Parenting Practices and Types of Adolescent Problematic Behaviour

Two remarkable studies changed the conception of measuring parental monitoring (^{Kerr & Stattin, 2000}; Stattin & ^{Kerr, 2000}). These works proposed that actually the level of parental knowledge about children’s life was the construct assessed by most of the studies rather than the effort to actively monitor children. The research of these authors presented parental knowledge as a variable comprised of three sources to get through them such a knowledge:

First, the children could tell them spontaneously, without any prompting (child disclosure). Second, parents could ask their children and their children’s friends for the information (parental solicitation). Third, parents could impose rules and restrictions on their children’s activities and associations, thereby controlling the amount of freedom children have to do things without telling them (parental control) (Stattin & ^{Kerr, 2000}, p. 1073).

Since then, several studies have analysed these sources of knowledge in relation to problematic behaviour. Child or adolescent disclosure is considered the main predictor of parental knowledge (^{Kerr, Stattin, & Burk., 2010}). Adolescent disclosure was consistently and directly related to lower levels of antisocial behaviour and substance use (e.g., Ahmad, Smetana, & Klimstra, 2015; ^{Hoeve et al., 2009}; ^{Willoughby & Hamza, 2011}). Regarding parental control and parental solicitation, results have not been so consistent. Some studies found that parental control (e.g., ^{Parra & Oliva, 2006}; ^{Willoughby & Hamza, 2011}) and parental solicitation (e.g., ^{Keijsers, Frijns, Branje, & Meeus, 2009}; ^{Lippold, Greenberg, Graham, & Feinberg, 2014}) were significantly related to lower levels of antisocial behaviour and substance use. Nevertheless, other research found that parental control and parental solicitation were not directly related to problematic behaviour (^{Gault-Sherman, 2012}; Muñoz, Pakalniskiene, & Frick, 2011; Rekker, Keijsers, Branje, Koot, & Meeus, 2017), whereas in other studies control and solicitation were even positively related to antisocial behaviour and substance use (^{Laird, Marrero, & Sentse, 2010}; ^{Willoughby & Hamza, 2011}). These contradictions may be based on differences in definition and operationalisation of the variables among studies. Nevertheless, the inconsistent effects of parental control and parental solicitation point out the need to evaluate these sources separately. This contributes to avoid biased results and mistaken conclusions about the significance of parenting variables labelled under the knowledge construct.

Another common parenting variable in research has been parental support. Parental support can be defined as a general construct of parent-child affective relationships, including warmth, attachment, and involvement (^{De Haan, Prinzie, & Dekovi, 2012}). Parental support has been often presented as a significant predictor of lower levels of antisocial behaviour and substance use (e.g., ^{Hoeve et al., 2009}; ^{Parra & Oliva, 2006}; ^{Wang, Dishion, Stormshak, & Willett, 2011}). However, several studies did not find significant direct prospective relationships of parental support, warmth, or attachment and aggressive behaviour and rule-breaking (^{De Haan et al., 2012}), delinquency (^{Meldrum & Barnes, 2017}; ^{Yun, Cui, & Blair, 2016}), and substance use (^{Pereyra & Bean, 2017}). Other research has even found that parental support was positively related to aggression, delinquency, nonviolent behaviour, and substance use in adolescence (^{Cutrín, Gómez-Fraguela, & Sobral, 2017c}), specifically in those adolescents with low levels of empathy or involved in gangs (^{Van der Graaff, Branje, De Wied, & Meeus, 2012}; ^{Walker-Barnes & Mason, 2004}). As similarly occurs with parental control and parental solicitation, the conceptual broadness of parental support may be one of the reasons for the contradictory findings as regards problematic behaviour. In addition, the existence of indirect relationships between different risk factors and behaviour may also explain such inconsistence. Therefore, it should be taken into account that other variables may be influencing the effects of parenting practices on problematic behaviour.

Mediation Effects of Parenting through Deviant Peers

The interest in research about the mediation effects of deviant peers on the relationship between parenting and problematic behaviour has substantially increased in the last decade. In this regard, evidence has shown that negative parenting practices are indirectly related to problematic behaviour through increasing the affiliation with antisocial peers in adolescence, and these results were found across contexts and in several types of problematic behaviour. For instance, research has indicated that the effects of low monitoring and low parental knowledge on the adolescent level of delinquency were mediated by deviant peer affiliations (e.g., Dynes, Domoff, Hassan, Tompsett, & Amrhein, 2015; O’Donnell, Richards, Pearce, & Romero, 2012; ^{Walters & Espelage, 2018}). In this line, some studies found that both low parental control and low maternal support were indirectly related to involvement in delinquent activities via deviant peer affiliations (^{Deutsch, Crockett, Wolff, & Russell, 2012}), and that low parental attachment was indirectly related to delinquency through the affiliation with deviant peers as well (^{De Vries, Hoeve, Stams, & Asscher, 2016}). Likewise, some studies have analysed these mediation effects on more specific types of problematic behaviour. For example, ^{Van Ryzin and Dishion (2013)} evidenced that family coercive interactions led to coercive joining with peers and, consequently, to violent behaviour in early adulthood. Other research indicated that parenting comprising monitoring, attachment, discipline, and guidelines was indirectly related to the level of violent behaviour in adolescence through the affiliation with deviant peers (^{Haggerty, Skinner, McGlynn-Wright, Catalano, & Crutchfield, 2013}). Similarly, ^{Trudeau, Mason, Randall, Spoth, and Ralston (2012)} found that parenting composed of affect, discipline, guidelines, and monitoring indirectly predicted, through deviant peers, externalizing behaviour including violent, aggressive behaviour and nonviolent, rule-breaking behaviour. Other studies also showed that deviant peer affiliations mediated the effect of permissive parenting on higher levels of aggressive behaviour and rule-breaking (^{Hinnant, Erath, Tu, & El-Sheikh, 2016}). As regards substance-related problems, some studies found that low rule-setting monitoring was indirectly related to higher levels of adolescent substance use by the mediation of being affiliated with substance-using peers (^{Kiesner, Poulin, & Dishion, 2010}). In this line, other research evidenced that the effects of monitoring and quality in family relationships were mediated by deviant peer affiliations for both tobacco and alcohol use (^{Van Ryzin, Fosco, & Dishion, 2012}). The recent study of Cox, Criss, Harrist, and Zapata-Roblyer (2017) specifically indicated that being affiliated with deviant peers mediated the effect of parenting on externalizing behaviour, whereas being affiliated with substance-using peers mediated the effect of parenting on adolescent substance use.

Gender Differences in Antisocial Behaviour and Risk Factors

Research has consistently found the existence of a gender gap in antisocial behaviour and delinquency, especially when violence is considered; that is, females present lower levels of delinquency, general antisocial behaviour, and violent behaviour in comparison with males (^{Lanctôt, 2015}; ^{Zahn-Waxler & Polanichka, 2004}). Nevertheless, current research has indicated that both females and males present similar levels of substance use in adolescence, including tobacco, alcohol, and other illicit drugs (e.g., ^{Colell, Sánchez-Niubò, & Domingo-Salvany, 2013}). As regards risk factors, parenting and peers seem to influence males and females differently but findings are not consistent. Literature has traditionally proposed that females are more influenced by family factors. Females tend to be more monitored by their parents and have more supportive and communicative relationships (^{Javdani, Sadeh, & Verona, 2011}; ^{Kerr et al., 2010}). However, some research has indicated that males seem to be more vulnerable than females to the lack of positive parenting practices (e.g., ^{McAdams, Salekin, Marti, Lester, & Barker, 2014}). On the other hand, evidence has proposed that males are more influenced by deviant peers because they are more frequently involved in antisocial groups during adolescence (e.g., ^{Trudeau et al., 2012}). Nevertheless, research has also suggested that females are more likely to affiliate with opposite-sex peers and, therefore, to be more exposed to the influence of older males (e.g., ^{Poulin, Denault, & Pedersen, 2011}).

The Current Study

The current work aimed to analyse the prospective effects of parenting practices on adolescent problematic behaviour taking into account the mediation effects of deviant peer affiliations. In the Spanish context, some attempts of analysing these effects have been previously carried out. In this regard, previous studies found that parental knowledge, parental support, and parent-adolescent conflict were indirectly related to violent and nonviolent behaviour (^{Cutrín, Gómez-Fraguela, Maneiro, & Sobral, 2017}; ^{Cutrín, Gómez-Fraguela, & Sobral, 2017a}), as well as substance use (^{Cutrín, Gómez-Fraguela, & Sobral, 2017b}) through the affiliation with deviant peers. The mediated relationship of parenting practices and antisocial behaviour through deviant peers was also found for Spanish juvenile offenders (^{Cutrín, Gómez-Fraguela, & Luengo, 2015}). Nevertheless, all these studies were carried out using a cross-sectional design and, therefore, leading to preliminary conclusions that should be tested at a longitudinal level in order to check the generalization of these results. In order to address this research gap in the Spanish context, the current study aimed to confirm the direct effects of parenting variables on later problematic behaviour, as well as the mediation effects of deviant peers on such relationship. This study also aimed to clarify some previous inconsistent results regarding the role of specific parenting practices. For that purpose, different parenting variables were analysed: two main positive parenting practices (parental knowledge and parental support) and, in a separate model, the three sources of parental knowledge proposed by the Kerr and Stattin’s classical works (adolescent disclosure, parental control, and parental solicitation; Kerr & ^{Stattin, 2000}; Stattin & ^{Kerr, 2000}). In addition, direct and mediation effects were analysed separately on several types of problematic behaviour (violent and nonviolent antisocial behaviour, and substance use) and also examined in female and male adolescents. This work, then, is proposed as a preliminary study – and to the best of our knowledge, the first in our context – regarding longitudinal mediated relationships between parenting and problematic behaviour through deviant peer affiliations in normative Spanish adolescents. The specific hypotheses proposed in the current study were: (1) parenting practices directly predict deviant peer affiliations and problematic behaviour; (2) deviant peer affiliations are directly related to problematic behaviour; (3) parenting practices indirectly predict problematic behaviour through increasing the affiliation with deviant peers; and (4) no gender differences exist in direct and mediated relationships.

Participants

The initial sample was composed of 666 adolescents in 1^st grade of compulsory secondary education [1º ESO] from 13 state secondary schools in Galicia (NW Spain). Three subjects were removed from the sample because more than 90% of their responses were missed. Thus, the final sample at time 1 (T1) was composed of 663 adolescents aged 12 to 15 (M = 12.49, SD = 0.68) and gender-balanced (54.3% male). From the final sample, 78.4% of adolescents lived with both parents, 16.9% lived only with their mother, 2.3% lived only with their father, and 2.4% lived with other relatives. Youths presented similar cultural and social characteristics regarding ethnicity and socio-economic background, most of them being white (more than 90%) and coming from middle and low-middle income contexts. In order to obtain data from the four provinces of Galicia, 24 schools were selected by convenience sampling. In those centres that agreed to participate (5 in rural and 8 in urban contexts), data were collected in all groups of 1^st grade and participation exceeded 90%. The follow-up at time 2 (T2) was carried out one year after the first data collection. The level of attrition between the two data collections was 25.1% (N_T2 = 499). Seven subjects were removed from the sample because more than 90% of their responses were missed. Significant differences were found between participants and non-participants in T2 regarding gender, χ²(1) = 4.61, p = .032, age, t(663) = 5.72, p < .001, and antisocial behaviour in T1, F(1, 602) = 37.296, p < .001, non-participants being mostly males, older, and showing higher frequency of antisocial behaviour than participants.

Measurements

Parental knowledge (T1). The degree of parental knowledge regarding adolescent’s whereabouts, activities, and friendships was measured in T1 by a self-reported 5-item scale based on previous scales (i.e., ^{Kerr & Stattin, 2000}; Stattin & ^{Kerr, 2000}; e.g., “Your parents know what you do during your free time”; α = .78). Items were scored from 0 (never) to 3 (always). The lambda coefficients in the structural models were between .60 and .69.

Parental support (T1). Parental warmth, responsiveness, and closeness were assessed in T1 by a self-reported 8-item scale validated in normative Spanish adolescents (Oliva, Parra, Sánchez-Queija, & López, 2007; e.g., “You feel supported and understood”; α = .90). This scale was scored from 0 (never) to 3 (always). The lambda coefficients in the structural models were between .60 and .84.

Adolescent disclosure (T1). The degree in which adolescents spontaneously reveal information about their life with their parents was measured in T1 using a self-reported 5-item scale based on previous scales (i.e., ^{Kerr & Stattin, 2000}; ^{Oliva et al., 2007}; Stattin & ^{Kerr, 2000}; e.g., “You tell them what you do when you go out”; α = .82). Items were scored from 0 (never) to 3 (always). The lambda coefficients in the structural models were between .49 and .85.

Parental control (T1). The level of rules and restrictions established by parents to control the amount of children’s freedom was assessed in T1 using a self-reported 5-item scale based on previous scales (i.e., Kerr & ^{Stattin, 2000}; ^{Oliva et al., 2007}; Stattin & ^{Kerr, 2000}; e.g., “Before you go out on Saturday, they require you to tell them where you are going and with whom”; α = .83). Items were scored from 0 (never) to 3 (always). The lambda coefficients in the structural models were between .59 and .78.

Parental solicitation (T1). The degree in which parents directly ask adolescents for information was evaluated in T1 using a self-reported 5-item scale based on previous scales (i.e., ^{Kerr & Stattin, 2000}; ^{Oliva et al., 2007}; Stattin & ^{Kerr, 2000}; e.g., “They start conversations with you about your free time”; α = .73). Items were scored from 0 (never) to 3 (always). The lambda coefficients in the structural models were between .50 and .75.

Deviant peers (T2). The presence of antisocial behaviour in the peer group was assessed in T2 by an 11-item scale developed and validated in normative Spanish adolescents (Cutrín, Maneiro, Sobral, & Gómez-Fraguela, 2018b). This scale presents two subscales to assess general antisocial behaviour (seven items; e.g., “They get into trouble in their free time”; α = .83) and substance use in the peer group (four items; e.g., “They take ‘legal’ drugs (tobacco or alcohol)”; α = .87). The global score presented an internal consistency of α = .88. Items were scored from 0 (never) to 3 (always). The lambda coefficients in the structural models were between .60 and .97.

Violent behaviour (T2). Violence and physical aggression were assessed in T2 by a 6-item scale of the short Spanish version of the Antisocial Behaviour Questionnaire validated in Spanish adolescents (ABQ; Luengo, Otero-López, Romero, Gómez-Fraguela, & Tavares-Filho, 1999) and used in previous studies with normative Spanish adolescents (e.g., Cutrín, Gómez-Fraguela, ^{Maneiro et al., 2017}; e.g., “Fighting and hitting someone”; α = .74). Items were scored from 0 (never) to 3 (very often). The lambda coefficients in the structural models were between .34 and .83.

Nonviolent antisocial behaviour (T2). Nonviolent forms of antisocial behaviour were evaluated in T2 by three 6-item scales of the ABQ (^{Luengo et al., 1999}) used in previous studies with normative Spanish adolescents (e.g., Maneiro, Gómez-Fraguela, Cutrín, & Romero, 2017): rule-breaking scale (e.g., “Spending the night out without permission”; α = .71), theft scale (e.g., “Taking something from class without permission with the intention of stealing it”; a = .73), and vandalism scale (e.g., “Setting fire to something: a dustbin, table, car, etc.”; a = .75). The global score presents an internal consistency of α = .81. Items were scored from 0 (never) to 3 (very often). The lambda coefficients in the structural models were between .55 and .94.

Frequency of substance use (T2). The frequency of tobacco, alcohol, and cannabis use throughout life was assessed in T2 by 3 items scored from 0 (never) to 5 (almost every day). The global score presents an internal consistency of α = .75. The lambda coefficients in the structural models were between .61 and .95.

Procedure

Compliance with ethical standards was taken into account throughout the investigation. The study was presented to the heads of 24 secondary schools. In the 13 centres that agreed to participate in the study, parental consent was requested and, subsequently, adolescent assent was obtained at the moment of the questionnaire implementation. Qualified psychologists visited these centres during school hours in order to explain the objectives and provide proper instructions to the adolescents who answered the self-reported scales in their classrooms. The total questionnaire was composed of 150 items to be answered in a class period of 50 minutes. Adolescent participation was voluntary, and anonymity and confidentiality of information were totally guaranteed.

Data Analysis

IBM SPSS Statistics 20 and MPLUS 7 were used to conduct the statistical analyses. Firstly, reliability, descriptive, and variance analyses across gender and controlling for age were carried out. Secondly, partial correlations among the variables of the study controlling for age were conducted for both genders. Lastly, two types of structural equation models were tested in order to analyse the existence of mediation effects separately in both genders. The first type examines the effects of two general parenting-related variables (i.e., parental knowledge and parental support) on adolescent problematic behaviour through affiliation with antisocial peers; also called ‘parenting model’. The parenting model was tested separately for violent behaviour, nonviolent antisocial behaviour, and substance use. The second type examines the effects of three specific sources of parental knowledge (i.e., adolescent disclosure, parental control, and parental solicitation) on adolescent problematic behaviour through the affiliation with antisocial peers; also called ‘sources model’. The sources model was also tested separately for violent behaviour, nonviolent antisocial behaviour, and substance use. A combination of maximum likelihood (ML) and bootstrapping (b = 5,000) were used in order to maximize accurate estimations under a non-normal distribution and estimate bias-corrected 95% confidence intervals for indirect effects (^{Hancock & Liu, 2012}). The goodness-of-fit indexes χ²/df, CFI, RMSEA, and SRMR were used to estimate the models, considering as criteria for an optimum fit χ²/df < 2-3, CFI > .95, RMSEA and SRMR < .05, and for an acceptable or reasonable fit χ²/df < 4, CFI > .90, and RMSEA and SRMR between .08 and .10 (^{Hu & Bentler, 1999}).

Preliminary Analyses

As can be seen in Table 1, descriptive results indicated that females reported significantly higher levels of parenting-related variables (i.e., parental knowledge, adolescent disclosure, parental control, and parental solicitation), except for parental support, which did not show significant differences between genders. On the other hand, males reported significantly higher frequencies of antisocial-related variables (i.e., antisocial peers, violent behaviour, nonviolent behaviour, and substance use). Effect sizes indicated that violent and nonviolent antisocial behaviour were the variables most influenced by gender (see Table 1).

Table 1 Descriptive Results of MANOVA for all Study Variables by Gender and Controlling for Age 

Note. η² = eta squared effect size.

Table 2 Partial Correlations among the Study Variables for both Genders and Controlling for Age 

Note. The coefficients in parentheses correspond to males coefficients.

Know (T1) = knowledge; Supp (T1) = support; Disclos (T1) = disclosure; Contr (T1) = control; Solicit (T1) = solicitation; D. peers (T2) = deviant peers; Viol (T2) = violent behaviour; Nonviol (T2) = nonviolent behaviour; Subst (T2) = substance use.

*p < .05. **p < .01. ***p < .001.

Table 3 Goodness-of-Fit Indexes of the Structural Equation Models 

***p < .001.

All the parenting-related variables were significantly and positively intercorrelated similarly in females and males (see Table 2). Parental knowledge, parental support, and adolescent disclosure were significantly and negatively correlated with deviant peer affiliations in both genders, and significantly and negatively correlated with problematic types of behaviour, exclusively in males in the case of violent behaviour. Parental solicitation was only significantly and negatively correlated with deviant peers, violent and nonviolent behaviour in males, and parental control was not significantly correlated with any antisocial-related variable in females nor males. Regarding antisocial-related variables, deviant peers, violent behaviour, nonviolent behaviour, and substance use were significantly and positively intercorrelated in both genders.

Structural Equation Models for Mediation: Direct and Indirect Effects

Figure 1 Conceptual parenting model examining direct and mediation effects of parenting variables (i.e., knowledge and support) and deviant peer. Affiliations on different types of problematic behaviour (i.e., violent and nonviolent antisocial behaviour and substance use). 

All the tested structural equation models have obtained mostly acceptable fit indexes (see Table 3). Regarding the parenting model (see Figure 1), three structural models were tested, one for each type of problematic behaviour. As shown in Table 4, parental knowledge and parental support were significantly correlated in the three structural models (r = .66, p < .001 for females and r = .60, p < .001 for males). Parental knowledge in T1 significantly and negatively predicted deviant peer affiliations in T2 for both genders in the three models, and deviant peers in T2 were significantly and positively related to violent behaviour, nonviolent behaviour, and substance use in T2 for both genders. In addition, parental knowledge in T1 also significantly and negatively predicted substance use in T2 for males. Parental support did not predict any variable in T2 (see Table 4). Regarding the explained variance by the parenting model, R² of violent behaviour was .13 and .50 for females and males, respectively; R² of nonviolent behaviour was .51 and .71 for females and males, respectively; R² of substance use was .77 and .52 for females and males, respectively; and R² of deviant peers ranged from .12 to .20 for females and from .16 to .29 for males in the three models.

Figure 2 Conceptual sources model examining direct and mediation effects of the sources of parental knowledge (i.e., disclosure, control, and solicitation) and deviant peer affiliations on different types of problematic behaviour (i.e., violent and nonviolent antisocial behaviour and substance use). 

Table 4 Results of Parenting Models Analysing Direct Effects of Parenting Variables (i.e., Knowledge and Support) and Deviant Peer Affiliations on Problematic Behaviour for both Genders 

Regarding the sources model (see Figure 2), three structural models were also tested, one for each type of problematic behaviour. As can be seen in Table 5, adolescent disclosure and parental control were significantly correlated in the three models (r = .44, p < .001 for females and r = .59, p < .001 for males) as well as adolescent disclosure and parental solicitation (r = .63, p < .001 for females and r = .69, p < .001 for males), and parental control and parental solicitation (r = .87, p < .001 for females and r = .92, p < .001 for males). Only adolescent disclosure in T1 significantly and negatively predicted deviant peer affiliations in T2 for females in the three models, and deviant peers in T2 was significantly and positively related to violent behaviour, nonviolent behaviour, and substance use in T2 for both genders. Regarding the explained variance by the sources model, R² of violent behaviour was .17 and .50 for females and males, respectively; R² of nonviolent behaviour was .53 and .74 for females and males, respectively; R² of substance use was .76 and .45 for females and males, respectively; and R² of deviant peers ranged from .18 to .24 for females and from .10 to .17 for males in the three models.

Table 5 Results of Sources Models Analysing Direct Effects of the Sources of Parental Knowledge (i.e., Disclosure, Control, and Solicitation) and Deviant Peer Affiliations on Problematic Behaviour for both Genders 

Lastly, Table 6 displays the mediation effects. For the parenting model (see Figure 1), the bootstrapping results indicated that only parental knowledge in T1 significantly predicted violent behaviour, nonviolent behaviour, and substance use in T2 through the mediation of deviant peer affiliations for both females and males. Although the indirect path “knowledge-peers-violent behaviour” was not associated to a significant value of probability for females (p < .05), the effects were included in a confidence interval which did not contain zero; therefore, this effect is considered significant. For the sources model (see Figure 2), the bootstrapping results indicated that no source of parental knowledge in T1 significantly predicted problematic behaviour in T2 through deviant peers. Although the coefficients was not associated to a significant value of probability (p < .05), for females indirect effects of adolescent disclosure through deviant affiliations on the three types of problematic behaviour were included in confidence intervals which did not contain zero; therefore, these effects are considered significant. For males the indirect path “knowledge-peers-nonviolent behaviour”, although included in a confidence interval which did not contain zero, is not considered significant because adolescent disclosure was not directly associated with deviant peers.

Table 6 Standardized Indirect Effects of all Parenting-Related Variables on Problematic Behaviour through Affiliation with Deviant Peers for both Genders 

Note. CI = Confidence Interval.

Limitations and Future Directions

Some limitations should be taken into account for the proper interpretation of the results of the current study. Firstly, although this study has overcome the main limitation of previous studies by using a longitudinal design, data were collected only in two times and, consequently, deviant peer affiliations and problematic behaviour were analysed in the same wave. More times of follow-up are needed in order to analyse the independent variables, the mediator, and the criterion variables in different longitudinal waves. Secondly, all the measures in the current study were self-reported by adolescents; therefore, the results may be partly influenced by shared method variance. Other sources of information, such as parents or peer reports, should be included in future data collections. Moreover, as stated before, future studies should focus on deeply analysing gender differences in the strength of relationships and their implications. In addition, no moderation effects among parenting variables or between socio-demographic covariates and parenting variables were tested. Moderated relationships as well as bidirectional effects could add possible explanations regarding the non significant effects found for some of the variables; therefore, other studies should take into account these effects to better explain problematic behaviour in adolescence.