. Undetected aggressors were more similar to detected aggressors than to non-aggressors. In addition, CPV-JI youth had higher scores in CPV-Q than CPV non-JI youth. The CPV-Q score that best classified detected versus undetected aggressors is 16 to 17 points, proving to be a useful instrument in classifying different types of CPV aggressors.
Child-to-parent violence (CPV) has received increasing attention in the scientific literature during the last decade, with research on the topic spreading to an increasing number of countries (Contreras et al., 2021). CPV is a type of violence in which sons and daughters show behaviors of physical, psychological, and/or financial violence to gain power, control (Cottrell, 2001), and dominance over parents (Howard & Rottem, 2008) in a repeated, conscious, and intentional manner (Pereira et al., 2017). According to Cottrell (2001), psychological violence includes, for example, threats, intimidation, and verbal behavior, such as insults or offensive comments. Physical violence refers to acts such as kicking, punching, or pushing. Financial violence includes behaviors such as stealing money or incurring debt that parents must pay. Control, power, and dominance over parents are reflected in behaviors such as making exacting demands or controlling family functioning. These types of violence can occur at the same time and, in fact, overlap (Cottrell, 2001).
Among other reasons, the scientific interest in this topic is due to the increase in reports of this type of crime. CPV is gathered in the Spanish Criminal Code as a “domestic violence” crime. CPV crime started to increase in Spain since 2005, and the data have remained high during the last few years. According to the data gathered in the Report of the Spanish Fiscalía General del Estado (2023), there has been a new increase in the last year, up to a total of 4,332 proceedings initiated for this crime, which confirms the persistence of this social problem.
However, these official data refer only to cases in which parents report their children. Most cases of CPV are undetected. This type of violence is kept private, as parents tend to be reluctant to report their children (Kennedy et al., 2010). Reluctance has been linked to feelings of guilt and shame on the part of parents, who are sometimes seen as responsible for the situation (Walsh & Krienert, 2007). All this leads to non-reporting until the situation becomes untenable. Judicial cases represent the last link in this type of violence. The development and consolidation of violent behaviors can be progressive, resulting in an escalation from psychological violence to more serious forms of violence, such as physical violence (Cottrell, 2001). Detected CPV and long-lasting violent behavior could represent the most severe, high-risk cases (Moulds et al., 2019).
Studies with community samples can provide relevant data on the extent of undetected cases of CPV. In such studies, it is difficult to determine the magnitude of the problem, because the reported rates are highly variable. A lack of agreement regarding the definition of CPV, the use of varied assessment instruments, and the different criteria for estimating CPV have contributed to the large variability in prevalence data. In general terms, studies in different countries such as Spain (Cano-Lozano, León & Contreras, 2021), Germany (Beckmann et al., 2021), the United States (Margolin & Baucom, 2004), and Chile (Jiménez-García et al., 2020), among others, reveal that physical violence has been placed between 2% and 8%, psychological violence between 20% and 70%, economic violence between 11% and 35%, and control and domain between 20% and 45%. These data reflect the high prevalence of CPV in the general population, most of which are undetected. Early identification of these undetected cases is important to intervene as soon as possible.
The present study compares CPV-Justice Involved youth (CPV-JI youth/detected CPV), CPV non-Justice Involved youth (CPV non-JI youth/undetected CPV), and non-CPV youth. It is the first study that compares detected CPV and undetected CPV, with the aim of examining differences in risk factors and pattern of violence. Comparing these groups would allow determination of whether the risk factors are similar in both groups, whether there is a progression in the profile of violence, or whether the individual, family, or social profile of detected aggressors is different from that of undetected aggressors. Exploring these aspects would also help to understand the causes that lead some cases to be reported and ultimately come to the attention of the justice system. The inclusion of a non-CPV group will also allow us to know if undetected aggressors are more similar to detected aggressors or to non-aggressors.
Risk Factors for Child-to-Parent Violence
From an ecological perspective (Hong et al., 2012), it is proposed that multiple variables influence the development of CPV at different levels (ontogenetic, microsystem, exosystem, and macrosystem). In this line, previous research has related CPV to individual, family, and social variables. Following Cottrell and Monk (2004), although in some cases many of these risk factors may be present, it is also possible that isolated variables exert a particularly strong influence on violent behavior. From this approach, the more risk factors are present, the more probability that CPV occurs.
At the individual level, different studies on CPV have analyzed, among other aspects, some cognitive and emotional variables. For example, Contreras and Cano-Lozano (2015) found that CPV-JI youth presented more deficits in cognitive skills to solve interpersonal problems compared with non-CPV youth, along with a more hostile perception of parental authority and the family environment in general. A large proportion of CPV-JI youth also show high levels of anger (Armstrong et al., 2018). Other studies with community population reveal similar results, indicating that CPV is related to hostility (Rosado et al., 2017), hostile attribution, anger, access to aggressive responses, and anticipation of positive consequences of the aggressive response (Calvete, Gámez-Guadix, & García-Salvador, 2015; Contreras et al., 2020b). In a recent study with young people from the community population, Simmons et al. (2020) demonstrated that those who abused their parents presented higher levels of anger compared with non-abusive ones. Regarding empathy, a recent study has pointed out that CPV aggressor in the judicial context present more empathy problems than CPV aggressors from a clinical context (Loinaz & Ma de Sousa, 2020). In the study by Martín et al. (2022), also conducted with a judicial sample, this variable was not associated with CPV. In studies conducted with adolescents from community populations, empathy was not a predictor of CPV (Calvete, Gámez-Guadix, & García-Salvador, 2015; Contreras et al., 2020b). Another relevant variable is the justification of violence. In this regard, Loinaz and Ma de Sousa (2020) recently found that a significantly higher percentage of CPV aggressors in the judicial context present more attitudes justifying violence compared with CPV aggressors from clinical contexts. In addition, CPV-JI youth report more antisocial attitudes than non-CPV youth (Contreras & Cano-Lozano, 2016b). However, Simmons et al. (2020) have found similar levels of attitudes justifying violence among abusive and non-abusive young people in a community sample. Regarding impulsivity, studies with judicial samples reveal higher levels of this variable in CPV-JI youth compared with non-CPV youth (Contreras & Cano-Lozano, 2015), and studies that include both judicial and community populations reveal a direct relationship with CPV (Del Hoyo-Bilbao et al., 2020). Similarly, the relationship between impulsivity and CPV has also been found in community samples (Calvete et al., 2011), and also that abusive youth present higher levels of impulsivity than non-abusive youth (Simmons et al., 2020). Furthermore, some studies reveal a pattern of drug use in a large proportion of adolescents who commit CPV offenses (Armstrong et al., 2018), higher levels of drug use in CPV-JI youth when compared with non-CPV youth (Contreras & Cano-Lozano, 2015; Hernández et al., 2020; Ibabe et al., 2014), and a positive relationship between drug use and CPV (Del Hoyo-Bilbao et al., 2020). Research with community samples also reveals an association between CPV and drug use (Beckmann et al., 2021; Calvete et al., 2011; Cano-Lozano et al., 2020; Ibabe et al., 2013).
Regarding family context variables, studies with judicial samples have found that CPV-JI youth present lower perceived parental warmth and more parental criticism/rejection compared with non-CPV youth (Contreras & Cano-Lozano, 2014). In this line, low parental warmth (Calvete, Orue, et al., 2015; Cano-Lozano et al., 2020) and maternal emotional rejection (Zhang et al., 2019) are also associated with CPV in community samples. In addition, exposure to violence in the family context is a significant predictor of CPV (Gallego et al., 2019). Studies with the judicial population have indicated that exposure to violence at home predicts CPV (Cuervo, 2021) and that CPV-JI youth report higher levels of exposure to violence at home compared to non-CPV youth (Contreras & Cano-Lozano, 2016a; Hernández et al., 2020) and to other-JI youth (Cano-Lozano et al., 2023; Cuervo, 2021). Similar results have been found in studies with community samples, observing a consistent relationship between CPV and exposure to violence at home (e.g., Calvete, Orue, et al., 2015; Cano-Lozano et al., 2022; Cano-Lozano, Navas-Martínez, & Contreras, 2021; Navas-Martínez & Cano-Lozano, 2022) and that abusive young people present higher levels of direct and vicarious victimization than non-abusive youth (Simmons et al., 2020).
Other authors have explored exposure to violence in other social contexts. In this respect, several studies show that CPV-JI youth present higher levels of exposure to violence at school and in the street compared with non-CPV youth (Contreras & Cano-Lozano, 2016a; Hernández et al., 2020) and other-JI youth (Cano-Lozano et al., 2023) and that school victimization predicts CPV in adolescents from community samples (Navas-Martínez & Cano-Lozano, 2022). Among other social variables, it has been found that a significantly higher percentage of CPV-JI youth have deviant peer groups compared with adolescents with CPV behaviors from clinical samples (Loinaz & Ma de Sousa, 2020). The association between CPV and deviant peer groups has also been found in adolescents from community populations (Calvete et al., 2011; Cano-Lozano et al., 2020).
Research on this topic suggests differences in risk factors between CPV-JI youth, other-JI youth, and non-CPV youth, and between CPV aggressors and non-aggressors in studies with community samples. However, the differences between CPV-JI youth (detected) and CPV non-JI youth (undetected) have not been analyzed. A worse psychological and social profile is expected in detected compared with undetected ones, but this has not been investigated to date. A recent study found differences between CPV aggressors from the judicial and clinical settings, with aggressors from the judicial system presenting a worse psychological, family, and social profile (Loinaz & Ma de Sousa, 2020). From an ecological perspective, the more risk factors are present, the more probability that CPV occurs (Cottrell & Monk, 2004). In addition, detected CPV could represent the most severe, high-risk cases (Moulds et al., 2019). Thus, there may be a continuum in the cases of CPV, so it is expected that those cases that reach the judicial setting (which are expected to be the more serious cases) would present a worse individual, family, and social profile than those CPV cases from the community population.
Assessment of Child-to-Parent Violence
Over time, various instruments have been developed to assess CPV, including some general family violence instruments adapted to this type of violence, and others specific to this type of violence. A recent systematic review by Ibabe (2020) examined empirical studies showing evidence of the psychometric properties of instruments that assess CPV between the ages of 10 and 25 years. Eleven instruments for measuring CPV were identified. To assess the quality of the selected instruments, The COnsensus-based Standards fot the selection of health Measurement INstruments (COSMIN) guidelines were applied (Mokkink et al., 2018; Prinsen et al., 2018) highlighting three of them: the Child-to-Parent Aggression Questionnaire (CPAQ; Calvete et al., 2013), Abusive Behavior by Children-Indices (ABC-I; Simmons, McEwan, et al., 2019), and the Child-to-Parent Violence Questionnaire (CPV-Q; Contreras et al., 2019).
The CPAQ has 10 parallel items (father/mother) and assesses the presence of psychological (seven items) and physical (three items) aggression during the past year. Recently, the psychometric properties of the Child-to-Parent Aggression Questionnaire–Revised (CPAQ-R), which includes both violent behaviors against parents and the reasons for these behaviors, have been examined (Calvete et al., 2023). This instrument was administered to children aged 13 to 18 years from a community population. Physical aggression is considered severe if it has occurred at least three times in the last year, while psychological aggression is considered severe if it has occurred at least six times in the last year.
The ABC-I has nine items, which measure three factors: physical aggression (three items), verbal aggression (two items), and coercive behavior (four items). A score of 16 or more is considered abusive. This instrument can be used with adolescents and young adults aged 14 to 25 years and their parents, and it should be administered in conjunction with the Beliefs About Child-to-Parent Abuse Questionnaire (BACPAQ; Simmons, Purcell, & McEwan, 2019) which investigates social norms about child-to-parent abuse by asking how often various behaviors must occur to be abusive.
The CPV-Q consists of 14 parallel items (father and mother) that measure psychological (four items), physical (three items), and financial violence (three items), together with behaviors of control and domain over their parents (four items). It also includes a scale with eight items on the reasons for the violence. In the Ibabe (2020) review, the CPV-Q obtained the highest rating as a promising instrument. This instrument was administered to children aged 12 to 18 years from a community population. It currently has adaptations to adolescents from a judicial population (Cano-Lozano et al., 2023), young people aged 18 to 25 years (Cano-Lozano, León, & Contreras, 2021), parents (Contreras et al., 2020a), and in other countries (Jiménez-García et al., 2020, 2022). The criterion for determining the presence of CPV behaviors was the presence of any CPV behavior two or more times in the last year.
As we have pointed out, both the CPAQ and the CPV-Q include a scale on the reasons for CPV which were grouped into two factors corresponding to reactive and instrumental reasons. On one hand, reactive reasons refer to the use of violence in response to a previous aggression or threat of aggression. On the other hand, instrumental reasons imply the use of violence to get what the adolescent wants.
Concerning assessment, an important challenge in the study on the subject is to clearly differentiate what is CPV from what is not CPV. Occasional aggressive episodes may be part of the normative behavior of young people, while a continuous pattern of these behaviors could be considered abusive (Ibabe, 2020). In either case, there are no clear cut-off points for considering a child to be abusive rather than just aggressive (Gallagher, 2008). Although the aforementioned assessment instruments have used different criteria for the presence or absence of CPV or greater or lesser severity, there are no studies that empirically support one or the other type of criterion. This analysis with cut points may help to identify empirically supported criteria for differentiating CPV cases. This could be useful not only for research but also for professional practice for early detection of potential cases in the community population.
In the present study, we set out to examine this aspect with the CPV-Q, assessment instrument which has excellent psychometric properties (Contreras et al., 2019) and obtains the best evaluation in the review by Ibabe (2020). Specifically, we aimed to identify the scores that best differentiate a group of CPV-JI youth (detected CPV) from a group of CPV non-JI youth (undetected CPV). In the judicial population, the judicial sentence serves as a gold standard to classify all of them as true CPV aggressors. The identification of criteria is required to correctly classify CPV aggressors in the community population (Gallego et al., 2019).
The Present Study
The present study compares three samples of adolescents, that is, CPV-JI youth (detected CPV), CPV non-JI youth (undetected CPV), and non-CPV youth, to examine what risk factors distinguish these three populations. Specifically, we set out the following objectives and hypotheses:
To examine what individual, family, and social risk factors differentiate CPV-JI youth (detected CPV), CPV non-JI youth (undetected CPV), and non-CPV youth. Detected CPV (Contreras & Cano-Lozano, 2014, 2015, 2016b; Hernández et al., 2020; Ibabe et al., 2014) and undetected CPV (Simmons et al., 2020) will present significantly higher scores on risk factors compared with non-CPV youth (Hypothesis 1a) and, in turn, detected CPV will obtain significantly higher scores on risk factors compared with undetected CPV youth (Loinaz & Ma de Sousa, 2020) (Hypothesis 1b).
To study which risk factors predict belonging to the group of CPV-JI youth (detected CPV) and CPV non-JI youth (undetected CPV). It is expected that detected CPV would present a worse individual, family, and social profile than undetected CPV youth (Loinaz & Ma de Sousa, 2020).
To analyze the differences in the pattern and reasons for CPV between CPV-JI youth (detected CPV) and CPV non-JI youth (undetected CPV). Detected CPV will present higher scores in the different types of CPV and the reactive and instrumental reasons for CPV than undetected CPV youth (Loinaz & Ma de Sousa, 2020).
To know which cut-off score on the types of violence obtained by the CPV-Q offered the best detection capacity in the classification in the CPV-JI youth (detected CPV) and CPV non-JI youth (undetected CPV) groups. Due to the absence of previous studies on the subject, no predictions are made.
Method
Participants
The sample included 306 youth (229 males, 77 females) aged 14 to 19 years (Mage = 15.7, SD = 1.3). Specifically, 83 CPV-JI youth (detected CPV) (27.1%) were youth with CPV-related offenses (63 males, 20 females, Mage = 16.2, SD = 1.3) from the judicial field. Judicial measures were internment (38.6%), family/educational cohabitation (34.9%), probation (20.5%), and socio-educational work (2.4%). Extrajudicial measures were imposed on 3.6% of the cases. The 108 CPV non-JI youth (undetected CPV) (35.3%) were youth who had committed CPV (80 males, 28 females, Mage = 15.6, SD = 1.2) from the community field but they had not been reported to the judicial system. The 115 non-CPV (37.6%) were youth who had not committed CPV (86 males, 29 females, Mage = 15.5, SD = 1.3) from the community field. The 83 CPV-JI youth were recruited from the Juvenile Justice Service of Jaén (Andalusia) and Oviedo (Asturias), regions of southern and northern Spain, respectively, under Organic Law 5/2000 on the Criminal Responsibility of Minors. The sample includes all young people with crimes related to CPV (which come from the Juvenile Justice Service of Jaén and Oviedo, in Spain, with the mandatory authorizations) and who gave their consent to participate in the study. The CPV non-JI and non-CPV groups were recruited from different public and concerted high schools in Jaén and Oviedo (provinces in the south and north of Spain) applying an intentional non-probabilistic sampling method. The sample includes young people between the ages of 14 and 19 who had the necessary authorizations and who gave their consent to participate in the study. They completed the CPV-Q (Contreras et al., 2019). From the full community sample of 1,580 youth, 108 youth with the highest CPV-Q scores (1 SD above the community sample mean), which corresponds with CPV-Q scores of 9 or more, and 115 youth with CPV-Q scores of 0 were selected. The selection of these two groups also ensured that the proportion of males and females was similar to that of the CPV-JI youth group. Table 1 presents information about demographic characteristics as well as the differences between groups in these variables.
Sociodemographic Variables
Note. Differences between groups. CPV-JI = Child-to-Parent Violence-Justice Involved; CPV non-JI = Child-to-Parent Violence non-Justice Involved; non-CPV = Non-Child-to-Parent Violence.
p < .05, **p < .01, ***p < .001.
Instruments
Participants provided information on sociodemographic data (age, sex, nationality), family data (socioeconomic status, marital status of parents), and judicial data (offenses, judicial measures, criminal record). Judicial data were also confirmed by juvenile justice technicians.
The Child-to-Parent Violence Questionnaire (CPV-Q) (Contreras et al., 2019)
The CPV-Q is composed of 14 parallel items (for the father and mother) that refer to different CPV behaviors: physical (three items), psychological (four items), financial (three items), and control/domain (four items) over their parents (father: α = .91; mother: α = .90). Participants indicate the frequency of these behaviors in the past year. Each of the items is answered using a 5-point scale ranging from 0 (has never occurred) to 4 (has occurred six or more times), with total scores ranging from 0 to 56 and by types of violence: physical (0–12), psychological (0–16), financial (0–12), and control/domain (0–16). In addition, it includes three items relating to reactive and five items relating to instrumental reasons for violence toward parents (α = .82), measured using a 4-point scale ranging from 0 (never) to 3 (always), with total scores ranging from 0 to 9 and from 0 to 15, respectively.
Social Information Processing Questionnaire (SIP) in Child-Parent Conflicts (Calvete, Gámez-Guadix, & García-Salvador, 2015)
Adolescents were asked to imagine three scenarios consisting of different conflicts with their parents. There were nine items for each scenario to assess five components of the SIP: hostile attribution, anger, access to aggressive response, the anticipation of positive consequences of aggressive action, and empathy (α = .83). Each item was answered using a 5-point response scale, ranging from 0 (not at all) to 4 (to a great extent), with total scores ranging from 0 to 36.
The Justification of Violence (JV) Subscale of the Irrational Beliefs Scale for Adolescents (Cardeñoso & Calvete, 2004)
It consists of nine items (α = .82), and adolescents were asked to answer to a 4-point response scale, ranging from 1 (strongly disagree) to 4 (strongly agree), with total scores ranging from 9 to 36.
The Impulsivity Subscale (IMP) of the Attitudes and Cognitive and Social Strategies Questionnaire (Moraleda et al., 1998)
It consists of seven items (α = .76), and adolescents were asked to answer to each item on a 4-point scale (1 = strongly disagree, 4 = strongly agree), with total scores ranging from 7 to 28.
Drug Use Questionnaire
This instrument of six items was designed ad hoc for this study (α = .85). The adolescents were asked to indicate how often they had used different drugs (tobacco, alcohol, marijuana, hashish, cocaine, speed/ecstasy) in the last year, on a scale from 1 (never) to 5 (daily), with total scores ranging from 6 to 30.
The Warmth Scale (WS, Fuentes et al., 1999)
The WS is composed of 20 parallel items (for the father and the mother), divided into two factors: (a) warmth-communication and (b) criticism-rejection of parents toward their children (α = .85). Each factor consists of 10 items rated on a scale ranging from 1 (never) to 5 (always), with total scores ranging from 10 to 50.
The Exposure to Violence in the Context of Home Subscale of The Exposure to Violence Scale (EV; Orue & Calvete, 2010; α = .90) includes three items for each direct and vicarious exposure to violence at home that are rated on a scale ranging from of 1 (never) to 5 (every day), with total scores ranging from 3 to 15.
The Exposure to Violence in the Context of School and Street Subscales of The Exposure to Violence Scale (EV; Orue & Calvete, 2010; α = .90) includes three items each for direct and vicarious exposure to violence at school and in the street (α = .90). Each item is rated on a scale ranging from 1 (never) to 5 (every day), with total scores ranging from 3 to 15.
Deviant Peer Questionnaire
This instrument was designed ad hoc for this study, based on the Deviant Peer Scale of Barnow et al. (2005). It consists of a total of four items, with which adolescents were asked to indicate whether their friends have been involved in criminal activities, shown violent behavior, skipped class, and/or used drugs (α = .77). The response scale ranged from 1 (none of them) to 4 (all of them), with total scores ranging from 4 to 16.
Design and Procedure
A descriptive study with cross-sectional surveys was designed. First, the study was approved by the Ethics Committee of the University of Jaén (Spain) (Ref. CEIH 270215-1). The research required the authorization of the Directorate General of Juvenile Justice of Andalusia and Oviedo and the collaboration of the Juvenile Justice Service of Jaén and Oviedo (Spain). Parental consent was also requested for us to evaluate their children. In addition, youth were also informed and their consent was requested. To ensure confidentiality, each participant received an identification code. No incentive was offered for participation. The assessment was carried out individually in the corresponding judicial institution and as a group in the classrooms of the educational center, and by members of the research team specifically trained in this protocol.
Data Analysis
The participants’ responses to the scales employed in the study were used to compute our dependent variables. The result for each variable consisted of the sum of the item scores corresponding to each factor or scale. After confirming that missing values were randomly distributed, multiple imputations were performed with the MICE package of R (Buuren & Groothuis-Oudshoorn, 2011). Imputation was only performed when there were less than 5% missing values for each case and variable.
These variables were Child-to-Parent Violence Questionnaire for the father and the mother (composed of four factors: physical, psychological, financial, and control/domain), and reactive and instrumental reasons for CPV. Individual variables are hostile attribution, anger, access to aggressive responses, positive consequences of aggression, empathy, justification of violence, impulsivity, and drug use. Family variables are warmth-father, warmth-mother, critical-father, critical-mother, direct home victimization, and vicarious home victimization, and social variables are direct school victimization, vicarious school victimization, direct street victimization, vicarious street victimization, and deviant peers. In our study, our grouping variable was CPV group, with three levels CPV-JI youth (detected CPV), CPV non-JI youth (undetected CPV), and non-CPV youth.
Based on the most complex condition (comparison of the three groups) of our proposed experimental design, we used the G*Power program (Faul et al., 2009) to calculate the statistical power of our contrast. This analysis showed that with a sample size of 306, across three groups, for an effect size F = 0.25, with a 95% confidence level, the estimated statistical power would be 0.98.
In the analytical process, we first compared the mean scores of all dependent variables, except CPV variables, for the three groups. This analysis will be used to evaluate the possible differences of each measure collected through the scales as a function of membership in each of the three groups (CPV-JI youth, CPV non-JI youth, and non-CPV youth). The effect size of the comparison between the three groups will be represented by partial eta squared (η2p), while a pooled standard deviation Cohen’s d will be used for the effect sizes for bivariate comparisons between groups. In addition, to determine the relationship between the variables evaluated in the study, we analyzed the Pearson correlation matrix of these variables, first as a whole (the three groups), and then separating the CPV groups (CPV-JI youth and CPV non-JI youth).
Next, we performed a multiple logistic regression in which the dependent variable to be predicted in the model analyzed was the CPV groups (CPV-JI youth/detected CPV and CPV non-JI youth/undetected CPV), and all the variables assessed through the scales were used as predictor variables in the model, with the exception of the variables assessed through the Child-to-Parent Violence Questionnaire. This analysis was proposed with the aim of analyzing which variables were predictive of belonging to the detected CPV group or to the undetected CPV group.
Then, through an analysis of variance (ANOVA), we analyzed whether there were differences between the two CPV groups (CPV-JI youth and CPV non-JI youth) for the mean scores of each of the CPV total scores in the case of fathers and mothers (CPV-F and CPV-M), as well as of the subfactors measured through this scale (psychological, physical, financial, and control/domain violence) and the reasons (reactive and instrumental).
Finally, using the receiver-operating characteristic (ROC) test, we analyzed the diagnostic capacity shown by the CPV scales on the type of CPV group (CPV-JI youth/detected CPV and CPV non-JI youth/undetected CPV). This analysis proposes, through logistic regression, to analyze to what extent one or more variables have the capacity to predict membership of each individual evaluated to a group or category (detected vs. undetected). CPV-JI group (detected) was given a value of 1 in the parameter-coding scheme and CPV non-JI group (undetected) was given a value of 0. Through this analysis, we can detect the percentage of those classified correctly versus incorrectly, which provides a cut-off point for the dependent variable and allows for classification of cases into each group. Sensitivity is the probability of correctly classifying an individual whose actual status is defined as positive with respect to the condition being studied by the test, which is why it is also called the fraction of true positives (positive predictive value [PPV]). Specificity is the probability of correctly classifying an individual whose true status is defined as negative. It is equal to the result of subtracting the false-positive fraction (negative predictive value [NPV]) from one. This analysis allows us to calculate (and represent) the area under the curve (AUC), which is the percentage of cases that the variable is able to classify correctly. The AUC is an overall measure of the accuracy of a diagnostic test. It is defined as the probability of correctly classifying a pair of positively diagnosed and negatively diagnosed individuals, randomly selected from the population, by the results obtained when applying the diagnostic test to them. Thus, the AUC gives us the probability that a positive person randomly selected from the population will be correctly detected. AUC values can range from 0.5 (random predictor) to 1 (perfect predictor). AUC values are usually labeled as poor, moderate, and excellent predictors, although the values that comprise these ranges vary greatly depending on the area of knowledge (De Hond et al., 2022). In the field of law and human behavior, it is recommended to interpret AUC values as 0.56 for small, 0.64 for medium, and 0.71 for large detection ability (Rice & Harris, 2005). For all hypothesis tests, the significance level was set at .05. The analyses in this study were carried out using jamovi (The Jamovi Project, 2022) and R software version 4.1.2 (R Core Team, 2022).
Results
Risk Factors in Detected CPV, Undetected CPV, and Non-CPV
Table 2 shows the mean scores and their standard deviations for each of the measures for the three groups (CPV-JI youth, CPV non-JI youth, and non-CPV youth). In addition, the results of the contrast of means between the three groups for each measure are presented, as well as the significant results in posteriori planned binary comparisons between the groups. As can be seen, except for empathy, all variables showed a significant effect. The analysis of the planned comparisons showed that, in all cases (except for positive consequences of aggression), the scores differed between CPV youth (detected and undetected) and non-CPV youth. Specifically, concerning individual factors, youth who assaulted their parents showed higher levels of hostile attribution, anger, access to aggressive responses, justification of violence, impulsivity, and drug use. In terms of family factors, aggressor youth report greater perceptions of criticism and less warmth from their parents, as well as higher levels of direct and vicarious home victimization. Finally, concerning social factors, youth aggressors report higher levels of exposure to violence at school and in the street, along with more deviant peer groups. Furthermore, except for the positive consequences of aggression, all comparisons between CPV non-JI youth (undetected CPV) and non-CPV youth showed a simple effect, with CPV non-JI youth showing higher scores on risk factors than non-CPV youth. In the case of the contrast between the two CPV groups (detected and undetected), it was significant for the variables aggressive response access, drug use, criticism from father, vicarious street victimization, and deviant peers, with CPV-JI youth (detected CPV) showing higher levels in all variables than CPV non-JI youth (undetected CPV).
CPV-JI, CPV Non-JI and Non-CPV Group Comparisons: One-Way ANOVA and Multiple Comparisons for Individual, Family, and Social Variables
Note. CPV-JI = Child-to-parent Violence-Justice Involved; CPV non-JI = Child-to-parent Violence non-Justice Involved; non-CPV = Non-Child-to-Parent Violence. Letters represent the different significative planned post hoc comparison: aCPV-JI youth vs. CPV non-JI youth. bCPV-JI youth vs. non-CPV youth. cCPV non-JI youth vs. non-CPV youth.
p < .05, **p < .01, ***p < .001.
Table 3 presents the results of the correlations for all the variables measured in the study, taking the entire study sample (the three groups) as a whole. As can be seen, all of the variables showed significant correlations with CPV (father and mother), except the variable empathy. Table 4, similar to that presented in Table 3, shows the correlations of all the variables evaluated in the study, but in this case, it analyzes these relationships according to CPV group membership. Specifically, the correlations of the CPV-JI youth group are shaded in gray and the correlations of the CPV non-JI youth group are represented with a white background. In both groups CPV significantly correlates with several variables, finding differences between CPV father and mother (see Table 4).
Full Correlation Matrix With All the Variables
Note. CPV-F = Child-to-Parent Violence–Father; CPV-M = Child-to-Parent Violence–Mother; RR = Reactive Reasons; IR = Instrumental Reasons; HA = Hostile Attribution; AN = Anger; ARA = Aggressive Response Access; PC = Positive Consequences of aggression; EM = Empathy; JV = Justification of Violence; IMP = Impulsivity; DU = Drug Use; W-F = Warmth-Father; W-M = Warmth-Mother; C-F = Criticism-Father; C-M = Criticism-Mother; DV = Direct Victimization; VV = Vicarious Victimization; DP = Deviant Peers.
p < .05, **p < .01, ***p < .001.
Correlations of the Variables According to CPV-JI and CPV Non-JI Groups
Note. The correlations with a gray background correspond to Child-to-Parent Violence-Justice Involved Group (CPV-JI) and those with a white background correspond to Child-to-Parent Violence non-Justice Involved Group (CPV non-JI). CPV-F = Child-to-Parent Violence–Father; CPV-M = Child-to-Parent Violence–Mother; RR = Reactive Reasons; IR = Instrumental Reasons; HA = Hostile Attribution; AN = Anger; ARA = Aggressive Response Access; PC = Positive Consequences of aggression; EM = Empathy; JV = Justification of Violence; IMP = Impulsivity; DU = Drug Use; W-F = Warmth-Father; W-M = Warmth-Mother; C-F = Criticism-Father; C-M = Criticism-Mother; DV = Direct Victimization; VV = Vicarious Victimization; DP = Deviant Peers.
p < .05, **p < .01, ***p < .001.
Then, a multiple logistic regression analysis was conducted to determine which variables (other than CPV variables) were able to predict the membership of individuals in the CPV-JI (detected) and CPV non-JI groups (undetected). Positive b values will reflect prediction about the CPV non-JI group and negative values about the CPV-JI group. The full model containing all predictors was statistically significant (R2 = .464, p < .001) and correctly classified 90% of cases (see Table 5). The independent variables that made a statistically significant contribution to the model were hostile attribution, anger, aggressive response access, drug use, and vicarious victimization in the street. Concretely, the predictor variables for the CPV-JI group were the social variable vicarious street victimization (b = −0.35), and the individual variables access to aggressive responses and drug use (b = −0.33 and −0.21, respectively), whereas the predictor variables for the CPV non-JI group were the individual variables hostile attribution (b = 0.15), and anger (b = 0.25).
Logistic Regression Model to Detect Predictor Variables of Membership in the CPV-JI and CPV Non-JI Groups
Note. CPV-JI = Child-to-Parent Violence-Justice Involved; CPV non-JI = Child-to-Parent Violence non-Justice Involved. LL and UL indicate the lower and upper limits of a confidence interval, respectively. AIC = Akaike information criterion, BIC = Bayesian Information Criterion.
p < .05, **p < .01, ***p < .001.
CPV Cut-Off Scores
Next, we set out to analyze the possible differences between the groups of CPV (detected and undetected) for the variables obtained through the Child-to-Parent Violence Questionnaire. Table 6 shows the means and standard deviations of each variable analyzed for each of the groups, as well as the results of the contrast of means. Except for control/domain (father and mother) and instrumental reasons, all variables showed different scores between the two groups; this difference is due to higher scores in the group of CPV-JI youth (detected CPV).
CPV-JI and CPV Non-JI Group Comparisons: One-Way ANOVA for Child-to-Parent Violence Variables
Note. CPV-JI = Child-to-Parent Violence-Justice Involved; CPV non-JI = Child-to-Parent Violence non-Justice Involved. CPV-F = Child-to-Parent Violence–Father; CPV-M = Child-to-Parent Violence–Mother.
p < .05, **p < .01, ***p < .001.
Finally, Table 7 shows the results after ROC analysis. This analysis allowed us to know which cut-off score in the variables obtained by CPV-Q offered the best detection capacity in the classification of subjects in the CPV-JI youth (detected CPV) and CPV non-JI youth (undetected CPV) groups. Through the Youden index, we were able to establish the best cut-off scores for the scales. This cut-off point is where the greatest balance is found between the percentage of true-positive rate (sensitivity) and the percentage of false-positive rate (specificity). The best total cut-off score for the father and the mother is 16 to 17 points. That cut-off score showed the best results in predicting the detected versus the undetected group of aggressors of violence toward the father (AUC = 0.666) and toward the mother (AUC = 0.653). These values can be interpreted as follows, in the case of AUC = 0.666, it would be interpreted as a 67% probability that a randomly selected CPV-JI youth would have more severe CPV toward father than a CPV non-JI youth. It should be noted that, in the case of physical violence, the cut-off point for both scales is 1, and this value can detect 57.50% of positive cases in the case of the fathers and 60.34% in the case of the mothers. However, in the case of psychological violence, a value of nine for fathers and seven for mothers showed a positive detection capacity of 73.17% and 60.66%, respectively. In the case of financial violence, a value of three for fathers and four for mothers showed a positive detection capacity of 58.33% and 55.77, respectively. Using the DeLong test, we could verify that, except for the control/domain in the parent scale, all the factors analyzed showed significant differences with a random classifier (AUC = 0.5). This means that the factors showed an AUC (detection level of the classifier variable) significantly different from the detection behavior of a factor with random detection (AUC = 0.5). In the case of control/domain on the mother’s scale, although it showed a significant effect, the detection direction was shown to be the opposite, as the AUC value is less than 0.5.
ROC Analyses: Discrimination Properties of CPV Measures for Detecting CPV-JI From CPV Non-JI Groups
Note. CPV-JI = Child-to-Parent Violence-Justice Involved; CPV non-JI = Child-to-Parent Violence non-Justice Involved. CPV-F = Child-to-Parent Violence–Father; CPV-M = Child-to-Parent Violence–Mother; PPV = positive predictive value; NPV = negative predictive value. AUC = area under the curve. Significance level in contrast with p(H0: AUC = 0.5) for the DeLong Test = *p < .05, **p < .01, ***p < .001.
Discussion
This study aimed to compare three samples of youth: CPV-JI youth (detected CPV), CPV non-JI youth (undetected CPV), and non-CPV youth. The purpose was to examine which risk factors distinguish these three groups. The differences between the detected and undetected aggressors groups in the pattern and types of violence were also analyzed using the CPV-Q. The study was focused on determining which factors differentiate the two groups and may lead some aggressors to reach the judicial system. To our knowledge, this is the first study to compare these types of cases.
Risk Factors in Detected CPV, Undetected CPV, and Non-CPV
First, Hypothesis 1a stated that both CPV-JI youth (detected CPV) and CPV non-JI youth (undetected CPV) would have significantly higher scores on risk factors compared with youth from the non-CPV group. The results confirmed this hypothesis for practically all the factors analyzed. Thus, youth who show violent behaviors toward their parents, regardless of whether they are detected or undetected aggressors, present higher scores in all individual, family, and social risk factors than youth without aggressive behaviors, except in empathy. This is in line with previous studies on the subject in which higher levels have been found in some variables in CPV-JI versus non-CPV youth, highlighting individual variables, such as deficits in cognitive skills to solve interpersonal problems, impulsivity (Contreras & Cano-Lozano, 2015), antisocial attitudes (Contreras & Cano-Lozano, 2016b), and drug use (Contreras & Cano-Lozano, 2015; Hernández et al., 2020; Ibabe et al., 2014), as well as some family variables, such as a hostile perception of parental authority and the family environment in general, the perception of parental criticism/rejection (Contreras & Cano-Lozano, 2014, 2015), and violence exposure at home (Contreras & Cano-Lozano, 2016a; Hernández et al., 2020), and social variables, such as violence exposure at school and in the street (Contreras & Cano-Lozano, 2016a; Hernández et al., 2020). In line with our results, in the study by Simmons et al. (2020), it was also found that abusive youth showed higher levels of anger compared with non-abusive youth. However, the variable anticipation of positive consequences of aggression only differentiated between CPV-JI youth and non-CPV youth. Furthermore, no differences in empathy were found between either group, which is consistent with other studies that have already pointed out the absence of a relationship between this variable and CPV (Calvete, Gámez-Guadix, & García-Salvador, 2015; Contreras et al., 2020b; Martín et al., 2022).
Therefore, adolescents who assault parents, whether detected or undetected, present psychosocial characteristics that differentiate them from adolescents who do not exercise CPV. However, in what variables do the two types of aggressors differ? Hypothesis 1b proposed that CPV-JI youth (detected aggressors) would present significantly higher levels of risk factors than CPV non-JI youth (undetected aggressors). This hypothesis was also confirmed. Specifically, concerning individual variables, the results showed higher levels of access to aggressive response and more drug use in the detected aggressors compared with the undetected aggressors. Regarding family variables, a greater perception of criticism/rejection by the father was observed, and, finally, regarding social variables, CPV-JI youth reported more vicarious victimization of violence in the street and more deviant peer groups compared with CPV non-JI youth. Therefore, according to our results, these would be the variables that characterize the most severe cases of CPV. Similarly, the study by Loinaz and Ma de Sousa (2020) also indicates that CPV-JI youth present a worse psychological and family profile, along with more deviant peer groups, when compared with CPV aggressors from clinical samples.
The second objective was to determine the specific risk factors that predict detected versus undetected CPV. The obtained data show that the predictors of detected CPV are the individual variables aggressive response access and drug use, together with the social variable vicarious victimization in the street. Some previous studies with community samples have already pointed out the importance of the aggressive response access variable as a predictor of CPV (Contreras et al., 2020b; Orue et al., 2019), and in studies with judicial samples, it is one of the variables that differentiate CPV-JI youth from other-JI youth (Cano-Lozano et al., 2023). Drug use is a variable that appears moderately linked to CPV in community samples (Beckmann et al., 2021; Calvete et al., 2011; Cano-Lozano et al., 2020; Ibabe et al., 2013). However, it is not a variable that differentiates CPV-JI youth from other-JI youth and it seems to be part of a more general pattern of antisocial and criminal behavior (Cano-Lozano et al., 2023; Contreras & Cano-Lozano, 2015; Hernández et al., 2020). Regarding vicarious victimization in the street, previous studies have also pointed out that it is a predictor of CPV in judicial sample (Contreras & Cano-Lozano, 2016a), being also a characteristic that differentiates CPV-JI youth from other-JI youth (Cano-Lozano et al., 2023).
So far, these results reflect and confirm higher scores on risk factors in detected CPV youth compared with undetected CPV youth. According to these data and previous studies, it seems that there may be a continuum or dimensional evolution in the cases of CPV, with the most serious cases being the aggressors who reach the judicial setting and would present a worse individual, family, and social profile than the undetected aggressors, and these, in turn, more than youth who do not present violent behaviors toward their parents. Of particular interest was the finding that undetected aggressors (CPV non-JI youth) were more similar to detected aggressors (CPV-JI youth) than to non-aggressors, hence the need to identify and intervene as soon as possible in this type of case at the individual, family, and social levels. Intervention should be adapted to the settings and the needs of each case; thus, more serious cases will have more therapeutic needs.
CPV Cut-Off Scores
The third objective of this study was to determine whether there are differences in the pattern and reasons for CPV between detected and undetected aggressors. As expected, we found differences between the two groups in total CPV and practically all dimensions. Specifically, total CPV scores differed between the two groups, with higher scores in the group of detected aggressors toward both the father and the mother than in the group of undetected aggressors. In addition, the detected aggressors group had higher scores on physical, psychological, and financial violence toward both the father and the mother and on reactive reasons for CPV. No differences were found between the two groups in control and domain behaviors and instrumental reasons for CPV. Loinaz and Ma de Sousa (2020), who compared a judicial and clinical sample of CPV aggressors, also found a higher prevalence of injuries and physical and financial violence toward the mother in the judicial sample, although, in the clinical sample, slightly higher psychological violence toward the father was observed. According to the results obtained, it seems again that there may be a continuum or evolution of violence from less severe to more severe. This seems to be less evident in the control and domain behaviors toward the parents and in the instrumental use of CPV, which are present to the same extent in both detected and undetected aggressors. This suggests that before the parents make the decision on reporting their children’s behavior at the Justice System, although there is a pattern of violent behavior (psychological, financial, control over parents) from the child toward the parents, the incident of physical violence is what leads (in most of the cases) to reporting/detection. For this reason, it is fundamental to encourage parents to seek help when there is increasing psychological or financial violence, in an effort to prevent physical aggression. Educating parents about this pattern seems very important and could be a reasonable and beneficial prevention step.
Finally, the CPV-Q score that best classified detected versus undetected aggressors in both the total score and each type of violence was identified. The best total cut-off score for the father and the mother is 16 to 17 points. The cut-off scores for physical violence were lower than for other types of violence. Specifically, the existence of only one type of physically violent behavior toward the father or the mother (a score of 1) is sufficient. The best cut-off scores for psychological violence, economic violence, and control and domain behaviors were 7 to 9, 3 to 4, and 9 to 12 points, respectively. Moreover, cut-off scores were useful for correctly classifying up to 74.42% to 67.15% of aggressors into the correct group. Therefore, the CPV-Q proved to be a useful instrument in the classification of different types of CPV aggressors (detected and non-detected). In addition, this analysis with cut points may help to identify empirically supported criteria for differentiating CPV cases. This could be useful not only for research but also for professional practice for early detection of potential cases in the community population. Early detection of cases is important to help extend support services to these young people. However, we must be cautious and not label young people in a way that could be stigmatizing, requiring other sources of information.
This study has some limitations that should be taken into account. First, the sample was not representative insofar, as it was small and limited to two geographical areas of Spain. Likewise, the proportion of males and females is unequal, although this is due to the fact that there are more males than females in judicial samples. This is the reason why this proportion was maintained in the rest of the groups. It could also be that males resort more often to physical violence than females. In addition, the data are based on the self-report of youth. Further studies should include data from other informants, such as parents.
Nevertheless, this study offers important contributions to this field of study and raises some questions that should be addressed in future studies. This is the first time that detected and undetected CPV aggressors have been compared in both the pattern of violence and risk factors at different levels. Future research should identify factors other than the severity of the violence and the psychosocial risk profile that may influence the decision of families to report this type of violence through the courts, to seek professional help in a clinical setting, or not to seek professional help. Another line of research to be developed is to address the relevance of gender both in the pattern and types of violence and in the risk factors of the different types of aggressors, in the same way as in other studies with justice involved youth (general and specific judicial profiles) and in the general population (e.g., Pusch & Holtfreter, 2018), as well as the relation between the aggressor’s gender and the victim’s gender. On the other hand, this work shows the potential of the CPV-Q as a screening tool in research to assess the presence of CPV, to classify types of aggressors, or as a measure to study the effectiveness of an intervention or to assess the evolution of cases. Other practical implications are the need to detect these types of cases as soon as possible and to intervene according to the context and needs present, attending to the psychological difficulties that the adolescent may present, as well as to the problems at the family level and those present in the social environment.
Footnotes
Authors’ Note:
The authors wish to express their gratitude to the General Directorate of Juvenile Justice and Cooperation of the Junta de Andalucía and the Juvenile Justice Service of Jaén, Spain, for the authorization and collaboration in the study. This research was funded by the Ministry of Economy and Competitiveness of Spain and the European Union through the European Regional Development Fund, grant number PSI2015-65950-R.
