I. Introduction

Adolescents’ delinquent behaviors can be regarded as anti-social behaviors. Adolescents who exhibit delinquent behaviors are more likely to experience an increased incidence of unemployment, psycho-emotional problems, suicide attempts, and crime history during their later adulthood^[6]. The range of delinquency is quite wide, from hanging out with peers at the mall to risky behaviors such as drinking and smoking. There is a growing concern that the level of adolescent delinquency can become more aggressive and physically/mentally harmful to others, including behaviors such as sexual harassment, bullying, and murder, and some adolescents exhibit multiple anti-social behaviors at once^[25]. Those delinquent behaviors are not limited to offline contexts. Increasing adolescents’ screen time also predicts the possibility of delinquency in an online context. The detrimental effect of online delinquency–such as cyber bullying, posting and disseminating fake information, and illegal downloading–on victims has well recognized^[15]. Therefore, substantial research on adolescent delinquency has been conducted to figure out predictors triggering the early onset and increasing and sustaining delinquent behaviors during adolescence, in both offline and online contexts. However, most studies have examined offline and online delinquency separately, focusing on individual characteristics such as self-control, delinquent peers, or family background factors. There is little known about the mechanism through which adolescents’ delinquent behaviors interact between their offline and online contexts, and there is little research that considers the effect of adolescents’ free time. Therefore, the purpose of this study is to figure out whether online behaviors and delinquency influence offline behaviors and delinquency, and vice versa, emphasizing a longitudinal perspective with a cross-lagged panel model.

This study can provide practical implications for guiding adolescents’ behaviors that occur in the online context as well as the offline context, and if there is an adverse cyclic mechanism between adolescents’ offline and online contexts, this will indicate when to time prevention and intervention efforts so as to break the detrimental cycle.

For adolescents, online space is not an optional supplement to the offline world, nor are they considered separate settings. Rather, it has become one of the ecosystems that interacts with other ecosystems. Adolescents meet most of their offline friends online^[31]. Extant research consistently shows that time with peers is likely to be unstructured and unsupervised, regardless whether it occurs in offline or online contexts, and therefore it opens more opportunities to be involved in delinquent behaviors. However, there is very little research that considers the reciprocal influences of time with peers in offline and online contexts on offline and online delinquent behaviors. Although substantial research on adolescent delinquency has been conducted, most studies examined what factors that could be observed in offline contexts predicted delinquency in online or offline settings. In other words, there is a lack of research examining the effect of offline factors on online behaviors and the effect of online factors on offline behaviors. Therefore, this study examines how adolescents’ time with peers in offline vs. online settings influence each other during the middle school years, and how offline and online time spent with peers affects offline and online delinquent behaviors. This cross-domain analysis will help elucidate what happens in the online context, which is difficult for parents and adults to monitor, and what the effects of adolescents’ online behaviors might be. As the effect of time spent with peers on delinquent behaviors apparently differs by gender^[5], analysis is conducted separately for boys and girls.

We arrived at three research hypotheses supporting opportunity theory from the developmental perspective:

Ⅱ. Literature Review

1. Theoretical background

As adolescents grow, they are likely to detach from parents and to form stronger attachments with their peers. Parents expect their adolescent children to be more independent, autonomous, and self-sufficient. Bronfenbrenner’s ecological perspective (1986) shows how children and adolescents extend their environment as they grow and the ways in which they interact with others within/across combinations of physical, temporal, and spatial contexts over time^[4]. An individual’s environment consists of nested systems that have bi-directional influences within and among those systems. A child learns to extend his/her experience from microsystem to mesosystem, exosystem and macrosystem^{[4, 19]}. The microsystem is a person’s experience of direct activities and interactions within a particular setting, such as home or school. The mesosystem includes the interactions between two or more microsystems. The exosystem consists of settings that have indirect effects on a person. The macrosystem is the ideological system or set of beliefs–for example, religion, gender, and politics–that layers over the other three ecosystems. Among these multiple ecosystems, adolescents’ mesosystems get more complicated as they are given more free time away from their parents and family, and are allowed to involve themselves in various activities with others. Adolescents’ offline and online contexts are an example of their mesosystem, as their experiences (e.g., mood, feeling, peer relationship quality, knowledge) online and offline influence each other.

The process of expanding their environment from family to peers, and the greater amount of free time they are granted, can expose adolescents to risk factors for delinquency. Therefore, research has focused on this developmental stage to examine adolescents’ initial inclinations toward delinquent behaviors. There are two traditional viewpoints about what may lead to adolescents’ delinquency. First, from the socialization perspective, adolescents’ behaviors, especially deviant or delinquent behaviors, are directly reinforced by delinquent peers^{[34, 37]} when they are viewed as being the norm in the group of peers. On the other hand, opportunity theory suggests that more free time increases the incidence of deviant and delinquent behaviors, regardless of whether that time is spent with delinquent peers or not^[28].

The ecological perspective provides the rationale that adolescents’ behaviors offline and online influence each other, and socialization and opportunity theory can explain whether and how adolescents’ free time, especially time spent with peers, is associated with their delinquent behaviors over time during their adolescence.

2. Adolescents’ peer influences and offline delinquency

From the developmental perspective, building relationships with peers in adolescence is expected to develop social competence and empathy^{[11, 36]}. However, hanging around with peers is generally regarded as an unstructured activity and is likely associated with deviant behaviors and delinquency^[26]. Time with peers is mostly spent without parents or authority figures who can control adolescents, which means that adolescents can easily engage in deviant behaviors^[28]. Adolescents’ delinquent behaviors are magnified when peers are present, as these behaviors are seen as a way of showing their loyalty to the group, maintaining social status, and avoiding being ridiculed^[36]. The amount of time spent with peers and its effects on adolescents’ delinquency has been extensively studied^{[3, 8, 20]}, and researchers have characterized time spent with peers as unstructured and unsupervised socializing^{[13, 28]}. Agnew and Peterson (1989)^[1] included the amount of time adolescents spent talking in person, dating, and talking on the phone, in addition to hanging around/wandering, as an unstructured and unsupervised socializing activity; they found that it did indeed increase delinquent behaviors. Certain circumstances–such as being together in public places without a specific purpose and without adult supervision–are more likely to place adolescents at risk of delinquent behaviors^[38]. Even though this time did not initially include deviant peers, it opened up more opportunities for deviance^[28]. Barnes et al. (2007)^[2] found that the amount of time spent with peers, independently from the protective effect of time spent with families, significantly increased the incidence of adolescents’ risky behaviors–including drinking, smoking, drug use, and sexual activity–as well as deviant behaviors such as fighting with parents, skipping a day of school, beating up others, picking others’ pockets, and breaking into a house. In addition, the detrimental influence of time spent with peers has long been observed to be stronger among boys than girls^{[10, 29, 32]}. Most previous studies examined offline time with peers and no study specified online time with peers or examined its effect on offline delinquency.

3. Adolescents’ peer influence and online delinquency

Adolescents spend a large amount of time online to communicate, study, search information, and entertain themselves. The Internet has facilitated one of the most important developmental tasks of adolescence: emotional connection with others. Increasingly, this can be accomplished online, with no time and space limits^[24]. In the 2000s, when Internet use became prevalent among adolescents, online communication – instant messaging, chat rooms or social networking sites – was found to be beneficial to adolescents’ friendship quality^{[7, 12, 35]}. However, at the same time, concerns have arisen because adolescents’ online activities often occur without parental proactive monitoring^[27] and personal information is made public and easily searchable^[9]. This means that adolescents have more opportunities to engage in unstructured and unsupervised online activities. Bullying in Internet chatrooms and harassment through cell phone texts or messages have been increasing^[33]. It appears that social learning theory and opportunity theory, both of which attempt to explain adolescent delinquency, may also be applied to adolescents’ online delinquent and deviant behaviors. The more delinquent behaviors, such as online piracy and illegal downloading, are exposed as common among peer groups, the greater the perception that those behaviors are acceptable^[23]. Also, more online time with peers is likely to involve online delinquent behaviors^[16].

Ⅲ. Methods

1. Data

This study draws upon data from the Korean Children & Youth Panel Survey (KCYPS). KCYPS is an ongoing longitudinal and nationally representative survey that has gathered information on educational and health status, attitudes, and behaviors, and family/school environment from children and youth in Korea since 2010. Three cohort groups1), 2,342 1^st graders from 98 elementary schools, 2,378 4^th graders from 95 elementary schools, and 2,351 1^st graders from 78 middle schools participated in the survey in 2010, and the same children and their families have been recontacted and asked to complete the survey every year. By 2016, the 1^st graders in the elementary school cohort had reached 1^st grade in middle school, the 4^th graders in the elementary school cohort had reached 1^st grade in high school, and the 1^st graders in the middle school cohort had entered post-secondary schools or the workforce.

This study focused on 2,351 1^st graders in the middle school cohort for whom information was available through 3^rd grade in middle school, in order to examine their experiences in middle school. With a response rate of 97%, 2,280 completed interviews when they became 2^nd graders in middle school in 2011. In 2012, of the eligible participants, 2,259 children completed interviews. After selecting a sample who had information from 1^st grade to 3^rd grade in middle school, a total of 2,229 children were included in our study. After applying a base-year weight, data represent the population of Korean middle school students who were 1^st graders in 2010.

2. Measures

1) Delinquency

Adolescents’ offline delinquent behaviors were measured with 10 items. Example items included whether they had ever smoked, drunk alcohol, threatened others, fought, had sex, or abused others during the last one year. With responses of 1 for yes and 0 for no, a sum of the items was used in the analysis: a score of 0 means that respondents never exhibited delinquent behaviors during the last one year, 1 means that they engaged in at least one delinquent behavior, and a higher score means that they exhibited multiple delinquent behaviors.

Online delinquent behaviors were measured with 6 items. Example items included whether they had posted false information on any websites, illegally downloaded software, stolen and used others’ identification, lied about their gender or age, hacked into other machines/ websites, or used abusive expressions while chatting and posting during the last one year. The sum of the responses on these 6 items were used in the analysis: 0 means that the respondent never engaged in online delinquent behaviors during the last one year, 1 means that they exhibited at least one of those delinquent behaviors online, and a higher score means that they had more online delinquent behaviors.

2) Socializing

Offline socializing was calculated using the amount of time spent playing with friends measured in hours and minutes the respondents reported spent playing with friends per day after school during typical weekdays (Monday to Friday).

Online playing with friends was measured as the frequency of communication and interacting using a computer or a cell phone. Adolescents were asked how often they chat/message, send/receive emails, join online communities, and own and/or visit others’ blogs, Facebook profiles, and Twitter accounts: 1 for very often and 4 for never. Adolescents also reported on the frequency with which they talk on the phone and send/receive messages with friends using a (cell) phone, with choices ranging from 1 for very often to 4 for never. Online playing with friends was defined as the sum of these activities using a computer or a (cell) phone. After reverse coding, higher scores indicate more frequent communication and interactions with friends online.

3) Demographic variables

Individual and family characteristics that may influence the outcome variables (i.e., delinquent behaviors) were included as control variables. Gender was coded as 0 for female and 1 for male. Mothers’ education level was categorized into two groups: equal to or less than high school graduation, and some college education or more. Whether parents were working and the family’s annual household income were also included. Affective parental monitoring was measured with 3 items; adolescents were asked how much they agreed with these items: “My parents know my whereabouts after school”, “My parents are well aware of how I am spending time all day”, and “My parents know when/what time I come back home.” Higher scores indicate more parental monitoring.

The quality of peer relationships in terms of communication skills was measured using items asking how much they agreed with these 3 statements: “My friends respect my opinions”, “My friends listen to me while I am talking”, and “I talk to my friends about my worries and concerns” with choices ranging from 1 for strongly agree to 4 for strongly disagree. After reverse coding, higher scores indicate better peer communication skills. The level of peer trust was measured with items asking how much they agreed with these 3 statements: “I understand my friends well”, “I can talk to my friends about whatever happens”, and “I trust my friends.” Higher scores indicate more peer trust. Whether adolescents attended a gender co-ed school or a single gender school was included. Whether adolescents had friends who smoked, drank alcohol, fought, threatened, had sex, gambled, or abused others was included in our analysis. Six aggressive behaviors were summed up and included as a control variable. Example items are: how often are they angry all day long, do they interfere with others, and do they attack whenever they were stopped from doing what they wanted. Five items on whether they were socially withdrawn were summed up. Example items are “I am not comfortable with others”, “I am shy”, “I feel challenged when I have to state my opinion”, and “It is very hard to stand in front of others”. Adolescents’ stress because of their parents’ over-parenting was measured with 3 items, because over-parenting is likely to increase adolescents’ online delinquent behaviors^[22]. Adolescents were asked how much they agreed with the following statements about their parents’ behaviors: “My parents do not believe I am as capable as other adolescents my age”, “My parents always want me to win”, and “My parents always meddle in everything.” Higher scores indicate that adolescents experienced more stress because of their parents’ over-controlling and over-monitoring behaviors.

3. Analysis

Descriptive analysis results are presented in Table 1. Adolescents’ offline and online delinquent behaviors, and offline and online playing with peers from 1^st grade to 3^rd grade in middle school, as well as their individual and family backgrounds, were compared by adolescent gender.

A cross-lagged panel model using Mplus 8 was used to test direct effects of offline and online time with peers on their offline and online delinquent behaviors from 1^st grade to 3^rd grade in middle school (Figure1). Using Mplus allows for an examination of predictors and outcomes simultaneously at multiple time points during the specific period. Online socializing at 1^st grade was pathed to online socializing at 2^nd grade. Online socializing at 2^nd grade was pathed to online delinquency at 2^nd grade and online socializing at 3^rd grade. Online socializing at 3^rd grade was pathed to online delinquency at 3^rd grade. Also online delinquency at 2^nd grade was pathed to online delinquency and online socializing at 3^rd grade. Offline socializing and delinquency at each grade was pathed with same method. To examine cross-domain effects, socializing and delinquency at each grade were also pathed across offline and online contexts. For example, online socializing at 1^st grade was pathed to offline socializing and delinquency at 2^nd grade. Maximum likelihood parameter estimates with robust standard errors(MLR) were used with sampling weights. All exogenous variables were allowed to covary. Model fit was evaluated by examining specific goodness-of-fit statistics. A Root Mean Square Error of Approximation (RMSEA) that is below .06, a Comparative Fit Index (CFI) greater than .96, and a Standardized Root Mean Square Residual (SRMR) of less than .09 all indicate good model fit^[18].

Ⅳ. Results

1. Descriptive analyses

Male adolescents exhibited more online and offline delinquent behaviors during all middle school years than did female adolescents. Frequency of online socializing was greater for females from 1^st grade to 2^nd grade in middle school. There was no difference in the amount of time adolescents spent interacting with peers offline between males and females in 1^st grade and 2^nd grade. But in 3^rd grade, males spent significantly more time with peers offline compared to females.

Forty percent of adolescents had parents who had some college education or more. About 44% of adolescents had both parents working. Almost one third of adolescents lived in a two-parent household with one parent working. Nine percent lived in a single-parent family, and the remaining 13% lived with grandparents, other relatives, or guardians. Affective parental monitoring did not differ by adolescent gender. The quality of peer communication skills and peer trust adolescents perceived was significantly greater for females than it was for males. Males were more likely to attend gender co-ed schools than were females. Compared to females, males had more delinquent friends during their middle school years. Aggressive behaviors in 2^nd grade were greater for females than for males, whereas there was no difference in aggressive behaviors between females and males in 3^rd grade. Socially withdrawn behaviors did not differ by adolescent gender in 2^nd and 3^rd grade. Perceptions of stress due to parents’ over – controlling/monitoring adolescents were greater for males than for females.

2. Cross-lagged panel analysis

Figure 2 shows results of direct paths from cross-lagged panel analysis for males. The structural model was quite a good fit to the data (Table2). Online time with peers in 1^st grade increased online time with peers in 2^nd grade (β=0.38, p<.001), which increased online time with peers in 3^rd grade (β=0.31, p<.001). And in turn, more online time with peers in 3^rd grade increased online delinquent behaviors in 3^rd grade as well (β=0.10, p<.01). Online delinquent behaviors in 2^nd grade were likely to increase online delinquent behaviors in 3^rd grade (β=0.29, p<.001).

Offline time with peers in 1^st grade increased offline time with peers in 2^nd grade (β=0.14, p<.01), and subsequently increased offline time with peers in 3^rd grade (β=0.16, p<.001). Offline delinquent behaviors in 2^nd grade that were increased by offline time with peers in 2^nd grade (β=0.14, p<.01) were in turn associated with increased offline time with peers in 3^rd grade (β=0.09, p<.01), and also increased offline delinquent behaviors in 3^rd grade (β=0.31, p<.001).

Cross-domain interactions were also found. More online time with peers in 2^nd grade was associated with increased offline delinquent behaviors in 2^nd grade (β=0.11, p<.05) and increased offline time with peers in 3^rd grade (β=0.07, p<.05). Online time with peers in 3^rd grade increased offline delinquent behaviors in 3^rd grade (β=0.10, p<.05). Offline time with peers in 2^nd grade increased online delinquent behaviors in 2^nd grade (β=0.14, p<.05) and online time with peers in 3^rd grade (β=0.07, p<.05). Offline delinquent behaviors in 2^nd grade increased online delinquent behaviors in 3^rd grade (β=0.09, p<.05).

Figure 3 shows results of direct paths from cross-lagged panel analysis for females. Online time with peers in 1^st grade increased online time with peers in 2^nd grade (β=0.47, p<.001), and subsequently increased online time with peers in 3^rd grade (β=0.29, p<.001). Online delinquent behaviors in 2^nd grade were associated with increased online delinquent behaviors in 3^rd grade (β=0.36, p<.001). There were no statistically significant associations between online time with peers and online delinquent behaviors. In offline contexts, offline time with peers in 1^st grade increased offline time with peers in 2^nd grade (β=0.28, p<.001), and it subsequently increased offline time with peers in 3^rd grade (β=0.31, p<.001). Offline delinquent behaviors in 2^nd grade increased offline delinquent behaviors in 3^rd grade (β=0.29, p<.01). For females, online/offline cross-domain interactions were not found.

V. Discussion and Conclusion

This study examined the reciprocal influences of time with peers in offline and online settings on adolescents’ offline and online delinquent behaviors during their middle school years. The average of offline and online delinquency was significantly greater for male adolescents than for female adolescents. Consistent with findings in the extant literature, female adolescents’ peer relationships were characterized by more online socializing time, better peer communication skills and a higher level of peer trust. On the other hand, male adolescents reported having more delinquent peers compared to females. Socio-demographic background factors, such as parental education level and family structure, did not differ by gender.

We found that offline time with peers in 1^st grade in middle school was positively associated with offline time with peers in 2^nd grade, and offline time with peers in 2^nd grade was positively associated with offline time with peers in 3^rd grade subsequently. The same pattern applied to online contexts for both males and females. Offline delinquent behaviors in 2^nd grade were positively associated with offline delinquent behaviors in 3^rd grade for both males and females. For the effects of time with peers on delinquency, adolescents’ offline socializing was observed to increase offline delinquency in 2^nd grade, which in turn increased offline socializing in 3^rd grade. More online socializing among males increased their online delinquency in 3^rd grade. But this negative influence of time with peers was only observed among male adolescents. Regarding online and offline cross-domain influences, males’ offline socializing increased their online delinquent behaviors, and online socializing increased their offline delinquency in 2^nd grade. Male adolescents’ online socializing in 3^rd grade increased their offline delinquency in 3^rd grade. And again, such complex cross-domain influences were not observed among female adolescents. For adolescent males, online friends are not much different from their offline friends, so the same peer approval behaviors may be exerted in both offline and online contexts. Offline delinquency may play a role in connecting with peers online when they are not together offline, and their delinquent behaviors may be maintained in the online context.

Surprisingly, cross-domain influences of offline and online socializing with offline and online delinquency were found only among male adolescents, not female adolescents. This finding of gender difference is supported by previous research showing that males’ peer approval and peer inclusion behaviors increased delinquency, which in turn increased the amount of time spent with peers^[30]. That no association between time with peers and delinquent behavior was found for female adolescents may be because female adolescents are likely to talk about their day and share their lives with their parents, and the time they spend with peers may not be as unstructured and unsupervised. This is consistent with Heimer and De Coster’s suggestion (1999)^[14] that female adolescents, compared to males, maintain strong bonds with family and parents, and this plays a protective role in female adolescents’ behaviors. Another possible explanation is that females use the Internet to communicate with peers whereas males use the Internet to play games^[17], which means that male adolescents may need to update, download, and join various websites for games. There may be more shares about free or low cost methods to obtain what they need, but probably these methods are illegal. In this case, the negative influences of time with peers on their behaviors easily translate between online and offline contexts, especially for males, because playing games is an important aspect of peer bonding among males. In order to develop effective interventions, it is necessary to identify what motivates male adolescents to engage in delinquent/illegal behaviors online. Also, identifying what factors have been playing a protective role for female adolescents is needed to examine whether the same factors can be used in interventions for male adolescents.

We should note some limitations of the current study. One limitation is that the measure of adolescents’ offline and online time with peers may be biased because the items were self-reported. The second limitation is that whether parents or adult figures were present, and where adolescents were when they spent time with peers online was not specified. Lastly, it was not identified whether adolescents talk to their parents about their online activities, which can be regarded as parental monitoring. However, despite these limitations, this study contributes to the literature by examining adolescents’ activities and behaviors in online and offline contexts simultaneously, using longitudinal data, and finding that male adolescents’ delinquent behaviors easily transfer between online and offline settings.

Adolescents now live in offline and online settings simultaneously, and their experiences in those settings come and go constantly, sometimes with no awareness of shifting between settings. Offline and online are the same world to adolescents. Our study found detrimental reciprocal influences between online and offline contexts, especially for male adolescents. Considering that most adolescents do not tell their parents about their online lives^[21], interventions need to take into account the ways in which adolescents’ online time can be monitored by parents, and how such adverse cross-domain influences can be prevented. Our study findings suggest several practical prevention and intervention implications in the field of youth welfare. Especially for male adolescents, careful and closer monitoring of their online behaviors is needed to prevent male adolescents’ delinquency. This will be well-achieved when parents at home and teachers at school provide consistent monitoring. What triggered offline delinquency may be found from studying adolescents’ online context, which can be difficult for adults to observe. What makes male adolescents willing to engage in illegal online behaviors can also be found from examining their offline peer relationship quality and offline delinquent behaviors. Also, our findings indicate that there might be a particularly effective time to intervene for male adolescents, given that the 2^nd grade of middle school is the most vulnerable time for peer relationships and risky contexts. Before male adolescents develop their delinquency during later adolescence, early interventions for online behaviors as well as offline behaviors should be planned. Additionally, education programs focusing on moral and ethical behaviors online are urgently needed, especially for male adolescents. Less punitive laws for online delinquency and a lack of information on what behaviors are illegal can make male adolescents less sensitive to the real harm that may result from online delinquency, and unaware of desirable and sound behavioral development. For female adolescents, there were very weak associations between online and offline behaviors; however, as a prevention program, close monitoring and a moral/ethical education program during the middle school years, as suggested for male adolescents above, is needed to prevent later delinquent behaviors.