2. Materials and Methods
2.1. Participants and Procedures
We used data collected in a 6-year longitudinal research project launched in the 2009–2010 school year, which involved more than 3000 students in 28 Hong Kong secondary schools. More information on the project is outlined elsewhere [23
]. The project was reviewed and approved by the Human Subjects Ethics Sub-Committee at a public university in Hong Kong (Hong Kong Polytechnic University). All related parties including adolescent participants, schools, and parents gave well-informed written consent prior to the commencement of the research project.
Among the 3328 students (age 12.59 ± 0.74 at Wave 1) who completed the survey at the first wave (Wave 1), 2905 students completed the survey at Wave 2 (attribution rate = 12.7%) and 2860 completed the questionnaires at Wave 3 (attrition rate = 14.1%). Across first three waves (Waves 1–3) during the junior secondary school stage (i.e., Grades 7–9), 2669 student participants were successfully matched and this matched sample was utilized in this study. The time interval between two consecutive waves was one year. The mean age of the matched sample was 12.56 ± 0.71 years at Wave 1. Among the participants in the matched sample, 1321 (49.5%) were boys, 1344 (50.4%) were girls, and 4 students (0.01%) did not indicate their gender information.
The matched sample (N = 2669) and the 659 students who withdrew from the study after Wave 1 (i.e., dropouts) were compared regarding their basic demographic profile, baseline delinquency level and parent-child subsystem qualities. It was found that the matched sample consisted of a higher percentage of girls (Chi square (degree of freedom is 1) = 39.70, p < 0.001, effect size (φ) = 0.11). Besides, the mean age of the matched sample (age 12.56 ± 0.71 at Wave 1) was slightly younger than that of the dropouts (age 12.72 ± 0.86 at Wave 1, F (1, 3233) = 35.52, p < 0.001, η2 = 0.02). With reference to delinquency, the matched sample showed a lower baseline level (F (1, 3115) = 50.54, p < 0.001, η2 = 0.01). While no significant difference was found in father’s behavioral control (F (1, 3190) = 0.26, p = 0.61), a slightly lower baseline level of both parents’ psychological control, a slightly higher baseline level of mother’s behavioral control and better relationships with both parents were observed among the matched sample (F values ranged from 5.40 to 35.52, p-values were <0.05, and η2 ranged from 0.002 to 0.01). As the effect size was not large in the observed differences, it was concluded that attrition is not a major problem in this study.
The project used a questionnaire including multiple measures [58
], among which adolescent delinquency and the three indicators of parent-child subsystem qualities were the focus of this study.
Delinquency was assessed using a validated scale measuring the frequency of participants’ involvement in delinquent behavior (i.e., “How often do you have each of the twelve delinquent acts listed below during the past one year?”) [59
]. The delinquent acts included “stealing”, “cheating others”, “truancy”, “running away from home”, “damaging others’ properties”, “beating others”, “gang fighting”, “speaking foul language”, “having sex with others”, “staying away from home overnight without parental consent”, “bullying”, and “trespassing”. Participants gave their answers on a 7-point reporting scale (0 = “never”; 6 = “more than 10 times”). In this scale, delinquent behavior was defined as actions that harm others or break certain rules, regulations or norms. Of note, these delinquent acts consisted of both illegal behavior and minor offences that are considered risk with reference to family, school, and social norms. The scale showed acceptable construct validity and reliability in previous studies [59
]. In the present study, confirmatory factor analyses showed that the scale was unidimensional and had adequate factorial validity and reliability across the three waves as indicated by the mean factor loading (over 0.50), average variance extracted (over 0.25), and composite reliability (over 0.70) for details, see [56
]. Besides, this scale showed acceptable internal consistency at Waves 1–3, with Cronbach’s α ranging from 0.79 to 0.84 and mean inter-item correlation coefficients ranging from 0.25 to 0.33 (see Table 2
). Furthermore, item-total correlations of the 12 items exceeded 0.20 across waves. Given these reliability figures, adolescent delinquency was indicated by the mean score of the 12 items in the present study.
Parent–Child Subsystem Qualities were measured by a Chinese scale entitled Parent-child Subsystem Quality Scale (PCSQS). The scale has been validated among Hong Kong Chinese adolescents and demonstrated good reliability in previous studies [61
]. The PCSQS included two subscales pertinent to paternal and maternal factors respectively. Three aspects of parental factors were measured by 17 items in each subscale. First, behavioral control subscale included seven items (e.g., “My father/mother asked me about what I did after school”). Second, the subscale of psychological control consisted of four items (e.g., “Dad/Mom often wants to change my mind or feelings for things”). Finally, the remaining six items belonged to parent-child relationship subscale (e.g., “I shared my feelings with my father/mother”). A 4-point measuring scale (1 = “strongly disagree”, 4 = “strongly agree”) was used. As shown in Table 2
, the Cronbach’s α across different measures included in the PCSQS ranged between 0.80 and 0.91 across different time points, indicating good internal reliability of the measures.
2.3. Data Analysis Plan
First, we performed reliability, descriptive, and correlational analyses for measures of delinquency and parental factors. Second, to address the first and the second research questions, individual growth curve (IGC) models were analyzed to examine how parental factors predict the initial level of adolescent delinquency (i.e., the first question) and the change rate of adolescent delinquency (i.e., the second research question). Similar to IGC models tested in previous studies [23
], time was the Level-1 predictor (i.e., within-subjects variable), which was nested into the Level-2 predictors (i.e., between-subject variables) including both control variable (i.e., gender) and parental factors. The values of time were set as “0”, “1”, and “2” at the three consecutive waves, respectively. Finally, four 2-level IGC models were analyzed and compared:
Model 1: “the unconditional mean model” which did not include Level-1 or Level-2 predictor;
Model 2: “the unconditional linear growth model” that only involved time as the Level-1 predictor;
Model 3: “the linear growth model” that further included gender as a Level-2 predictor (covariate);
Model 4: “the linear growth models” that further included all the parental factors as key Level-2 predictors. Specifically, Model 4a, 4b, and 4c tested parents’ behavioral control, psychological control, and parent-child relationship quality, respectively.
Gender–based IGC models were also tested to examine potential moderating effect of children’s gender on parenting influences. In this study, model fit was marked by three frequently used indices: the −2log likelihood, Akaike information criterion (AIC), and Bayesian information criterion (BIC). For all the indices, a smaller value informs a better model fit. To facilitate IGC modeling, gender was coded as follows: “female” = “−1”, “male” = “1”. In addition, parental factors were standardized. Furthermore, to reduce the distribution skewness of adolescent delinquency, a natural logarithmic transformation was conducted prior to IGC analyses.
Finally, we carried out several multiple regression analyses to investigate the relative contributions of paternal factors and maternal factors to adolescent delinquency (i.e., the third research question). Parental factors’ cross-sectional predictive effects were examined at each wave. For longitudinal predictive effects, we investigated how baseline parental factors measured at Wave 1 predict later adolescent delinquency behavior measured at Wave 2 and Wave 3. Furthermore, to explore moderating effect of children’s gender, the interactions between each parental factor and children’s gender were tested in the regression analyses.
The current study investigated the influence of three parental characteristics on the development of adolescent delinquency among Hong Kong Chinese adolescents during the junior high school years. We addressed three research questions in this study. First, we separately investigated paternal and maternal impacts on the initial level of adolescent delinquency. Three aspects of parental characteristics were referred to, including not only behavioral and psychological control but also parent-child relationship quality. Second, we examined these parental factors’ effects on the rate of change in adolescent delinquency. Finally, paternal and maternal factors’ relative cross-sectional and longitudinal contributions to adolescent delinquency were also examined. While some previous studies have focused on one or two of these questions among Western adolescents [3
], no research has addressed them all together in one single study in Chinese contexts.
For the level of adolescent delinquency, boys showed a higher initial delinquency level than female peers and this difference remained during the 3-year period, which echoes previous findings on gender differences [31
]. Besides, a similar increasing trend of delinquency level was observed among boys and girls. Furthermore, children’s gender did not significantly affect the linkages between parenting factors and delinquency, which appeared to be in line with the findings of recent meta-analyses [2
]. Regarding parental factors’ predictive effect (i.e., the first research question), as expected, behavioral control of both parents and their relationships with children negatively predicted the initial adolescent delinquency level (i.e., Hypotheses 1a, 1b, 1e, and 1f were supported). The findings were congruous with past research which suggested that positive parenting in terms of behavioral control and good parent-child relationship protect children from problem behaviors [2
]. However, the present study did not identify significant predictive effects of both parents’ psychological control on the initial level of adolescent delinquency (i.e., Hypotheses 1c and 1d were not supported). One possible explanation is that the present Chinese adolescents tended not to perceive parental psychological control as negative parenting.
In Western countries where autonomy and individual independence are emphasized, it has been theoretically and empirically verified that paternal psychological control which harms children’s sense of autonomy and self-representation concepts (e.g., self-identity and self-esteem) damages children’s psychological well-being and leads to problem behaviors [2
]. In contrast, Chinese culture values collectivistic interdependence and family obligations (e.g., respect for parents, filial piety, etc.). Hence, parents’ psychological control might not be as harmful as it is in Western societies. Chinese children who are socialized to comply with traditional values might have more positive feelings toward psychologically controlling parents [65
]. Indeed, Chinese children are more inclined to perceive parents’ control as a way to show love and concern but less likely to regard authoritarian parenting behaviors as controlling or intrusive [66
]. Furthermore, past research has suggested that Chinese parents’ dysfunctional parenting did not increase children’s problem behavior [68
]. Another possibility is that Chinese people may regard certain elements of psychological control (such as guilt induction) as a culturally acceptable way of child rearing [70
However, contrary to the above thesis, Luebbe et al. [71
] found that Chinese parents’ psychological control also result in internalizing problems such as depression and hopelessness in their children. Similarly, Shek and collaborators’ recent work underscored a significant and relatively robust positive relationship between parents’ psychological control and children’s Internet addiction among Chinese adolescents in Hong Kong [23
]. In addition, the present individual growth curve (IGC) analyses did not reveal a significant predicting effect of psychological control on the initial level of adolescent delinquency. Yet, the regression analyses suggested that there was a positive correlational association between both parents’ psychological control and children’s delinquency, although the effect sizes were not very high. Concerning these findings, we may conjecture that parental psychological control is also likely to negatively affect adolescent development in Chinese communities, but the magnitude of the effect may depend on the indicator of developmental outcome. It is possible that psychological control of Chinese parents has a stronger influence on other developmental outcome measures (e.g., depression and Internet addiction) than on adolescent delinquency. More research is needed to test this possibility by involving a more comprehensive set of measures including both internalizing (e.g., depression) and externalizing problems (e.g., delinquency) and comparing the predictive effects of parental psychological control.
With reference to the second research question, in line with the general consensus, we also observed an increasing trend of adolescent delinquency during junior high school years. Besides, the level of delinquency among boys increased faster than that of girls. This gender difference is also consistent with previous findings [8
]. However, all the hypotheses relating parental factors to the rate of change in adolescent delinquency were not supported. While maternal behavioral control and relationship between children and both parents did not show significant predictive effects, a higher level of paternal behavioral control was associated with a faster growth in adolescent delinquency, which is discordant with our expectation. This unexpected effect of fathers’ behavioral control was inconsistent with previous findings that parents’ firmer behavioral control or better relationship between father and child predicted a slower increase or a faster decline in adolescent externalizing behavior [3
]. However, our finding is concordant with some previous research which found that positive parenting characteristics including sufficient behavioral control and better relationship between children and parents predicted a slower drop in adolescent Internet addiction [23
One can argue that high paternal behavioral control’s protective effect may weaken over time, which result in the present association between high baseline paternal behavioral control and faster increasing in adolescent delinquency. It is quite plausible since that the parenting–delinquency link is less significant among older children with the influence of parents weakening as children mature and other social relations exerting greater influences [2
]. The less significant effect of paternal behavioral control over time can also be reflected from the present observation that parental factors accounted for less variance in adolescent delinquency at Wave 3 (6.8%) than at Wave 1 (9.0%). Nevertheless, adolescents with stronger baseline paternal behavioral control demonstrated a lower delinquency level across the three waves than other participants with a relatively lower paternal behavioral control, suggesting the critical role of paternal behavioral control. Additionally, the effect of parent-child relationship qualities did not show a similar declining trend as paternal behavioral control. It is possible that the influence of some parental factors (e.g., behavioral control) is more likely to decline than others (e.g., relationship quality) [23
]. Given the equivocal findings in the past research and the present study, the predictive effects of parental characteristics on the growth rate of adolescent delinquency are not conclusive. Hence, more replication studies should be carried out in future.
For the third research question that concerned differential cross-sectional and longitudinal predictive effects of different parental factors, three observations can be highlighted. First, for cross-sectional effect, compared with maternal factors, paternal parenting characteristics explained a slightly higher percentage of variance in adolescent delinquency across the three years. However, for longitudinal effects, baseline father-child and mother-child subsystem factors accounted for a similar percentage of variance in future adolescent delinquency. Based on these findings, we contended that paternal impact may be equally strong or even stronger as compared with maternal impact on the development of adolescent delinquency. This interpretation could be compatible with previous findings showing similar or greater effects of paternal parenting [2
]. This conclusion makes sense because fathers are less engaged in socializing children but they play a more authoritative and dominant role than mothers in family, especially in Chinese communities [75
The second observation is that paternal impact may be stronger in one specific aspect, while there will be a greater maternal impact in other aspects. Specifically, the present study found that at Waves 1 and 2, fathers’ behavioral control was more influential than mothers’ behavioral control whereas mothers’ relationship with children exerted stronger impact than father-child relationship. Such differences became insignificant at Wave 3, possibly due to gradually decreased parental influence on children as aforementioned. The findings indicate that relative maternal and paternal impacts should be considered in relation to a specific parental factor. For example, Bean et al. [76
] reported a significant predictive effect of fathers’, but not mothers’ support on children’s depression and a significant effect of mothers’, but not fathers psychological control on children’s academic achievement. Recently, Shek et al. [64
] found that fathers were more influential on adolescent Internet addiction by behavioral control while mothers were more influential by psychological control. This pattern of different maternal impact versus paternal impact can help explain the mixed findings in past studies as most of them only investigated one aspect of parenting such as psychological control [77
Finally, parent-child relationship qualities, particularly mother-child relationship quality showed the most robust longitudinal predictive effect on adolescent delinquency. This observation echoes one finding of present IGC modeling that the baseline parent-child relationship qualities did not predict the growth rate of delinquency (i.e., the effect of baseline parent-child relationship qualities on delinquency level did not decline over time). These findings confirm the assertion that not only parents’ monitoring and control of adolescent activities, but also children’s satisfaction with control and willingness to communicate, prevent deviant behaviors. It supports scholars’ previous contention that a good relationship between parents and children operates preventively and is “a two-way process including both the parents’ solicitation or knowledge and control of their children’s behavior and the children’s willingness to make their parents part of their lives” [39
] (pp. 1083). Even more, without a good parent-child relationship, children may perceive strong parental behavioral control as intrusive and gain less positive influence or even negative influence from parental behavioral control [35
]. Thus, our findings help draw further attention to the need for including parent-child relationships when investigating parental control.
The findings of this study have implications for the development of prevention or intervention parent training programs. Notably, there is a dearth of parent training programs focusing on delinquency in Hong Kong. In Piquero et al.’s [78
] study, 78 studies that evaluated the parent training programs on delinquency were reviewed, but only one study was conducted in Hong Kong. Because interventions need to be guided by theory and based on valid empirical research findings, our findings on the associations between multiple parenting characteristics and adolescent delinquency serve as the theoretical framework for the parent training in Hong Kong. First, the significant effects of parental behavioral control, particularly paternal behavioral control, suggest that parent training programs should train parents to rule and actively monitor their children so as to enhance parents’ knowledge on children’s whereabouts and conduct. Second, given that parent-child relationship demonstrated robust effects on delinquency, parent training should also promote parental skills in improving the relationship, communication, and mutual trust with children, which in turn enhance children’s self-disclosure. Finally, training programs should involve not only mothers but also fathers. This is consistent with previous findings that parent training brought more benefits to children if fathers also attended the training [79
Despite the pioneering nature of the study and its theoretical and practical implications, it has several limitations. First, the present study relied on adolescent self-report. Past research suggested that effect sizes of the links between parenting and adolescent problems depend on the informants employed [2
]. It is commented that children tend to overestimate the negative features of their parents while parents are more inclined to overestimate positive qualities of their own parenting practice [2
]. While it remains unknown as to which informants evaluate parenting most objectively, child report measures are widely used, especially in longitudinal research. It is arguable that children know themselves better than others and what actually matters are children’s perception and experience. Another related issue is social desirability associated with self-report measures. While it is difficult to eliminate social desirability bias, the present study strived to reduce it as much as possible through emphasizing anonymity and instructing the participants to give honest answers. Nevertheless, future research will certainly benefit from collecting data through multiple informants (e.g., children, parents, teachers and peers) as well as different measuring methodologies (e.g., self-report and objective observation). Third, internalizing problems such as depression and externalizing problems such as delinquency can co-occur in a homotypic (either
comorbid internalizing or
comorbid externalizing problems within individuals) or heterotypic (comorbidity of both
externalizing problems within individuals) way [12
]. However, the present study only measured delinquency. Hence, we are not sure about the relationships between adolescent delinquency and other emotional or behavioral problems; and whether such relationships would moderate the parenting effect. Future study should involve more forms of adolescent problems and test their interactions. Finally, our study only involved Hong Kong Chinese adolescents. To improve the universality of the present findings, future research should address relevant research questions in other Chinese subcultures such as mainland China and Taiwan, and compare findings obtained from different samples.