More Than Just Friends

Abstract

The study of peer-group processes has a rich history in criminology. The dramatic growth in online social network websites has fundamentally changed peer-group interaction; however, relatively little research has considered how socialization processes observed in traditional interaction translate to online interaction. Using a sample of 583 undergraduate students from a mid-southern university, this study explores the concurrency between self-reported offending and exposure to criminal behavior in social network websites. Results demonstrate a strong, positive association between individual behavior and exposure to criminal behavior in online networks, suggesting that the processes underlying traditional social interaction also characterize online interaction. These results underscore the importance of online networks for understanding the etiology of criminal behavior.

Keywords

online social networks peer networks offending

Introduction

A wealth of criminological research has established peer influence to be among the strongest predictors of crime and delinquency (see Pratt et al., 2010, for a recent review). The concurrency between individual and peer behavior is one of the most consistently reported findings in the literature (Pratt et al., 2010; Warr, 2002). This association has been observed for both self-report and official data, across social contexts, and for a variety of antisocial and otherwise criminal behaviors. The dramatic increase in online social networking websites (SNS for short) over the past two decades fundamentally changed the ways in which adolescents and young adults interact with their peers. Young people now spend considerably more time interacting digitally than even 5 years ago (Lenhart, Purcell, Smith, & Zickhur, 2010; Smith & Brenner, 2012). The advent of SNS has altered the composition of youth social networks. Adolescents and young adults typically have larger online networks than personal networks and their friendships are no longer limited to the traditional geographic confines of the neighborhood and school (Acar, 2008).

SNS have the potential to serve as powerful agents of socialization, as youth are exposed to diverse attitudes and models of behaviors from a wide range of actors. Indeed, emerging research suggests certain health-related behaviors, such as medical decision making, are diffused throughout SNS (Centola, 2010; Lefebvre & Bornkessel, 2013). It bears to reason that attitudes or messages supporting maladaptive behaviors, like violence or substance use, may be similarly diffused throughout virtual networks; however, the potentially criminogenic influence of SNS has received little attention in the empirical research.

The following analysis takes a step toward bridging this gap in the literature by examining the association between exposure to criminal behavior in one’s SNS and self-reported offending. Most contemporary explanations of peer influence focus on one of two competing processes: social learning and behavioral homophily. The social learning perspective suggests that attitudes and models of behavior are transmitted through interpersonal interaction within social networks. The behavioral homophily perspective argues that “birds of a feather flock together” or people tend to choose friends who are similar to themselves.

While social learning and behavioral homophily may be viewed as competing hypotheses, both reflect a degree of behavioral concurrency that is imperative to establish before further theoretical tests are performed. Therefore, this study is oriented around a relatively straightforward question—does the concurrency between individual and peer behavior observed in traditional social networks exist in online social networks? The answer to this question is the first step to identifying how online peer behavior might influence individual off-line offending. While scholars have previously considered the influence of traditional criminological theories on online offending, this study focuses specifically on off-line behavior. The empirical analyses unfold through a series of negative binomial models regressing exposure to criminal behavioral on SNS on self-reported offending among a sample of students at a mid-southern university. The implications of the findings are discussed in terms of their contribution to the broader criminological literature as well as in relation to the role of SNS in the etiology of criminal behavior.

Theoretical Framework

Two theoretical perspectives are commonly employed to explain the concurrency between individual and peer behavior: social learning theory and behavior homophily. While they are often viewed as competing explanations, both predict a strong degree of similarity between individual behavior and the behaviors to which one is exposed to by peers. Social learning theory assumes attitudes favoring deviance are the result of interaction with intimate social groups (Akers, 1998; Sutherland & Cressey, 1974). The key premise of the theory revolves around the notion that “most of the learning in criminal and deviant behavior is the result of direct, and indirect, social interaction in which the words, responses, presence, and behavior of other persons directly reinforces behavior” (Akers & Jensen, 2006, p. 40). Differential reinforcement, or the balance of anticipated and actual rewards or punishments following a behavior, will impact the likelihood the behavior will continue (Akers, 1998). Definitions in favor of and against law violation are learned through exposure to others and serve as the content of what is differentially reinforced. These definitions encompass an individual’s perception and attitudes toward various behaviors. Additionally, individuals may model or imitate the behaviors of their peers, especially when these behaviors are positively reinforced. Evidence suggests imitating both primary and secondary groups serves as a means of learning normative definitions, which may then affect delinquent behavior (Strayer, Wareing, & Rushton, 1979). While social learning theory predicts that learning takes place in intimate social groups, larger peer groups may have the capacity to influence behavior through the strength of weak ties (Granovetter, 1973). Individuals may receive social reinforcement by a larger number of peers in which there is frequent contact, as well as exposure to information or behaviors that individuals embedded in smaller networks may not typically be exposed.

Conversely, the association between individual and peer behavior may reflect a tendency of individuals to select friends based on shared interests. This perspective, typically referred to as behavior homophily, indicates delinquent or criminal behavior may be one such characteristic that brings friends together (Glueck & Gleuck, 1950; Gottfredson & Hirschi, 1990). Thus, any association between individual behavior and the behavior displayed in one’s SNS may reflect the greater likelihood for offenders to select into online networks where criminal behavior is common. The general consensus in the criminological literature is that both learning and selection processes are at work, such that offenders tend to associate with other offenders, which then reinforces future behavior. However, some research indicates learning processes may have a potentially greater effect than selection (Simons-Morton & Chen, 2006; Urberg, Luo, Pilgrim, & Degirmencioglu, 2003).

Literature Review

SNS Usage

SNS refer to a broad category of websites that allow individuals to interact with others through a public or semipublic, user-created profile (Boyd & Ellison, 2007). The key distinction between SNS and other forms of online media is the ability for users to engage in interpersonal interaction while online. Although there are a multitude of SNS, Facebook is the most popular by far. As of December 31, 2012, there were 1.06 billion active users and an average of 618 million daily users (Facebook, 2013). Given the population estimate of 2012, one out of seven people in the world has an active Facebook account (U.S. Census Bureau, 2013). Twitter, another popular SNS, is used by approximately 15% of adults in the United States and accessed by roughly 8% of the U.S. population on a typical day (Smith & Brenner, 2012). SNS have seen dramatic growth over the course of the past decade (Lenhart & Madden, 2007), indicating people are increasingly turning to SNS to either create or maintain their social networks.

This trend is most notable among adolescents and young adults. For instance, recent estimates indicate that 72% of Internet users access SNS on a regular basis (Smith & Brenner, 2012). Young adults, particularly those between the age of 18 and 29, access these websites at a much higher rate. An estimated 89% of Americans in this age range regularly visit SNS. Moreover, 82% of those between the age of 14 and 17 and over half of those between the age of 12 and 13 are regular users of SNS. The typical youth between the age of 8 and 18 spends an average of 34 minutes using social media per day (Rideout, Foehr, & Roberts, 2010).

There are a few notable differences between online networks and traditional social networks. For one, online networks tend to be much larger than traditional social groups (Acar, 2008). While the quality of relationships is similar, SNS allow users to connect with significantly more friends than would be possible in a traditional peer group. In terms of online behavior, research by Pempek, Yermolayeva, and Calvert (2009) suggests that the majority of users spend their time observing content posted by others rather than generating their own content. Their research also demonstrated that individuals are more likely to interact with preexisting friends relative to those they met through the SNS.

An important distinction between traditional and online networks is that friendships can transcend traditional neighborhood and school boundaries. In other words, networks are more likely to consist of individuals who neither live in one’s neighborhood nor attend one’s school. This has two potential consequences. First, individuals interact with a greater pool of potential peers, increasing exposure to modes of behavior they might not otherwise encounter. Second, online networks allow users to maintain friendship ties as they change residences and schools. This may limit the ability for some adolescents to knife off connections to delinquent pasts, as they attend college and secure employment after high school.

Social Learning and SNS

In order for tenants of social learning theory to be applicable to the types of relationships developed, maintained, and nurtured via SNS, certain key criteria must be met. First, a level of trust needs to be established in order to produce interactions which may affect behavior. That is, users must believe the information they are viewing is an accurate depiction of reality. Acquisti and Gross (2006) reported 78% of public information provided on SNS reflects accurate information. More importantly, self-reports indicated that only 2% of respondents knowingly provided inaccurate information on SNS. Given the low percentage of users who admit to posting false information, it is possible that individuals may believe other users also post relatively high amounts of true information. It may be inferred that most participants on these websites should believe they are viewing accurate information about their peers.

The quality and intimacy of relationships fostered through SNS continues to grow. In recent years, users have become much more private (Dey, Jelveh, & Ross, 2012). A user may limit what could be perceived as negative behavior if an outside audience has access to their profile. A sense of privacy creates a more realistic environment in which to socialize. Consistent with this trend, the distinction between off-line and online networks has become less apparent in recent years. As a result, users now identify as strongly with their online communities as with their own families (Lehdonvirta & Räsänen, 2011). This indicates a high level of trust in online relationships. Users believe the principles applying to face-to-face contact mirror those of electronic methods (McKenna & Bargh, 1998). Research continues to support the notion that SNS is an extension of in-person interaction. Importantly, online interaction typically occurs at a much higher rate than in-person interaction, suggesting SNS may serve as an additional avenue of reinforcement.

Collectively, these trends indicate that SNS may provide a unique form of social interaction with intimate social groups. Consistent with the core tenets of social learning theory, we may reasonably expect attitudes and behaviors observed and reinforced in online communities will influence individual behavior. Emerging research suggests that processes of reinforcement and diffusion are indeed present in some facets of online interaction. For instance, one study reports depressed college students are more likely to discuss their symptoms when they receive positive feedback from their peers online (Moreno et al., 2011). Public health research indicates medical information shared on SNS has a direct effect on users’ decisions for chronic disease management, medication, and approach to diet and exercise (Lefebvre & Bornkessel, 2013). Likewise, Centola (2010), in an experimental approach, demonstrates that health-related behaviors are diffused throughout online networks, with the greatest level of diffusion being achieved within large networks characterized by a greater number of “weak ties.”

Selection in SNS Networks

Alternatively, concurrency between individual and peer behavior on SNS may reflect the tendency of individuals to affiliate with others who have similar interests, opinions, and engage in similar behaviors. For instance, Lewis, Gonzalez, and Kaufman (2011) report shared interests between individual and peer interests on Facebook to a large degree reflect preexisting similarities. The authors contend online interaction has less to do with influencing neighbors and more to do with strengthening social ties among those whom users already resemble. Likewise, the gap between online and off-line networks continues to narrow. A majority of individuals use SNS to maintain existing relationships instead of finding new ones (Subrahmanyam, Reich, Waechter, & Espinoza, 2008). Multiple studies conclude there are a few differences between the online and off-line communities within which college students interact (Boyd & Ellison, 2007; Pempek, Yermolayeva, & Calvert, 2009; Subrahmanyam et al., 2008). Therefore, SNS relationships may be formed for the purpose of digitally maintaining ties with preexisting networks. As a result, behavioral concurrency may reflect underlying similarities that initially contributed to the formation of the online friendship.

Online Networks and Criminal Behavior

Much of the criminological research on SNS focuses on their capacity as platforms for cyber-victimization. On the whole, this research indicates youths are increasingly likely to be exposed to criminal behavior online, either by being a personal victim of harassment or threatening behavior or by witnessing exchanges between a cyber-bully and his or her victim. Harassment takes on many forms including repetitive hateful messages, tricking victims into share personal information, denigration by posting untrue information, and excluding or ostracizing victims (Marcum, 2010). While these behaviors encompass traditional forms of bullying, SNSs provide a forum that allows victims to be harassed or threatened at all times of the day regardless of the physical proximity to the offender (Holt & Bossler, 2014). In a review of literature between 2004 and 2010, Luxton, June, and Fairall (2012) report that lifetime cyber-bullying victimization rates ranged from 20.8% to 40.6% while offending rates ranged from 11.5% to 20.1% among Internet users.

Similarly, SNS allow for the facilitation of gang-related violence (Patton, Eschmann, & Butler, 2013). Popular video hosting websites are an avenue for the dissemination of insults and threats, which may lead to violent behavior. Studies have shown online gang behavior as an extension of street behavior, with the exception that members are able to reach a much larger audience than may otherwise have been possible (Pyrooz, Decker, & Moule, 2013). King, Walpoole, and Lamon (2007) found 70% of gang members reported it was easier to find and maintain online friendships compared to forming these relationships on the streets. SNS can also be used as a means to build social capital, and such sites are frequently employed to recruit new members and coordinate action, as well as to broadcast gang affiliation and to boast about fights or murders (Patton et al., 2013). This process takes place through techniques that mirror street methods used to develop a collective and group identity (Pyrooz et al., 2013). Law enforcement agencies have infiltrated SNS in order to monitor illicit activities which has driven technological and investigative innovation in policing strategies. Currently, 96% of agencies use SNS in some capacity and 86% use SNS for the purpose of criminal investigations (International Association of Chiefs of Police, 2013).

Other studies focus on the capacity of SNS to reinforce extreme behaviors characteristic of hate groups. Viewers of these websites are exposed to hate-inspired violence through motives conveying lessons of discrimination, techniques for committing acts of terrorism, and rationalizations for racial superiority (Hawdon, 2012; McDonald, Hortman, Strom, & Pope, 2009). These websites can be viewed as a vehicle of reinforcement, strengthening preexisting attitudes and biases. Recruitment of these individuals into hate groups and terrorist organizations benefits from the cheap outreach on websites, control over group image, and anonymity. Virtual communities allow regular interaction and encourage others to become supportive of movements (Bowman-Grieve, 2009). Additionally, SNS provide a platform for sexual deviants, such as pedophiles, to share images, locate victims, and maintain relationships with similar individuals. The need for a concealable identity for these offenders becomes irrelevant due to the limitless bounds of the Internet (McKenna & Bargh, 1998). The behavior that may otherwise remain hidden from society is reinforced through identifying others with similar views.

The unique attributes present in the online environment allow offenders to engage in a variety of other offenses. The Internet is an appealing instrument for locating victims, finding opportunities, and identifying co-offenders (Newman & Clarke, 2003). Cybercrime, or “offenses where special knowledge of cyberspace is used to violate the law” includes identity theft, pornography, cyber-trespassing, and hacking (Furnell, 2002; Wall, 2001). Holt, Bossler, and May (2012) argue traditional explanations of criminal behavior, notably social learning and low self-control, also help explain multiple forms of cybercrime. The authors stress that future research should continue to apply the key tenets of traditional criminological theory to Internet-based offending.

Multiple studies have utilized these traditional theories in examining various forms of cybercrime. For example, peer associations were found to be the most significant predictor of digital piracy (Higgins, Marcum, Freiburger, & Ricketts, 2012; Holt, Bossler, & May, 2012; Holt & Morris, 2009). Elements of social learning theory have been incorporated into forms of cyber offending including imitation of peers in digital piracy (Hinduja, 2003; Holt & Copes, 2010), definitions favorable to software law violation (Higgins & Marcum, 2011), and positive reinforcement in piracy (Ingram & Hinduja, 2008). Other studies have amended traditional perspectives to account for the unique attributes of the Internet, such as Reyns, Henson, and Fisher’s (2011) cyber-lifestyle routine activities theory and Morris and Higgins’s (2009) integration of self-control, social learning, micro anomie, and techniques of neutralization to explain digital piracy. In their assessment of cybercrime scholarship, Holt and Bossler (2014) stress that the increased growth of technological advancements should result in criminological research addressing the development process of adolescents, especially in terms of how virtual socialization has affected the age-crime curve.

To the best of the authors’ knowledge, only two studies to date have applied criminological theory to explain how SNS usage affects off-line criminal behavior. As part of a larger study on peer effects, Weerman, Bernasco, Bruinsma, and Pauwels (2013) examined the effects of multiple forms of social interaction on delinquency. The authors reported the association between online socialization was significantly weaker than other forms of interaction, and in some cases, the empirical models failed to detect any effect of online interaction on self-reported offending. Meldrum and Clarke (2013) expanded upon this study by applying Osgood, Wilson, O’Malley, Bachman, and Johnstons’ (1996) and Osgood and Anderson’s (2004) concept of unstructured socialization to help explain the link between online and off-line behavior. The authors posited the rewards from the presence of peers, lack of authority figures, and unstructured time spent online would increase the risk of delinquent behavior. Consistent with this hypothesis, the authors reported a strong relationship between the amount of virtual time spent socializing and substance use and delinquency among a sample of middle school students. This underscores the potential role that SNS may play in the etiology of crime and delinquency.

Current Study

Given the increasing use of SNS, in-person interaction is being supplemented, and in some cases, replaced by online interaction. Relatively little is known about how online interaction contributes to offending behavior. On one hand, traditional social learning perspectives would predict that exposure to offending online might reinforce an individual’s likelihood to offend off-line. Conversely, the behavioral homophily perspective would indicate that individuals with a greater proclivity toward deviance would select in online networks where criminal behavior is common. While this study is not able to address temporal ordering—required for a true causal examination of the competing processes—it is able to explore the degree to which some of the key factors highlighted by previous criminological examinations also apply to online interaction. Both the social learning and behavioral homophily portend a strong degree of concurrency between individual behavior and the behavior of those who make up one’s online social network.

While the work of Meldrum and Clarke (2013) demonstrates a strong association between virtual time spent socializing and delinquency, the authors were unable to account for the characteristics of the individuals with whom respondents were interacting and the types of information to which they were exposed. Additionally, the generalizability of their study remains limited due to the sampling of a middle and high school from a rural and poor county in a southeastern state during the spring 2008. While data from 2008 would not typically be considered outdated, the rapid growth of SNS over the past 10 years resulted in considerable change in usage patterns. Facebook, the most popular SNS, was only 4 years old during this time period and was surpassed in popularity by Myspace and Friendster (Joinson, 2008). While there were 1.06 billion users of this website as of December 2012, there were only 100 million in August 2008 (Facebook, 2009, 2013).

This study builds upon this small yet growing body of research by exploring the relationship between exposure to behavior within SNS and users’ self-reported offending. Prior research has conceptualized unstructured virtual socialization as the amount of time spent communicating with friends via phone, text messaging, or e-mail (Meldrum & Clarke, 2013). This study focuses solely on online interaction with peers and includes the specific types of behavior to which respondents are exposed online. The empirical analyses are framed around a single overarching hypothesis: The level of criminal behavior displayed or discussed in an individual’s online social network will be positively associated with self-reported offending.

Method

Data

Data for this analysis come from a survey of 583 undergraduate students at a mid-southern university. College-aged students are of particular interest when analyzing SNS. As of 2011, 18 to 24 year-olds made up approximately 49% of Facebook users (Holt, 2013) and as many as 90% of college students have a SNS profile (Ellison, Steinfield, & Lampe, 2007; Morgan, Snelson, & Elison-Bowers, 2010). Additionally, college students are frequent offenders and have been used extensively for empirical criminological research (Mazerolle & Piquero, 1998; Nagin & Paternoster, 1993; Tibbetts & Hertz, 1996). Therefore, undergraduates are a suitable sample for this study.

The survey contained a series of questions gauging respondents’ online behavior, exposure to criminal behavior online, self-reported offending, and demographic characteristics. Two general education classes, Oral Communication and English Literary Heritage, were targeted as the sampling pool. While every major at the university requires a set number of general education credits, these two classes are the only ones that cannot be substituted by an alternative class. Thus, every student must take these two classes, and enrolled students are the most representative of the typical undergraduate student in terms of both academic major and sociodemographic characteristics. Surveys were administered in class between January and March 2013.¹ The response rate was rather high, with only one survey returned incomplete.

Given the nature of the study, respondents who reported they did not use SNS were removed from the analysis. This led to the removal of 13 respondents (or 2.23% of the total sample). Finally, to be consistent with prior research, the sample was restricted to students who were 29 or younger at the time of the survey, reflecting the typical college student (Bye, Pushkar, & Conway, 2007). This led to the removal of an additional 40 respondents. Outliers were present in two of the independent variables—the number of minutes spent online and the size of the network. The top 1% of the sample was trimmed, resulting in the removal of an additional eight respondents. Collectively, these restrictions resulted in a final analytic sample of 522 respondents.

Missing values on the independent variables were imputed using chained equations through the ICE command in Stata in order to preserve the original sample size (Royston, 2005). The regression models were estimated on five imputed data sets, and the results reported here reflect the average effects across the imputed data sets. Preliminary analyses revealed relatively few systematic differences between respondents with complete valid data and those missing on one or more of the key covariates. Of note, respondents with missing data were more likely to report fewer friends in their social network. Importantly, the vast majority of respondents with missing data were only missing on one variable (n = 82); the remaining were missing on two (n = 11) or three (n = 3) variables. No respondent was missing information on more than three covariates. Five cases missing on the dependent variable were removed list-wise from the sample. The results reported here comport with models in which missing data on the independent variables was handled through list-wise deletion.

Dependent Variable

The dependent variable is respondents’ self-reported offending in the 2 years leading up to the survey.² It was measured as a cumulative scale indicating the number of different offenses a respondent reported during the reference period. These included abusing an intimate partner, illegally carrying a weapon, physical fighting, selling drugs, driving under the influence (DUI), setting fire to property, theft, and vandalism. These eight offenses represent various behaviors disclosed in preliminary interviews and were common among the targeted age-group. An offending variety scale was created by summing these dichotomous measures. Variety scales are favored for their high reliability and validity and the fact that they are not compromised by high frequency of crimes perceived as being nonserious (Sweeten, 2012). Just over half of the sample (52.03%) reported engaging in at least one of the offenses and 27.3% engaged in at least two offenses. Descriptive statistics for all variables are presented in Table 1.

Table 1.

Descriptive Statistics.

	N	Percentage	Mean	SD	Min	Max	Missing
Sex	520				0	1	2 (.38%)
Male	200	38.46	—	—
Female	320	61.54	—	—
Race	522						0
White	246	47.13	—	—	0	1
Black	203	38.89	—	—	0	1
Other	73	13.98	—	—	0	1
Hesitant to post due to job	521				0	1	1 (.19%)
Yes	334	64.11	—	—
No	187	35.89	—	—
Parental access to SNS profile	520				0	1	2 (.38%)
Yes	284	54.62	—	—
No	236	45.38	—	—
Minutes spent on SNS	506		490.20	556.46	5	4320	16 (3.07%)
Friends on SNS profile	463		668.57	633.07	0	4786	59 (11.3%)
Age	522		20.37	2.00	18	29	0
GPA	500		3.12	0.52	1.78	4	22 (4.21%)
Self-offending Scale	517		1.03	1.39	0	8	5 (.96%)
SNS Offending Scale	517		2.83	2.30	0	8	5 (.96%)

Note. N = 522. Min = minimum; Max = maximum; SD = standard deviation; SNS = social networking websites; GPA = grade point average.

Independent Variables

Exposure to criminal behavior online

A cumulative scale was also used to measure exposure to criminal behavior in an online social network. Respondents were given examples of common SNS for reference including Facebook, Twitter, Reddit, Instagram, YouTube, Pinterest, Flickr, Google+, and MySpace.³ The distribution of the most popular sites accessed by these students is presented in Table 2. Notably, the vast majority of respondents reported using Facebook, Twitter, and/or Instagram. Respondents were asked how frequently they viewed comments or posts in their SNS depicting a variety of criminal and deviant behaviors. The measure of exposure to online criminal behavior is comprised of the same 8 items as the self-report offending scale. Around 81% of respondents reported being exposed to at least one behavior and 45.64% reported being exposed to at least two.

Table 2.

Frequency Distribution of Top Five Most Popular SNS Sites as Nominated by Respondents.

	Frequency	Percentage
Facebook	475	90.99
Twitter	316	60.53
Instagram	286	54.79
YouTube	59	11.30
Tumblr	35	6.70

Note: SNS = social networking websites. Respondents were allowed to select up to three social networking sites, thus percentages do not add up to 100.

SNS usage

In order to account for the various usage patterns of respondents, data were collected regarding the amount of time spent using SNS during a typical week, the size of the respondents’ social network, respondents’ willingness to post information, and whether or not the respondents’ parents accessed their profiles. Time spent online was initially measured as the number of minutes a respondent spent online during a typical week ( $\overset{ˉ}{x} = 49 0. 2$ ; SD = 556.46). The variable was log transformed in the regression models to reduce skew and ensure multivariate normality (Beauchamp & Olson, 1973). Size of the social network represents the number of friends or subscribers to a respondent’s SNS profile ( $\overset{ˉ}{x} = 668.57$ ; SD = 633.07). This variable was also log transformed. Both of these continuous variables were mean centered in the regression models. Two additional variables were included in the model to gauge respondents’ hesitancy to post information on SNS. The first asked, “Do your parents have access or subscribe to your online social networking profile” and the second asked “Have you ever been hesitant to post information due to your current job or potential future employment?”

Control variables

Additional variables were included which have traditionally been associated with offending and SNS usage. Age is measured in years at the time of the survey ( $\overset{ˉ}{x} = 20.37$ ). Race differentiates respondents who identified as non-Hispanic White (47.13%), non-Hispanic Black (38.89%), and respondents identifying as any other racial/ethnic group, including Hispanics (13.98%).⁴ Sex was a dichotomous variable differentiating male (38.46%) and female (61.54%) respondents and grade point average (GPA) was a continuous variable ranging from 1.78 to 4 ( $\overset{ˉ}{x} = 3.12$ ). Note the sample statistics comport with available data on the racial/ethnic, gender, and GPA distribution of the university population during this time frame.⁵

Analytic Strategy

Given the overdispersed, count-based measure of the offending scale, the empirical association between exposure to online offending and individual offending was assessed through a series of negative binomial regression models (Long & Freese, 2006; Osgood, Finken, & McMorris, 2002). The regression equation takes on the general form presented in Equation 1 such that:

\begin{aligned} μ = exp (β_{0} + β_{1} (e x p o s u r e t o o n l i n e o f f e n d i n g) + β_{2} - β_{5} (S N S u s a g e) + β_{6} - β_{9} (c o n t r o l s)) \end{aligned}

In this model, μ reflects the expected count of self-reported delinquent activities, β₁ is coefficient for exposure to criminal behavior in one’s social network, and β₂–β₉ are the coefficients for remaining covariates. The regression coefficients can be interpreted as the difference in the log-odds of the expected counts of delinquent activities for a one-unit increase in the predictor variables. The exponentiation of this coefficient converts to the incident rate ratio (IRR), or the expected change in the count of delinquent activities for a one-unit increase in the independent variable. For sake of clarity, results are discussed in terms of the IRR.

Admittedly, the general offending scale may be problematic as minor offenses are overrepresented. Therefore, two sensitivity analyses were performed to demonstrate the validity of such a scale. The first analysis applied logistic regression to examine the association between exposure to each type of behavior and the likelihood that a respondent reported engaging in that specific behavior. The second analysis separated the offending variety scale into violent and nonviolent behaviors and used negative binomial regression models to assess the associations between exposure to violent/nonviolent behaviors online and a respondent’s own reports of violent and nonviolent offending.

Results

Table 3 provides a zero-order tetrachoric correlation matrix for the various types of self-reported criminal behaviors and specific behaviors respondents reported witnessing in their online networks. There are strong correlations between exposure to criminal behavior online and self-reported behavior, indicating a strong association between online exposure and off-line offending. Significant associations across the multiple types of behavior indicate that exposure to or participation in one type of behavior increases the risk of exposure or participation in other behaviors. These results indicate that exposure to specific forms of offending in social networks is associated with increases in self-reported offending more generally.

Table 3.

Correlation Matrix.

	SNS Partner Abuse	SNS Weapon	SNS Fight	SNS Sell Drugs	SNS DUI	SNS Fire	SNS Steal	SNS Vandalism	Self-Partner Abuse	Self-Weapon	Self-Fight	Self-Sell Drugs	Self-DUI	Self-Fire	Self-Theft	Self-Vandalism
SNS partner abuse	1.000	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—
SNS weapon	0.588*	1.000	—	—	—	—	—	—	—	—	—	—	—	—	—	—
SNS fight	0.698*	0.732*	1.000	—	—	—	—	—	—	—	—	—	—	—	—	—
SNS sell drugs	0.729*	0.784*	0.735*	1.000	—	—	—	—	—	—	—	—	—	—	—	—
SNS DUI	0.507*	0.490*	0.639*	0.628*	1.000*	—	—	—	—	—	—	—	—	—	—	—
SNS fire	0.712*	0.554*	0.461*	0.497*	0.279*	1.000	—	—	—	—	—	—	—	—	—	—
SNS steal	0.524*	0.548*	0.564*	0.559*	0.473*	0.523*	1.000	—	—	—	—	—	—	—	—	—
SNS vandalism	0.562*	0.607*	0.703*	0.642*	0.586*	0.649*	0.690*	1.000	—	—	—	—	—	—	—	—
Self-partner abuse	0.210*	0.332*	0.306*	0.395*	0.290*	0.338*	0.118	0.215*	1.000	—	—	—	—	—	—	—
Self-weapon	0.210	0.617*	0.434*	0.408*	0.239*	0.145	0.353*	0.297*	0.345*	1.000	—	—	—	—	—	—
Self-fight	0.378*	0.357*	0.573*	0.330*	0.312*	0.347*	0.327*	0.347*	0.338*	0.550*	1.000	—	—	—	—	—
Self-sell drugs	0.124	0.500*	0.319*	0.563*	0.250*	0.077	0.286*	0.202	0.238	0.646*	0.446*	1.000	—	—	—	—
Self-DUI	0.100	0.232*	0.207*	0.298*	0.563*	0.013	0.125	0.176*	0.272*	0.367*	0.431*	0.641*	1.000	—	—	—
Self-fire	0.056	0.028	0.055	0.068	−0.112	0.618*	−0.247	0.087	0.493*	0.438*	0.404*	0.409	0.081	1.000	—	—
Self-theft	0.148	0.151	0.120*	0.073	0.101	0.243*	0.367*	0.157	0.233*	0.423*	0.354*	0.437*	0.335*	0.435*	1.000	—
Self-vandalism	0.140	0.141	0.286*	0.283*	0.402*	0.374*	0.476*	0.473*	0.215	0.363*	0.384*	0.311*	0.357*	0.304	0.582*	1.000

Note: SNS = social networking websites; DUI = driving under the influence.

*p < .05.

Table 4 presents the results of the multivariate analyses. Model 1 presents the baseline bivariate negative binomial regression of exposure to criminal behavior within online social networks and self-reported offending. Consistent with expectation, the results of this model demonstrate a direct, statistically significant association between exposure to criminal behavior in SNS and self-reported offending. Each additional criminal behavior a respondent was exposed to online was associated with a 23.7% ( $e^{0 .213} - 1$ ) increase in the expected count of personal offending.

Table 4.

Negative Binomial Regression of Exposure to Criminal Behavior, SNS Access, and Demographic Controls on Self-Reported Offending.

	Model 1		Model 2		Model 3
	Baseline		Controls		Online Behavior
	β	SE	β	SE	β	SE
Intercept	−.697	.103***	−.553	.785	−.553	.807
Age	—	—	.035	.030	.042	.030
Female	—	—	−.563	.109***	−.569	.111***
Black	—	—	.065	.125	.101	.125
Other race	—	—	−.059	.170	−.010	.170
GPA	—	—	−.175	.120	−.161	.121
Minutes on SNS	—	—	—	—	−.018	.048
Number of friends on SNS	—	—	—	—	.150	.061*
Hesitant to post due to job	—	—	—	—	−.300	.118*
Parental access to SNS profile	—	—	—	—	−.143	.110
Exposure to offending on SNS	.213	.024***	.206	.024***	.187	.025***
Log likelihood	−683.13		−630.75***		−553.62***
χ²(df)	—		104.76 (5)		154.26 (4)

Note. N = 517. GPA = grade point average; SE = standard error; SNS: social networking websites; GPA: grade point average.

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

The second model (Model 2) introduces demographic controls including age, gender, race/ethnicity, and GPA. Results from this model suggest that levels of offending among females were roughly lower than males by 43% ( $e^{- 0.563} - 1$ ). None of the other control variables were significantly associated with self-reported offending. Importantly, the coefficient for exposure to offending through SNS remained significant after the inclusion of the control variables, which indicates that the association between exposure to offending online and self-reported behavior cannot be attributed to factors like age, race/ethnicity, or gender. Notably, the inclusion of the sociodemographic control variables provided a significantly better fit to the data, suggesting sociodemographic differences explain variation in self-reported offending above and beyond exposure to offending online (χ² = 104.76; df = 5).⁶

Model 3 introduces the online behavior controls including the number of minutes spent online, number of friends, hesitancy to post personal information due to job, and parental access to SNS profile to Model 2. As with the previous model, the coefficient for gender was significant and negative, suggesting females reported significantly lower levels of offending than males. Hesitancy to post information emerged as a significant protective factor against self-reported offending, as those who were hesitant to post on SNS due to their job or future potential employment had an expected count of offending 25.9% lower than those who were not hesitant to post ( $e^{- 0.300} - 1$ ). This may reflect the tendency among individuals who are invested in their future to refrain from any activities that may jeopardize their employment prospects. Finally, the size of the online social network emerged a significant risk factor for self-reported offending such that the larger the network, the higher the level of self-reported offending among respondents. A one-unit increase in the amount of friends one has on their online network was associated with a 16% increase in the expected count of personal offending ( $e^{0.150} - 1$ ).

Notably, the coefficient for exposure to online behavior remained relatively unchanged in the full model. Each additional criminal behavior a respondent was exposed to was associated with a 20.6% increase in the expected count of personal offending ( $e^{0.187} - 1$ ). This demonstrates that the observed relationship cannot be attributed to factors associated with differences in SNS usage. Interestingly, these findings fail to comport with those of Meldrum and Clarke (2013), as there was no significant association between time spent online and self-reported offending. The inclusion of the network usage variables provided a significantly better fit to the data than Model 2, suggesting that network usage differences explain variation in self-reported offending above and beyond sociodemographic differences (χ² = 154.26; df = 4).

Results from the two sensitivity analyses are consistent with findings using the general offending scale. The first analysis, which uses logistic regression to examine each specific type of behavior, indicates that each individual behavior to which a respondent is exposed is associated with self-reports of that behavior (see Table 5). The second analysis separates the behaviors into violent and nonviolent models (see Table 6). While the control variables exert varying influence on self-reported offending in both models, the results from this analysis demonstrate that exposure to offending online is a significant risk factor for both violent and nonviolent offending.

Table 5.

Item Specific Logistic Regression of Exposure to Criminal Behavior, SNS Access, and Demographic Controls on Self-Reported Offending.

	Model 1 Abusing Partner (N = 517)		Model 2 Carrying Weapon (N =518)		Model 3 Fighting (N = 518)		Model 4 Selling Drugs (N = 517)		Model 5 DUI (N = 517)		Model 6 Setting Fire (N = 517)		Model 7 Stealing (N = 517)		Model 8 Vandalism (N = 517)
	β	SE	β	SE	β	SE	β	SE	β	SE	β	SE	β	SE	β	SE
Intercept	2.055	3.301	−3.134	2.673	−5.115	2.408*	−3.06	3.203	−2.591	1.476	−8.024	6.475	−1.471	1.827	−2.113	2.507
Age	−.236	.134	.042	.097	−.002	.084	.097	.116	.147	.055**	.051	.236	.037	.070	.030	.095
Female	.916	.444*	−1.083	.336***	−.648	.300*	−1.430	.447***	−.536	.218*	−1.692	.960	−1.078	.256***	−1.608	.360***
Black	1.328	.481**	.124	.378	.951	.353**	.157	.474	−.324	.244	2.299	1.231	−.393	.301	.002	.383
Other Race	−.296	.833	−.282	.525	.591	.453	−.577	.829	−.137	.332	1.395	1.522	.255	.357	−.335	.555
GPA	−.513	.392	−.161	.394	.339	.356	−.685	.507	−.401	.231	.520	.956	−.131	.265	−.263	.384
Minutes on SNS	.152	.172	.004	.147	.114	.131	−.329	.174*	−.043	.100	−.270	.339	.015	.111	−.165	.147
Number of friends on SNS	.187	.248	.200	.197	.177	.174	−.064	.233	.325	.117**	.010	.477	.156	.135	.229	.183
Hesitant to post due to job	−.064	.394	−.285	.346	−.117	.307	−.546	.467	−.708	.228**	−.177	.937	−.601	.278*	−.477	.366
Parental access to SNS profile	−.043	.370	−.415	.332	−.140	.295	−.714	.433	−.271	.214	−1.440	.982	−.046	.257	.197	.332
Exposure to abusing partner	1.625	.398***	–	–	—	—	—	—	—	—	—	—	—	—	—	—
Exposure to carrying weapon	—	—	2.360	.412***	—	—	—	—	—	—	—	—	—	—	—	—
Exposure to fighting	—	—	—	—	2.292	.605***	—	—	—	—	—	—	—	—	—	—
Exposure to selling drugs	—	—	—	—	—	—	2.431	.548***	—	—	—	—	—	—	—	—
Exposure to DUI	—	—	—	—	—	—	—	—	1.662	.227***	—	—	—	—	—	—
Exposure to setting fire	—	—	—	—	—	—	—	—	—	—	3.027	.871***	—	—	—	—
Exposure to stealing	—	—	—	—	—	—	—	—	—	—	—	—	.970	.254***	—	—
Exposure to vandalism	—	—	—	—	—	—	—	—	—	—	—	—	—	—	1.656	.352***

Note: SNS: social networking websites; GPA: grade point average; DUI: driving under the influence.

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

Table 6.

Negative Binomial Regression of Exposure to Violent and Nonviolent Criminal Behavior, SNS Access, and Demographic Controls on Self-Reported Offending.

	Model 1		Model 2
	Violent Behaviors		Nonviolent Behaviors
	β	SE	β	SE
Intercept	−1.260	1.442	−.960	.780
Age	−.034	.053	.063	.029*
Female	−.322	.179	−.631	.111***
Black	.482	.212*	−.111	.128
Other race	.027	.296	−.053	.168
GPA	−.102	.203	−.195	.118
Minutes on SNS	.041	.079	−.041	.048
Number of friends on SNS	.101	.109	.147	.066*
Hesitant to post due to job	−.074	.184	−.412	.124***
Parental access to SNS profile	−.170	.174	−.160	.111
Exposure to violent offending	.696	.099***	—	—
Exposure to nonviolent offending	—	—	.227	.035***

Note: N = 525. SE = standard error; SNS = social networking websites; GPA = grade point average.

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

Discussion

SNS have had a profound impact on interpersonal communication over the course of the last decade. Recent evidence suggests an increasing number of people are interacting online and they are doing it with greater and greater frequency (Lenhart et al., 2010; Smith & Brenner, 2012). SNS transcend traditional geographic boundaries, allowing users to connect with a broader range of potential friends while maintaining ties with old acquaintances. Emerging criminological research suggests online interaction has the potential to influence behavior; however, relatively little research has examined the association between exposure to offending online and self-reported offending behavior.

The forgoing analyses took a first step at bridging this gap in the literature by examining whether the behavioral concurrency observed among individuals within traditional social networks is also characteristic of online social networks. Consistent with expectation, the results demonstrate a strong, positive association between self-reported offending and exposure to criminal behavior in one’s SNS. Thus, the processes underlying traditional social interaction also seem to characterize online interaction.

Two unique attributes of online social networking also emerged as significant predictors of offending in the multivariate models: network size and hesitancy to post information due to a job or future employment. Respondents embedded within larger online networks engaged in greater levels of offending. This may indicate a greater potential for users to be exposed to deviant messages than individuals with smaller networks. While social learning theory would suggest that learning takes place in intimate peer groups, this finding may be consistent with the logic of the strength of weak ties, whereby social reinforcement may take place more strongly in larger social networks, characterized by less personal relationships (Granovetter, 1973). This finding is consistent with other research on SNS-related behaviors. For example, Donath and Boyd (2004) hypothesized participation in online networks would indeed increase exposure to weak ties due to cheap and easy access to a larger peer group. Unfortunately, this analysis is limited in its ability to examine network size since data are absent on the number of peers posting about specific behaviors and how often they post about such behaviors. Given that weak ties allow for exposure to information and resources not found in one’s immediate environment (de Zúñiga &Valenzuala, 2011), future studies would benefit from examining the effects of online network size on personal offending.

On the other hand, those who were hesitant to post personal information due to employment concerns reported significantly lower levels of offending. This might indicate individuals who are more invested in their futures offend at lower rates on the whole. Consistent with classic control theory (Hirschi, 1969), respondents may be less likely to engage in criminal behavior if they are committed to their future since they may believe they have something to lose by engaging in such behavior. Employers are increasingly turning to social media during the hiring process of potential applicants (Joos, 2008), therefore individuals could be cognizant of the image they are portraying on SNS and may be less likely to post information that could jeopardize the attainment of a job. Thus, both offending and online behavior are influenced by future goals.

Parental access was not a significant predictor of self-reported offending in the current analysis, which may suggest that parental influence matters less among college students. Classic control theory posits that individuals who are attached to their parents would be less likely to engage in criminal behavior. This attachment may be less influential in a sample that has reached a stage of independence by attending college. Additionally, the dependent variable measures off-line self-reported behavior and does not examine respondents’ own behavior on SNS.

Prior research in this area revealed that frequency of virtual socialization was positively associated with self-reported delinquency (Meldrum & Clarke, 2013); however, the present analyses were unable to replicate this finding. Meldrum and Clarke applied Osgood and colleagues’ (1996) concept of unstructured socialization to explain the link between virtual socialization and delinquency. The inability to replicate this finding in this study may be attributable to the age of the analytic sample relative to the sample used by Meldrum and Clarke (2013). Given the focus on college students in this study, unstructured socialization online should matter less, as most socialization at this age occurs away from the watchful eyes of parents and strangers. Although previous research suggests individuals who engage in street crime are active in SNS (Moule, Pyrooz, & Decker, 2013), it seems that those who offend at the highest levels will have less time to interact with peers in an online setting. Additionally, Rogers (2010) characterizes offenders as lacking empathy, open to new ideas, and self-efficient. Moule, Pyrooz, and Decker (2013) point out this may limit the amount of time and the willingness of offenders to use SNS. It is possible the relationship between time spent socializing with friends online may not be as strong as previously hypothesized. In fact, recent evidence suggests virtual time spent with peers has a weaker relationship with delinquency than public unsupervised socializing (Weerman, Bernasco, Bruinsma, & Pauwels, 2013).

The findings of this study build on the small body of research examining online interaction and offending. It moves beyond prior work by not only considering time spent online but also identifying the types of behavior to which users are exposed. Net of the effect of demographic characteristics, frequency of usage, size of the network, and factors related to hesitancy to post information, there was a strong, positive association between exposure to offending in SNS and self-reported offending. Unfortunately, the empirical models presented here do not allow for the disentanglement of the two competing processes—social learning and peer group selection. Although we cannot speak much beyond these data, peer network research in criminology suggests peer effects reflect some combination of learning and selection (Matsueda & Anderson, 1998). We may reasonably surmise a similar process is reflected here, such that individuals may select friends based on similar characteristics and the behaviors they are exposed to through these networks may serve to reinforce future offending. Rather than serve as a critical limitation, this demonstrates a clear need for future research to unpack these competing underlying processes. A longitudinal study examining criminal behavior and SNS usage patterns will be a crucial step in identifying a causal relationship by establishing temporal order.

This study relied on a self-reported measure of peer offending through the posts a respondent is able to view on their SNS profile. It has been noted that using respondents’ reports of peer behavior may be influenced by assumed similarity or projection (Byrne & Blaylock, 1963). Scholars sometimes caution against the use of such measures since the reported peer behavior could possibly be an extension of the respondent’s own behavior (Gottfredson & Hirschi, 1990; Kandel, 1996). While this may be the case, the current analysis does not explicitly consider whether peers commit crimes, rather respondents report about posts that discuss or display various criminal activities. This measure may better capture favorable definitions or positive reinforcements of such behaviors compared to directly asking what behaviors are engaged in by peers.⁷ With this in mind, replicating this study with a direct measure of online peer behavior may further strengthen the results. Additionally, future research should control for off-line peer offending in order to disentangle the effects of off-line versus online peer behavior. It is possible that behaviors discussed or displayed on SNS reflect the same behaviors one would be exposed to through traditional interaction, thereby limiting the influence of SNS involvement.

The use of a college-based sample, while suitable for the purposes of this research, has inherent limitations. The sample may not be representative of the larger population, which limits the external validity of the study. However, empirical criminological research has extensively utilized similar college-based samples, and studies have shown that very few statistically significant differences exist in these samples compared to the general population (Wiecko, 2010). Future studies can expand SNS research into additional age-groups, which will be imperative as younger individuals increase their use of virtual networks.

Future studies should also examine different types of SNS in order to consider the various networks and communities that exist within these sites. Some, like Facebook, are characterized as being both open to a large network of friends yet private in terms of personal messaging. Other sites, like Reddit, often have larger communities that are relatively more open in nature.⁸ Some sites have a specific focus on pictures (Instagram), videos (YouTube), or short messages with limited text (Twitter). These differing characteristics may influence the type of criminal activity that is discussed or displayed. While the current analysis is unable to differentiate between networks,⁹ future research should consider the influence of specific SNS.

Given the precipitous increase in SNS usage over the course of the last decade, it seems virtual social networks will continue to shape interpersonal interaction for the foreseeable future. Understanding how and why such interaction influences behavior is paramount for extending criminological theory and research. Studies have consistently found that interaction with delinquent and criminal peers is associated with personal offending. This study has replicated this finding in a new environment yet to be fully explored by scholars. SNS has fundamentally changed the mechanisms of socialization, yet many questions surrounding the effects of these new virtual networks on criminal behavior remain. Interaction with peer groups takes place at any hour of the day, at any location, and with a limitless amount of friends. Although this study suggests an association between self-reported offending and exposure to criminal behavior in one’s SNS, it is possible these websites may have additional effects on criminal behavior. Cyber-socialization may uncover new explanations of criminal behavior and ultimately how peers influence one another.

Footnotes

Acknowledgments

Data for these analyses were collected as part of the first author’s master thesis research at the University at Memphis. The authors are indebted to Margaret Vandiver, K.B. Turner, Burt Burraston, T. J. Taylor, and Mari Katherine Webb for their invaluable feedback on earlier drafts of this article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Acar

(2008). Antecedents and consequences of online social networking behavior: The case of Facebook. Journal of Website Promotion, 3, 62–83.

Acquisti

Gross

(2006). Imagined communities: Awareness, information sharing, and privacy on the Facebook. In Wang

Kobsa

(Eds.), Privacy enhancing technologies (pp. 36–58). Berlin, Heidelberg, Germany: Springer.

Akers

(1998). Social learning and social structure: A general theory of crime and deviance. Boston, MA: Northeastern University Press.

Akers

Jensen

(2006). The empirical status of social learning theory of crime and deviance: The past, present, and future. Taking Stock: The Status of Criminological Theory, 15, 37–76.

Beauchamp

Olson

(1973). Corrections for bias in regression estimates after logarithmic transformation. Ecology, 54, 1403–1407.

Boyd

Ellison

(2007). Social network sites: Definition, history, and scholarship. Journal of Computer-Mediated Communication, 13, 210–230.

Bowman-Grieve

(2009). Exploring stormfront: A virtual community of the radical right. Studies in Conflict and Terrorism, 32, 989–1007.

Bye

Pushkar

Conway

(2007). Motivation, interest, and positive affect in traditional and nontraditional undergraduate students. Adult Education Quarterly, 57, 141–158.

Byrne

Blaylock

(1963). Similarity and assumed similarity of attitudes between husbands and wives. The Journal of Abnormal and Social Psychology, 67, 636.

10.

Centola

(2010). The spread of behavior in an online social network experiment. Science, 329, 1194–1197.

11.

Dey

Jelveh

Ross

(2012). Facebook users have become much more private: A large-scale study. In Pervasive Computing and Communications Workshops (PERCOM Workshops), 2012 IEEE International Conference (pp. 346–352). Lugano, Switzerland. Retrieved October 15, 2014 from http://cis.poly.edu/∼ratan/facebookusertrends.pdf

12.

de Zúñiga

H. G.

Valenzuela

(2011). The mediating path to a stronger citizenship: Online and offline networks, weak ties, and civic engagement. Communication Research, 38, 397–421.

13.

Donath

Boyd

(2004). Public displays of connection. BT Technology Journal, 22, 71–82.

14.

Ellison

Steinfield

Lampe

(2007). The benefits of Facebook “friends:” Social capital and college students’ use of online social network sites. Journal of Computer-Mediated Communication, 12, 1143–1168

15.

Facebook. (2009). Annual report 2008. Retrieved July 25, 2013, from http://investor.fb.com/annuals.cfm

16.

Facebook. (2013). Annual report 2012. Retrieved July 25, 2013, from http://investor.fb.com/annuals.cfm

17.

Furnell

(2002). Cybercrime: Vandalizing the information society. Boston, MA: Addison-Wesley.

18.

Glueck

(1950). Unraveling juvenile delinquency. New York, NY: Commonwealth Fund.

19.

Gottfredson

Hirschi

(1990). A general theory of crime. Stanford, CA: Stanford University Press.

20.

Granovetter

(1973). The strength of weak ties. American Journal of Sociology, 78, l.

21.

Hawdon

(2012). Applying differential association theory to online hate groups: A theoretical statement. Research on Finnish Society, 5, 39–47.

22.

Higgins

G. E.

Marcum

(2011). Digital piracy: An integrated theoretical approach. Raleigh, NC: Carolina Academic Press.

23.

Higgins

G. E.

Marcum

Freiburger

T. L.

Ricketts

M. L.

(2012). Examining the role of peer influence and self-control on downloading behavior. Deviant Behavior, 33, 412–423.

24.

Hinduja

(2003). Trends and patterns among online software pirates. Ethics and Information Technology, 5, 49–61.

25.

Hirschi

(1969). Causes of delinquency. Berkeley, CA: University of California Press.

26.

Holt

(Ed.). (2013). Crime on-line: Correlates, causes, and context. Durham, NC: Carolina Academic Press.

27.

Holt

Bossler

(2014). An assessment of the current state of cybercrime scholarship. Deviant Behavior, 35, 20–40.

28.

Holt

Bossler

May

(2012). Low self-control, deviant peer associations, and juvenile cyberdeviance. American Journal of Criminal Justice, 37, 378–395.

29.

Holt

T. J.

Copes

(2010). Transferring subcultural knowledge on-line: Practices and beliefs of persistent digital pirates. Deviant Behavior, 31, 625–654.

30.

Holt

T. J.

Morris

R. G.

(2009). An exploration of the relationship between MP3 player ownership and digital piracy. Criminal Justice Studies, 22, 381–392.

31.

Ingram

J. R.

Hinduja

(2008). Neutralizing music piracy: An empirical examination. Deviant Behavior, 29, 334–366.

32.

International Association of Chiefs of Police. (2013). IACP Center for Social Media Survey 2013. IACP Center for Social Media. Retrieved September 13, 2013, from http://www.iacpsocialmedia.org/

33.

Joinson

(2008). Looking at, looking up or keeping up with people? Motives and uses of Facebook. In Proceedings of the twenty-sixth annual SIGCHI conference on human factors in computing systems (pp. 1027–1036). New York, NY: ACM.

34.

Joos

J. G.

(2008). Social media: New frontiers in hiring and recruiting. Employment Relations Today, 35, 51–59.

35.

Kandel

(1996). The parental and peer contexts of adolescent deviance: An algebra of interpersonal influences. Journal of Drug Issues, 26, 289–315.

36.

King

Walpole

Lamon

(2007). Surf and turf wars online—Growing implications of internet gang violence. Journal of Adolescent Health, 41, S66–S68.

37.

Lefebvre

Bornkessel

(2013). Digital social networks and health. Circulation, 127, 1829–1836.

38.

Lehdonvirta

Räsänen

(2011). How do young people identify with online and offline peer groups? A comparison between UK, Spain and Japan. Journal of Youth Studies, 14, 91–108.

39.

Lenhart

Madden

(2007). Social networking websites and teens: An overview. Washington, DC: Pew Internet and American Life Project.

40.

Lenhart

Purcell

Smith

Zickuhr

(2010). Social media and mobile internet use among teens and young adults. Washington, DC: Pew Internet and American Life Project.

41.

Lewis

Gonzalez

Kaufman

(2011). Social selection and peer influence in an online social network. Proceedings of the National Academy of Sciences, 109, 68–72.

42.

Long

Freese

(2006). Regression models for categorical dependent variables using Stata. College Station, TX: Stata Press books.

43.

Luxton

June

Fairall

(2012). Social media and suicide: A public health perspective. American Journal of Public Health, 102, S195–S200.

44.

Marcum

(2010). Examining cyberstalking and bullying: Causes, context, and control. In Holt

T. J.

(Ed.), Crime on-line: Correlates, causes, and context (pp. 175–192). Raleigh, NC: Carolina Academic Press.

45.

Matsueda

Anderson

(1998). The dynamics of delinquent peers and delinquent behavior. Criminology, 36, 269–308.

46.

Mazerolle

Piquero

(1998). Linking exposure to strain with anger: An investigation of deviant adaptations. Journal of Criminal Justice, 26, 195–211.

47.

McDonald

Hortsmann

Strom

Pope

(2009). The impact of the internet on deviant behavior and deviant communities. Washington, DC: Institute for Homeland Security Solutions. Retrieved October 15, 2014 from http://sites.duke.edu/ihss/files/2011/12/IRW-Literature-Reviews-Deviance-and-the-Internet.pdf

48.

McKenna

Bargh

(1998). Coming out in the age of the internet: Identity demarginalization through virtual group participation. Journal of Personality and Social Psychology, 75, 681–694.

49.

Meldrum

Clark

(2013). Adolescent virtual time spent socializing with peers, substance use, and delinquency. Crime & Delinquency. doi:10.1177/0011128713492499

50.

Moreno

Jelenchick

Egan

Cox

Young

Gannon

Becker

(2011). Feeling bad on Facebook: Depression disclosures by college students on a social networking site. Depression and Anxiety, 28, 447–455.

51.

Morgan

Snelson

Elison-Bowers

(2010). Image and video disclosure of substance use on social media websites. Computers in Human Behavior, 26, 1405–1411.

52.

Morris

R. G.

Higgins

G. E.

(2009). Neutralizing potential and self-reported digital piracy: A multitheoretical exploration among college undergraduates. Criminal Justice Review, 34, 173–195.

53.

Moule

Pyrooz

Decker

(2013). From “what the f#@% is a facebook?” to “who doesnt use facebook?”: The role of criminal lifestyles in the adoption and use of the internet. Social Science Research, 42, 1411–1421.

54.

Nagin

D. S.

Paternoster

(1993). Enduring individual differences and rational choice theories of crime. Law and Society Review, 27, 467–496.

55.

Nardi

P. M.

(2006). Doing survey research: A guide to quantitative methods. Boston, MA: Allyn & Bacon.

56.

Newman

Clarke

(2003). Superhighway robbery. Portland, OR: Taylor & Francis.

57.

Osgood

Anderson

(2004). Unstructured socialization and rates of delinquency. Criminology, 42, 519–550.

58.

Osgood

Finken

McMorris

(2002). Analyzing multiple-item measures of crime and deviance II: Tobit regression analysis of transformed scores. Journal of Quantitative Criminology, 18, 319–347.

59.

Osgood

Wilson

O’Malley

Bachman

Johnston

(1996). Routine activities and individual deviant behavior. American Sociological Review, 61, 635–655.

60.

Patton

Eschmann

Butler

(2013). Internet banging: New trends in social media, gang violence, masculinity and hip hop. Computers in Human Behavior, 29, A54–A59.

61.

Pempek

Yermolayeva

Calvert

(2009). College students’ social networking experiences on Facebook. Journal of Applied Developmental Psychology, 30, 227–238.

62.

Pratt

Cullen

Sellers

Thomas Winfree

Madensen

Daigle

… Gau

. (2010). The empirical status of social learning theory: A meta-analysis. Justice Quarterly, 27, 765–802.

63.

Pyrooz

Decker

Moule

(2013). Criminal and routine activities in online settings: Gangs, offenders, and the internet. Justice Quarterly, 30, 1–29.

64.

Reyns

B. W.

Henson

Fisher

B. S.

(2011). Being pursued online: Applying cyberlifestyle–routine activities theory to cyberstalking victimization. Criminal Justice and Behavior, 38, 1149–1169.

65.

Rideout

Foehr

Roberts

(2010). Generation M² . Media in the Lives of 8-to 18-year-olds. Retrieved July 9, 2013, from http://myweb.wwu.edu/karlberg/444/readings/GenM2.pdf

66.

Rogers

(2010). Diffusion of innovations. New York, NY: Simon and Schuster.

67.

Royston

(2005). Multiple imputation of missing values: Update of ICE. The Stata Journal, 5, 527–536.

68.

Simons-Morton

Chen

(2006). Over time relationships between early adolescent and peer substance use. Addictive Behaviors, 31, 1211–1223.

69.

Smith

Brenner

(2012). Twitter use 2012. Pew Internet & American Life Project. Retrieved June 28, 2013, from http://alexa.pewinternet.com/˜/media/Files/Reports/2012/PIP_Twitter_Use_2012.pdf

70.

Strayer

Wareing

Rushton

(1979). Social constrains on naturally occurring preschool altruism. Ethology and Sociobiology, 1, 3–11.

71.

Subrahmanyam

Reich

Waechter

Espinoza

(2008). Online and offline social networks: Use of social networking sites by emerging adults. Journal of Applied Developmental Psychology, 29, 420–433.

72.

Sutherland

Cressey

(1974). Criminology (9th ed.). New York, NY: J. B. Lippincott.

73.

Sweeten

(2012). Scaling criminal offending. Journal of Quantitative Criminology, 28, 533–557.

74.

Tibbetts

S. G.

Herz

D. C.

(1996). Gender differences in factors of social control and rational choice. Deviant Behavior, 17, 183–208.

75.

Tourangeau

Rips

L. J.

Rasinski

(2000). The psychology of survey response. Cambridge, England: Cambridge University Press.

76.

Urberg

Luo

Pilgrim

Degirmencioglu

(2003). A two-stage model of peer influence in adolescent substance use: Individual and relationship-specific differences in susceptibility to influence. Addictive Behaviors, 28, 1243–1256.

77.

U.S. Census Bureau. (2013). U.S. and world population clock. Retrieved July 25, 2013, from http://www.census.gov/popclock/

78.

Wall

(2001). Cybercrimes and the internet. In Wall

D. S.

(Ed.), Crime and the internet (pp. 1–17). New York, NY: Routledge.

79.

Warr

(2002). Companions in crime: The social aspects of criminal conduct. Cambridge, MA: Cambridge University Press.

80.

Weerman

Bernasco

Bruinsma

Pauwels

(2013). When is spending time with peers related to delinquency? The importance of where, what, and with whom. Crime & Delinquency. doi:10.1177/0011128713478129

81.

Wiecko

F. M.

(2010). Research Note: Assessing the validity of college samples: Are students really that different? Journal of Criminal Justice, 38, 1186–1190.