Вы находитесь на странице: 1из 20

810943

research-article2018
YASXXX10.1177/0044118X18810943Youth & SocietyWood and Graham

Article
Youth & Society
1­–20
“Safe” and “At- © The Author(s) 2018
Article reuse guidelines:
Risk”: Cyberbullying sagepub.com/journals-permissions
DOI: 10.1177/0044118X18810943
https://doi.org/10.1177/0044118X18810943
Victimization and journals.sagepub.com/home/yas

Deviant Health Risk


Behaviors in Youth

Frank R. Wood Jr.1


and Roderick Graham1

Abstract
This study explores the links between cyberbullying victimization and a
set of health risk behaviors associated with juvenile delinquency (cigarette
smoking, marijuana usage, alcohol usage, and sexual frequency). These links
are examined with data from the 2015 Youth Risk Behavior Survey (n =
9,122). Using cluster analysis, respondents are categorized into two groups:
“safe” students who report on average no engagement in the behaviors
measured, and “at-risk” students who report on average moderate to high
levels of engagement in sexual frequency, marijuana usage, and alcohol
usage. Findings suggest that cyberbullying victimization increases the odds
of a student being categorized into the “at-risk” cluster. This effect holds
controlling for physical bullying, a proxy measure of self-control, and
demographic variables.

Keywords
cyberbullying, health risk behaviors, general strain theory, juvenile
delinquency

1Old Dominion University, Norfolk, VA, USA

Corresponding Author:
Roderick Graham, Assistant Professor, Department of Sociology & Criminal Justice, Old
Dominion University, Norfolk, VA 23529, USA.
Email: rgraham@odu.edu
2 Youth & Society 00(0)

Introduction
Cyberbullying can be defined as “any behavior performed through elec-
tronic or digital media by individuals or groups that repeatedly communi-
cates hostile or aggressive messages intended to inflict harm or discomfort
on others” (Tokunaga, 2010, p. 278). Examples of cyberbullying include
posting insults on a victim’s social media page, sending hateful text mes-
sages to a victim, or sharing via the Internet embarrassing or degrading
images of a victim. In this study, the focus is on students who report being a
victim of cyberbullying.
Although cyberbullying victimization is a relatively new phenomenon,
scholars have identified several adverse consequences. Scholars have
linked cyberbullying victimization to school avoidance (Payne & Hutzell,
2015), low self-esteem (Palermiti, Servidio, Bartolo, & Costabile, 2017;
Patchin & Hinduja, 2010), feelings of fear and powerlessness (Hoff &
Mitchell, 2009), aggressive behaviors toward others (Calvete, Orue,
Estévez, Villardón, & Padilla, 2010), juvenile delinquency (Hay, Meldrum,
& Mann, 2010), bringing a weapon to school (Keith, 2017), and adverse
psychological outcomes (Aboujaoude, Savage, Starcevic, & Salame, 2015;
Kowalski & Limber, 2013).
The current research will add to this body of literature by linking cyber-
bullying victimization to a specific set of health risk behaviors: behaviors that
are associated with juvenile delinquency. Using nationally representative
data from the Youth Risk Behavior Survey (YRBS) conducted in 2015, the
effect of cyberbullying victimization is examined for cigarette smoking, alco-
hol and marijuana use, and sexual frequency. Although the YRBS monitors a
wide range of health behaviors (e.g., amount of fruit intake), these four are
selected because they have been historically and theoretically linked with
deviance and future criminal acts. The research question asked is

Research Question 1: Does being a victim of cyberbullying increase the


likelihood of health risk behaviors associated with juvenile delinquency?

Literature Review
Defining Cyberbullying
Ybarra, Boyd, Korchmaros, and Oppenheim (2012) pointed out that defini-
tions of cyberbullying vary widely within the literature. However, there are a
few common themes that most definitions share. Olweus (1993) defined bul-
lying as when a person is exposed, repeatedly and over time, to negative
Wood and Graham 3

actions on the part of one or more other persons, and he or she has difficulty
defending himself or herself. This definition is common in the bullying litera-
ture and has been imported into the literature on cyberbullying. Accordingly,
Hinduja and Patchin (2010) defined cyberbullying as “willful and repeated
harm inflicted through computers, cell phones, and other electronic devices”
(p. 1). Similarly, Tokunaga (2010) defined cyberbullying as “any behavior
performed through electronic or digital media by individuals or groups that
repeatedly communicates hostile or aggressive messages intended to inflict
harm or discomfort on others” (p. 278).

Research on Cyberbullying Victimization


The research on cyberbullying victimization can be roughly divided into two
domains. First, scholars have explored the psychological impacts of cyber-
bullying. In a recent review of these impacts, Foody, Samara, and Carlbring
(2015) argued that psychological and emotional impacts represent the largest
problem in cyberbullying victimization. Scholars have identified positive
relationships between cyberbullying and school avoidance (Payne & Hutzell,
2015), low self-esteem (Palermiti et al., 2017; Patchin & Hinduja, 2010),
antisocial emotions such as fear and powerlessness (Hoff & Mitchell, 2009),
and higher levels of stress and anxiety (Fredstrom, Adams, & Gilman, 2011).
There have been several high-profile cases of young people committing sui-
cide because of being cyberbullied. Because of the nature of these cases, they
have necessarily commanded a large amount of attention in the media.
Scholarly work has demonstrated that this concern has an empirical founda-
tion. A recent review of relevant studies shows that cyberbullying is consis-
tently linked to various psychopathologies, including suicidal ideation
(Aboujaoude et al., 2015). The second dimension of cyberbullying research
explores the effect of cyberbullying victimization on subsequent acts of crime
and deviance. This includes violent and aggressive behaviors toward others
(Calvete et al., 2010), the tendency to cyberbully others (Chapin & Coleman,
2017), juvenile delinquency (Hay et al., 2010), and bringing a weapon to
school (Keith, 2017). This study will add to this second dimension.

Cyberbullying and Health Risk Behaviors


This study explores the relationship between cyberbullying and health risk
behaviors. The Centers for Disease Control and Prevention (CDC) has identi-
fied and measured a wide variety of health risk behaviors. These behaviors
“contribute to the leading causes of death and disability among youth and
adults” (CDC, 2018). This includes irregular regular consumption of fruit and
4 Youth & Society 00(0)

vegetables, not eating breakfast, physical inactivity, smoking, alcohol use,


marijuana use, unprotected sex, and sex with multiple partners. A large body
of research has associated these health risk behaviors with, among other fac-
tors, low income and low self-control (de Winter, Visser, Verhulst, Vollebergh,
& Reijneveld, 2016), immigration status (Jones, Pezzi, Rodriguez-Lainz, &
Whittle, 2016), and adverse childhood experiences such as childhood abuse
and family dysfunction (Downey, Gudmunson, Pang, & Lee, 2017; Hertz,
Jones, Barrios, David-Ferdon, & Holt, 2015).
This study, situated within the criminology and sociology literature, will
focus on health risk behaviors that are associated with crime and deviance.
By “associated with crime and deviance,” it is meant that these behaviors in
teens have historically been connected to delinquent behaviors. For example,
the use of alcohol by teens and concurrent or subsequent juvenile delinquency
have been repeatedly linked (Felson, Savolainen, Aaltonen, & Moustgaard,
2008), as has drug use (Wilson & Petersilia, 2011). The association with risky
sexual behavior (having sex with multiple partners, having sex after drug and
alcohol use, sex without birth control) is less empirically established.
However, work by Armour and Haynie (2007) suggested that early sexual
activity increases the risk of future deviant behavior.

Theorizing the Relationship Between Cyberbullying and Deviant


Health Risk Behaviors
This study explains the link between cyberbullying victimization and partici-
pation in deviant health risk behaviors by applying general strain theory
(GST). GST has been applied frequently to explain the link between cyber-
bullying and deviant behavior (Agnew, 2001; Hay et al., 2010; Jang, Song, &
Kim, 2014; Keith, 2017; Patchin & Hinduja, 2011). The foundational propo-
sition underpinning GST is that the presence of unwanted or noxious stimuli,
called strains, produces negative emotions that can lead to deviant behavior.
GST posits three categories of strains: (a) not being able to achieve one’s
goals, (b) the presentation of noxious or negatively valued stimuli, and (c) the
loss of positively valued stimuli. Cyberbullying is a strain of the second
type—the hostile or aggressive messages intended to inflict harm or discom-
fort on others are noxious stimuli. To cope with the emotions produced by
strain, a person may engage in deviant behavior. In a later article, Agnew
(2001) extended and further specified GST by describing the types of strain
most likely to lead to crime and deviance. These strains are perceived as
unjust and high in magnitude, are associated with low social control from
authorities, and create an inventive for coping through criminal means. Peer
abuse (bullying), Agnew argues, is a type of strain that can lead to crime and
Wood and Graham 5

deviance. It is likely to be perceived as unjust because young people will


think they are being treated unfairly. It is likely to be perceived as high in
magnitude because peers are central to the lives of young people. It occurs in
spaces where there are few authority figures present—bullying happens on
playgrounds or in moments when teachers and adults are not monitoring.
Finally, students who bully are often engaged in other deviant acts, thereby
modeling to the victim forms of criminal coping.
Several studies have linked strain to cyberbullying. Hay et al. (2010)
found that both physical bullying and cyberbullying are associated with
delinquency. Both forms of bullying were positively associated with external
acts of deviance toward others, and internal acts of self-harm. Jang et al.
(2014) linked physical bullying to cyberbullying, with the former acting as a
strain and the latter being a response. Keith (2017) found that being bullied
was positively associated with two coping mechanisms, avoidance behavior
and bringing a weapon to school.

Additional Factors
There are at least two additional factors that can impact the rate of deviant
health risk behaviors. One factor is the amount of self-control an individual
possesses. Individuals with low self-control have greater difficulties judging
the long-term consequences of their actions. As a result, they are more likely
to find deviance and crime attractive (Gottfredson & Hirschi, 1990). Low
levels of self-control has been linked to immediately satisfying but ultimately
detrimental health drug use and risky sexual behavior (Baron, 2003; Benda,
2005; Ford & Blumenstein, 2013). A second confounding factor may be the
quality of peers the individual associates with. Individuals who interact with
peers exhibiting deviant or criminal behaviors are themselves more likely to
participate in those exhibited behaviors (Akers, 1996). Studies have shown
that young people who have peers modeling and reinforcing sexual activity
and substance abuse are more likely engage in these behaviors (Dembo,
Wareham, & Schmeidler, 2007; Martino et al., 2011; Pereyra & Bean, 2017).
Self-control and deviant peer associations are discussed here to add nuance
to the current study. This study, undergirded by previous research, assumes that
the reason why one should expect cyberbullying to be associated with health
risk behaviors is because of emotional strain. However, other reasons not explic-
itly linked to cyberbullying can impact deviant behavior. And so while levels of
self-control or deviant peer associations would not theoretically occupy the
position of a causal mechanism between cyberbullying victimization and devi-
ant health risk behaviors, and therefore explain a statistical correlation, they
nonetheless can impact observed rates of deviant health risk behaviors.
6 Youth & Society 00(0)

Data
Data for this study comes from the 2015 YRBS conducted by the CDC. Variations
of the YRBS have been used in several studies by scholars interested in cyberbul-
lying within an American context (Duong & Bradshaw, 2014; Grinshteyn &
Yang, 2017; Rice et al., 2015). The YRBS monitors several types of health risk
behaviors including sexual behaviors associated with unintended pregnancy and
sexually transmitted diseases, alcohol use, drug use, and tobacco use. The YRBS
employs a three-stage cluster sample design to collect a representative sample of
ninth- through 12th-grade students. First, 180 schools were randomly selected
from a sampling frame of all regular public, Catholic, and private schools in the
United States, Grades 9 to 12. One hundred twenty-five schools responded (a rate
of 69%) and participated. Next, equal probability sampling was used to select
classes from each participating school. Over 18,165 students were issued a ques-
tionnaire of which 15,624 responded (a rate of 86%). The overall response rate
was 60% (69% × 86%). The analytic method employed, clustering, requires
complete cases. The sample size for the current study was 9,122.

Independent Variables
The primary independent variable was cyberbullying. This was measured as
“During the past 12 months, have you ever been electronically bullied?
(Count being bullied through e-mail, chat rooms, instant messaging, web-
sites, or texting.) Yes or No.” Physical bullying is also measured. This is
measured as “During the past 12 months, have you ever been bullied on
school property? Yes or No.”
Above, additional factors that scholars have theorized and demonstrated
to impact rates of deviant health risk behaviors—low self-control and deviant
peer associations—were discussed. Ideally, statistical models can control for
these factors to isolate the effects of cyberbullying. However, the YRBS does
not allow for direct measures of these factors. Self-control can be indirectly
measured through self-reported grades, as higher levels of self-control have
been linked to more successful academic outcomes (Duckworth & Seligman,
2005; Felson & Staff, 2006; Stadler, Aust, Becker, Niepel, & Greiff, 2016).
The measure used asked the student “During the past 12 months, how would
you describe your grades in school?” The student could reply, mostly As (0),
mostly Bs (1), mostly Cs (2), mostly Ds (3), or mostly Fs (4).

Dependent Variables
There were four dependent variables in the analysis: multiple sexual partners,
marijuana use, and alcohol use, all in the past 30 days. These variables were
Wood and Graham 7

measured at the ordinal level. These variables are listed here with the variable
labels and numerical codes for statistical analysis in brackets.
Students were asked about their sexual activity: “[Number of sexual part-
ners] During the past 3 months, with how many people did you have sexual
intercourse?” Students could respond, I have never had sexual intercourse
(0); I have had sexual intercourse, but not during the past 3 months (1); 1
person (2); 2 people (3); 3 people (4); 4 people (5); 5 people (6); 6 or more
people (7).
Students were asked about their use of marijuana: “[Marijuana Use]
During the past 30 days, how many times did you use marijuana?” Students
could respond, 0 times (0), 1 or 2 times (1), 3 to 9 times (2), 10 to 19 times
(3), 20 to 39 times (4), 40 or more times (5), 100 or more times (6).
Students were asked about the number of alcoholic drinks they had con-
sumed. “[Alcohol Use] During the past 30 days, on how many days did you
have at least one drink of alcohol?” Students could respond, 0 days (0), 1 or
2 days (1), 3 to 5 days (2), 6 to 9 days (3), 10 to 19 days (4), 20 to 29 days (5),
all 30 days (6).
Students were asked about the number of cigarettes they had smoked:
“[Cigarette Smoking] During the past 30 days, on how many days did you
smoke cigarettes?” Students could respond, 0 days (0), 1 or 2 days (1), 3 to 5
days (2), 6 to 9 days (3), 10 to 19 days (4), 20 to 29 days (5), all 30 days (6).

Controls
The control variables included in statistical models are gender, race/ethnicity,
grade, and sexual orientation. These variables are all categorical. The refer-
ence category for gender is respondents who identified as male, for race/
ethnicity it is respondents who identified as White, for grade it is respondents
who are in ninth grade, and for sexual orientation it is respondents who iden-
tified as straight. Table 1 shows descriptive statistics for variables used in the
analysis.

Method
Clustering Deviant Behavior
Although most studies isolate individual deviant behaviors in regression
models, a more realistic situation is that health risk behaviors co-occur
(Guilamo-Ramos, Litardo, & Jaccard, 2005; Hair, Park, Ling, & Moore,
2009). Therefore, cluster analysis is used. Specifically, the Partitioning
around Medoids (PAM) algorithm and the clustering around large applica-
tions (CLARA) method are used to judge empirically how deviant behaviors
8 Youth & Society 00(0)

Table 1.  Descriptive Statistics (n = 9,122).

Categorical (% of total)

Gender
 Female 4,693 (51.4)
 Male 4,429 (48.6)
Race
 White 4,218 (46.2)
  American Indian/Alaska Native 89 (1.0)
 Asian 405 (4.4)
  Black or African American 797 (8.7)
  Hawaiian/Pacific Islanders 34 (<1.0)
 Hispanic 1,487 (16.3)
  Multiple (Hispanic) 1,673 (18.3)
  Multiple (Non-Hispanic) 419 (4.6)
Grade
  9th grade 2,226 (24.4)
  10th grade 2,219 (24.3)
  11th grade 2,370 (26.0)
  12th grade 2,307 (25.3)
Sexual orientation
 Straight 8,176 (89.6)
 Gay/lesbian 154 (1.7)
 Bisexual 537 (5.9)
  Not sure 255 (2.8)
Bullying types No Yes
 Bullying 7,368 (80.8) 1,754 (19.2)
 Cyberbullying 7,777 (85.3) 1,345 (14.7)
  M SD Range
Continuous
  GPA (Self-control proxy) 1.00 0.91 0-4
  No. of sexual partners 0.84 1.22 0-7
  Marijuana use 1.31 2.04 0-6
  Alcohol use 1.05 1.97 0-6
  Cigarette smoking 0.32 1.12 0-6

Note. 0 = mostly As, 1 = mostly Bs, 2 = mostly Cs, 3 = mostly Ds, 4 = mostly Fs. GPA = grade
point average.

are clustered within a sample population, and then predict cluster member-
ship. More common methods of partitioning use the arithmetic mean in
Wood and Graham 9

clustering and have smaller data sets. This method is chosen because it is
designed for analyses of a large number of cases, and because the variables
clustered are ordinal. Moreover, because cyberbullying and physical bullying
are associated with each other (Finkelhor, Shattuck, Turner, Ormrod, &
Hamby, 2011; Garnett & Brion-Meisels, 2017; Hay et al., 2010; Keith, 2017;
Payne & Hutzell, 2015; Resett & Gamez-Guadix, 2017), models will specify
the effects of these two phenomenon in distinct models, and their effect on
the likelihood of being in a certain cluster.
For a detailed analysis of the clustering method employed see Kaufman
and Rousseeuw (1990). The method is described here in brief. The PAM
algorithm selects at random a representative object(s) from the data set and
clusters remaining objects that are closest in Euclidean distance around these
initial objects. Through a process of test and replace, the representative
objects are selected that, after objects are clustered around them, maximize
the intraclass similarity (distance between objects in the same cluster), and
minimize the interclass similarity (distance between objects of different clus-
ters). These objects, called medoids, are used to represent the cluster. CLARA
is an extension of PAM for larger data sets. CLARA selects a series of smaller
samples from the original sample, applies PAM, and then returns the best
clustering model.

Comparing Physically Bullied and Not Bullied Subsets


Prior research suggests both overlap and uniqueness with respect to physical
bullying phenomena and cyberbullying phenomena (Hong et al., 2016; Keith,
2017; Kowalski & Limber, 2013; Low & Espelage, 2013; Resett & Gamez-
Guadix, 2017; Sticca & Perren, 2013; Waasdorp & Bradshaw, 2015). Some
studies have included both cyberbullying and bullying in the same regression
model (e.g., Keith, 2017). Other studies have created subsets of students who
are either bullied, cyberbullied, both, or none, and then predicted the causes
or effects of being in these population (Hertz et al., 2015; Kowalski & Limber,
2013; Resett & Gamez-Guadix, 2017; Waasdorp & Bradshaw, 2015). This
later analytic approach is more justified statistically, as these two phenomena
are likely correlated and should not be predicted in the same regression
model. Therefore, models will measure the effects of cyberbullying control-
ling for bullying by dividing the sample population into discrete subsets of
students. One subset will be composed of students who have been physically
bullied, and a second subset will be composed of students who have not been
physically bullied.
10 Youth & Society 00(0)

Table 2.  Chi-Square Test (n = 9,122).

Cyberbullying

Bullying No Yes
No 6,884 (6,282) 484 (1,086)
Yes 893 (1,495) 861 (259)

Note. Expected scores in parenthesis; χ2 = 2,200, df = 1, p value ⩽ .001.

Analysis
Testing the Independence of Physical and Cyberbullying
The chi-square test shows that the null hypothesis that bullying and cyberbul-
lying are independent is rejected (Table 2). This conforms to prior research
also showing a link between physical bullying and cyberbullying (Hong
et al., 2016; Keith, 2017; Low & Espelage, 2013). Given the interdependence
of these two phenomena, regression models must be designed to take this into
account. Therefore, the sample population is divided into discrete subsets of
students who have been bullied and students who have not, and then apply
the effect of cyberbullying to each one.

Clustering Deviant Behavior


One of the central questions when doing cluster analysis is the decision of
what number of clusters is valid. Ultimately, that decision is subjective.
However, it can be informed by at least two factors: (a) the optimum amount
of variation explained—such that upon extracting x number of clusters, x +
1 does not appreciably improve the overall amount of variation explained, (b)
theory—such that the clusters produced conform to a priori assumptions
about the organization of cases.
To determine the optimum variation explained, the silhouette width is
used. The silhouette width is the average of each observation’s silhouette
value. The silhouette value measures the degree of confidence in the cluster-
ing assignment of a given observation, with well-clustered observations hav-
ing values near +1 and poorly clustered observations having values near −1.
Observation i is defined as
bi − ai
S (i ) = ,
max ( bi , ai )
Wood and Graham 11

where ai is the average distance between i and all other observations in the
same cluster, and bi is the average distance between i and the observations in
the “nearest neighboring cluster,” which is defined as
min dist ( i, j )
bi = ∑
Ck ∈ C ( i ) j∈C n ( Ck )
,
k

where C(i) is the cluster containing observation i, dist(i, j) is the distance


between observations i and j, and n(C) is the cardinality of cluster C. The
silhouette width is between the interval [−1,1] . The best solution (i.e., num-
ber of clusters) produces the highest silhouette width.
Using the silhouette width as a measure, a two-cluster solution produces
the optimum amount of variation explained. A two-cluster solution explains
about 73% of the variance in the four health risk behaviors. Three and four
cluster solutions are possible, but do not appreciably increase the amount of
variance explained. Theoretically, a small number of clusters conforms the
notion that deviant health risk behaviors should co-occur.
The representative case of the first cluster, called the medoid, is sexual
frequency = 0, cigarette smoking = 0, alcohol usage = 0, and marijuana
smoking = 0. These are young people who do not partake in the health risk
behaviors measured. This cluster comprises 5,974 respondents. This cluster
is labeled “safe” because the average case within this cluster reports no
engagement in the activities measured.
The representative case of the second cluster is indicated by respondents
who report a sexual frequency of 2 (having sex with two people over the last
3 months), cigarette usage of 0 (smoking cigarettes 1 or 2 days over the last
30 days), alcohol consumption of 2 (drinking alcohol 20 to 29 days over the
last 30 days), and marijuana usage of 3 (smoking marijuana 20 to 39 times
over the past 30 days). This second cluster is comprised of 3,148 respon-
dents. This cluster is labeled as the “at-risk” cluster because the average case
in this cluster reports moderate to high levels of engaging in the activities
measured.
The relationship between clusters and types of bullying is shown Table 3.
Table 3 shows a statistical association between cluster type and type of bully-
ing victimization at the .001 level (i.e., the null hypothesis that the two cate-
gories are independent is rejected). We see higher than expected scores for
the cells in which cyberbullying and at-risk intersect. The observed value for
the cell containing respondents who were cyberbullied and categorized as
at-risk was 222. Whereas the expected value if the two categories were inde-
pendent of each other was 167. Similarly, the observed value for the cell
containing respondents who were cyberbullied and physically bullied, and
12 Youth & Society 00(0)

Table 3.  Chi-Square (n = 9,122).

Safe cluster At-risk cluster


Not bullied 4,653 (4,508) 2,231 (2,376)
Physically bullied only 609 (585) 284 (308)
Cyberbullied only 262 (317) 222 (167)
Both physically bullied and cyberbullied 450 (564) 411 (197)

Note. Expected scores in parenthesis; χ2 = 110.61, df = 3, p value ⩽ .001.

categorized as at-risk was 411. The expected value for that cell was 197. The
subsequent analysis will explore this relationship further controlling for other
factors.

Regression Analysis
Because there are two clusters, the proper procedure is binary logistic regres-
sion analysis. We report the odds ratios—the increase or decrease of being in
the “at-risk” cluster—in Table 4. An odds ratio greater than 1 for an indepen-
dent variable indicates that the variable’s effect is to increase the odds of
being in a cluster. Conversely, an odds ratio of less than 1 indicates that the
variable’s effect is to decrease the odds of being in a cluster.
The population of students is subset into those who were not physically
bullied (Models 1 and 2) and those who were physically bullied (Models 3
and 4). Two models are presented for both subsets, with the first model pre-
dicting the odds without a self-control measure (Models 1 and 3) and the
second model including this measure (Models 2 and 4). This is done to show
how the self-control measure may impact effects of cyberbullying on deviant
health risk behaviors.
Models 1 and 2 (the subset of students who were not physically bullied)
show that being a victim of cyberbullying significantly increases the odds of
being in the at-risk cluster. Specifically, a student who was cyberbullied was
87% more likely (1.87 − 1) to be in the “at-risk” cluster than a student who
was not cyberbullied in Model 1, and 77% more likely in Model 2.
Demographic variables are also associated with being in the “at-risk” cluster.
In both models, identifying as female and Asian lowers the odds when com-
pared with their respective reference group, while identifying as multiple
race/Hispanic, being in higher grade levels, and being bisexual increase the
odds when compared with their reference groups. The risk increases mono-
tonically for grades, with each higher grade increasing the odds in compari-
son to the ninth-grade reference group. Model 2 added self-reported grades as
Wood and Graham 13

Table 4.  Odds-Ratios for Predictors of “At-Risk” Cluster.

Have not been Have been physically


physically bullied bullied
(n = 7,367) (n = 1,754)

  Model 1 Model 2 Model 3 Model 4


Intercept 0.21*** 0.11*** 0.22*** 0.14***
Female 0.73*** 0.88* 0.72** 0.79*
Race/ethnicity (White = Reference)
  American Indian 1.24 0.84 1.33 0.96
 Asian 0.26*** 0.28*** 0.29*** 0.32**
  African American 1.19* 0.98 1.29 1.14
  Hawaiian/Pacific Islanders 1.19 1.05 3.80 4.70
 Hispanic 1.04 0.82* 0.88 0.68
  Multiple (Hispanic) 1.38*** 1.17* 1.28 1.06
  Multiple (non-Hispanic) 1.03 0.98 1.58* 1.42
Grade (9th grade = Reference)
  10th grade 1.69*** 1.72*** 1.56*** 1.58***
  11th grade 3.27*** 3.44*** 2.70*** 2.84***
  12th grade 4.48*** 4.90*** 4.11*** 4.65***
Sexual orientation (Straight = Reference)
 Gay/lesbian 1.14 1.04 1.08 1.10
 Bisexual 1.63*** 1.27* 1.95*** 1.72***
  Not sure 0.82 0.77 1.90* 1.63
Self-control — 1.79*** — 1.56***
Has been cyberbullied 1.87*** 1.77*** 2.23*** 2.12***
Nagelkerke R2 12.3% 18.5% 15.7% 20.4%

Note. *= < .05, **= < .01, ***= < .001.

a measure of self-control. Self-reported grades were significant, with each


unit increase associated with a 79% increase in the odds of being in the at-risk
cluster. The self-control measure was included to judge its impact on the
cyberbullying effect. This measure somewhat decreases the impact of cyber-
bullying, but its most noticeable effect was to increase the predictive power
of the model, with Model 1 explaining 12.3% of the variance and Model 2
explaining 18.5% of the variance.
Models 3 and 4, for the subset of students who were physically bullied,
share the same general characteristics as Models 1 and 2. Namely, being a
victim of cyberbullying significantly increases the odds of being in the at-risk
cluster controlling for demographic variables, including self-control only
14 Youth & Society 00(0)

marginally impacts this relationship, and the predictive power of the model
increases markedly with the inclusion of self-control. An interesting addi-
tional finding is that cyberbullying matters more for the physically bullied
subset of students (Models 3 and 4) than for the students who were not bul-
lied (Models 1 and 2). And so, in Model 2 cyberbullying victimization is
associated with a 77% increase in the odds of being in the at-risk cluster.
However, in Model 4, being a victim of cyberbullying is associated with a
112% increase.

Discussion
The question asked in this study was “Does being a victim of cyberbullying
increase the likelihood of health risk behaviors associated with juvenile
delinquency?” The findings suggest an answer in the affirmative. Specifically,
cyberbullying victimization is positively associated with being categorized
into a cluster of students who report higher rates of sexual frequency, alcohol
consumption, and marijuana usage. The magnitude and direction of the
cyberbullying effect is consistent for populations that have been only cyber-
bullied and for those that have been both cyberbullied and physically bullied.
The effect is also consistent controlling for standard demographic variables
and a proxy measure of self-control.
This study has added to the cyberbullying literature in two ways. First, a
clustering approach was applied to the effects of cyberbullying. Deviant
behavior co-occurs, and cluster analysis reflects that reality. Using cluster
analysis, four behaviors were clustered together—alcohol usage, cigarette
smoking, marijuana usage, and sexual frequency. Second, the effects of
cyberbullying victimization has been extended to deviant health risk behav-
iors. There have been numerous studies linking cyberbullying victimization
to adverse outcomes (Fredstrom et al., 2011; Hoff & Mitchell, 2009; Palermiti
et al., 2017; Patchin & Hinduja, 2010; Payne & Hutzell, 2015). However, the
links between cyberbullying and the deviant health risk behaviors explored
here—sexual frequency, alcohol, and drug use—have been understudied.
This study is a step toward addressing that gap.
There are practical implications to these findings. Cluster analysis is a data
reduction procedure that does not need to be informed by a priori theoretical
considerations. These attributes make this method less appealing for research-
ers who look for variations in data and are testing or extending previous theo-
retical understandings. However, these same characteristics make this
procedure useful for practitioners and education professionals. The proce-
dure used here could be applied to any given population of students who have
responded to a small number of questions about bullying and behavior.
Wood and Graham 15

Working under the assumption that deviant behavior co-occurs, professionals


could use the methods detailed in this study to identify what behaviors cluster
in their particular school context and how cyberbullying is associated with
those behaviors. The findings could give practitioners a more empirically
sound basis upon which to narrow and focus their interventions.
Although this study did not test or extend a theory, the statistical asso-
ciation between cyberbullying victimization and deviant health risk behav-
iors is explained by leveraging GST (Agnew, 1992, 2001). In other words,
it is suggested that the causal mechanism linking being victimized online
and deviant health risk behaviors is attempting to cope with emotional
“strains.” Criminological research has pointed out other possible reasons
why a young person would partake in deviant behavior, including interact-
ing with deviant peers (Dembo et al., 2007; Martino et al., 2011; Pereyra
& Bean, 2017) or possessing low levels of self-control (Baron, 2003;
Benda, 2005; Ford & Blumenstein, 2013). These factors could mitigate the
relationship proposed in this study. At the least, as is shown in the current
study, these factors will likely contribute to deviant behavior over and
above strains from cyberbullying.
As such, these additional factors point toward a path for future research.
Future statistical models will need to incorporate more precise measures of
strain, peer associations, and self-control into statistical models. Alternatively,
future work may wish to assume significant statistical associations, and
instead explore qualitatively how students experience the strains accompany-
ing cyberbullying victimization. As an example, Agnew (2001) listed four
factors linking strain to crime and deviance—the strain being perceived as
unjust, high in magnitude, is associated with low social control from the per-
petrator, and generates pressure to engage in crime. As these are primarily
subjective measures there is a need for interviews, ethnographies, or other
types of qualitative investigation.
Before closing, some shortcomings of this study are given. First, the mea-
sure of cyberbullying used is rather coarse, and does not identify different
dimensions of cyberbullying (Menesini, Nocentini, & Calussi, 2011). As an
example, in their study linking cyberbullying perpetration with strain, Patchin
and Hinduja (2011) asked a battery of questions tapping into types of cyber-
bullying that was then turned into a scale. Second, the YRBS does not allow
us to control for many standard sociological variables, including family com-
position and economic class. These variables are important as they can miti-
gate or exacerbate the effect of cyberbullying. Third, a correlation between
cyberbullying victimization and deviant behavior can be established, but not
a causal relationship. It is plausible that the relationship is in the opposite
16 Youth & Society 00(0)

direction of the one proposed here: some deviant behaviors, especially high
sexual frequency, could lead to cyberbullying.

Conclusion
National discussions about cyberbullying tend to focus on cases in which a
young person who has been victimized commits suicide. These high-profile
cases should not be neglected, as they point to some of the most damaging
effects of being a victim of cyberbullying. However, it is more often the case
that cyberbullying victimization has less sensational, but no less significant
consequences at the population level. Namely, cyberbullying is one of the
many factors that can lead to any number of deviant and criminogenic behav-
iors. These behaviors can be costly, as parents, schools, and, when these
young people become adults, the criminal justice system must expend emo-
tional and economic resources to address them.

Declaration of Conflicting Interests


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

Funding
The author(s) received no financial support for the research, authorship, and/or publi-
cation of this article.

ORCID iD
Roderick Graham https://orcid.org/0000-0001-6295-6620

References
Aboujaoude, E., Savage, M. W., Starcevic, V., & Salame, W. O. (2015). Cyberbullying:
Review of an old problem gone viral. Journal of Adolescent Health, 57(1), 10-
18. doi:10.1016/j.jadohealth.2015.04.011
Agnew, R. (1992). Foundation for a general strain theory of crime and delinquency.
Criminology, 30, 47-88.
Agnew, R. (2001). Building on the foundation of general strain theory: Specifying the
types of strain most likely to lead to crime and delinquency. Journal of Research
in Crime and Delinquency, 38, 319-361.
Akers, R. L. (1996). Is differential association/social learning cultural deviance the-
ory? Criminology, 34, 229-247.
Armour, S., & Haynie, D. L. (2007). Adolescent sexual debut and later delinquency.
Journal of Youth and Adolescence, 36, 141-152. doi:10.1007/s10964-006-9128-4
Wood and Graham 17

Baron, S. W. (2003). Self-control, social consequences, and criminal behavior:


Street youth and the general theory of crime. Journal of Research in Crime and
Delinquency, 40, 403-425. doi:10.1177/0022427803256071
Benda, B. B. (2005). The robustness of self-control in relation to form of delinquency.
Youth & Society, 36, 418-444. doi:10.1177/0044118X04268071
Calvete, E., Orue, I., Estévez, A., Villardón, L., & Padilla, P. (2010). Cyberbullying in
adolescents: Modalities and aggressors’ profile. Computers in Human Behavior,
26, 1128-1135. doi:10.1016/j.chb.2010.03.017
Centers for Disease Control and Prevention. (2018). Adolescent and school health:
Youth Risk Behavior Surveillance System (YRBSS). Retrieved from https://www
.cdc.gov/healthyyouth/data/yrbs/index.htm
Chapin, J., & Coleman, G. (2017). The cycle of cyberbullying: Some experi-
ence required. The Social Science Journal, 54, 314-318. doi:10.1016/j.sos-
cij.2017.03.004
Dembo, R., Wareham, J., & Schmeidler, J. (2007). A longitudinal study of cocaine
use among juvenile arrestees. Journal of Child & Adolescent Substance Abuse,
17, 83-109. doi:10.1300/J029v17n01_05
de Winter, A. F., Visser, L., Verhulst, F. C., Vollebergh, W. A. M., & Reijneveld, S.
A. (2016). Longitudinal patterns and predictors of multiple health risk behav-
iors among adolescents: The TRAILS study. Preventive Medicine, 84, 76-82.
doi:10.1016/j.ypmed.2015.11.028
Downey, J. C., Gudmunson, C. G., Pang, Y. C., & Lee, K. (2017). Adverse childhood
experiences affect health risk behaviors and chronic health of Iowans. Journal of
Family Violence, 32, 557-564. doi:10.1007/s10896-017-9909-4
Duckworth, A. L., & Seligman, M. E. P. (2005). Self-discipline outdoes IQ in predict-
ing academic performance of adolescents. Psychological Science, 16, 939-944.
doi:10.1111/j.1467-9280.2005.01641.x
Duong, J., & Bradshaw, C. (2014). Associations between bullying and engaging in
aggressive and suicidal behaviors among sexual minority youth: The moderating
role of connectedness. Journal of School Health, 84, 636-645.
Felson, R. B., Savolainen, J., Aaltonen, M., & Moustgaard, H. (2008). Is the associa-
tion between alcohol use and delinquency causal or spurious?* Criminology, 46,
785-808. doi:10.1111/j.1745-9125.2008.00120.x
Felson, R. B., & Staff, J. (2006). Explaining the academic performance-delinquency
relationship. Criminology, 44, 299-320. doi:10.1111/j.1745-9125.2006.00050.x
Finkelhor, D., Shattuck, A., Turner, H. A., Ormrod, R., & Hamby, S. L. (2011).
Polyvictimization in developmental context. Journal of Child & Adolescent
Trauma, 4, 291-300. doi:10.1080/19361521.2011.610432
Foody, M., Samara, M., & Carlbring, P. (2015). A review of cyberbullying and sug-
gestions for online psychological therapy. Internet Interventions, 2, 235-242.
doi:10.1016/j.invent.2015.05.002
Ford, J. A., & Blumenstein, L. (2013). Self-control and substance use among college
students. Journal of Drug Issues, 43, 56-68. doi:10.1177/0022042612462216
18 Youth & Society 00(0)

Fredstrom, B. K., Adams, R. E., & Gilman, R. (2011). Electronic and school-based
victimization: Unique contexts for adjustment difficulties during adolescence.
Journal of Youth and Adolescence, 40, 405-415. doi:10.1007/s10964-010-9569-7
Garnett, B. R., & Brion-Meisels, G. (2017). Intersections of victimization among mid-
dle and high school youth: Associations between polyvictimization and school
climate. Journal of Child & Adolescent Trauma, 10, 377-384. doi:10.1007/
s40653-017-0183-7
Gottfredson, M. R., & Hirschi, T. (1990). A general theory of crime. Stanford, CA:
Stanford University Press.
Grinshteyn, E., & Yang, Y. T. (2017). The association between electronic bullying
and school absenteeism among high school students in the United States. Journal
of School Health, 87, 142-149.
Guilamo-Ramos, V., Litardo, H. A., & Jaccard, J. (2005). Prevention programs for
reducing adolescent problem behaviors: Implications of the co-occurrence of
problem behaviors in adolescence. Journal of Adolescent Health, 36, 82-86.
doi:10.1016/j.jadohealth.2003.12.013
Hair, E. C., Park, M. J., Ling, T. J., & Moore, K. A. (2009). Risky behaviors in
late adolescence: Co-occurrence, predictors, and consequences. Journal of
Adolescent Health, 45, 253-261. doi:10.1016/j.jadohealth.2009.02.009
Hay, C., Meldrum, R., & Mann, K. (2010). Traditional bullying, cyber bullying, and
deviance: A general strain theory approach. Journal of Contemporary Criminal
Justice, 26, 130-147. doi:10.1177/1043986209359557
Hertz, M. F., Everett Jones, S., Barrios, L., David-Ferdon, C., & Holt, M. (2015).
Association between bullying victimization and health risk behaviors among
high school students in the United States. Journal of School Health, 85, 833-842.
doi:10.1111/josh.12339
Hinduja, S., & Patchin, J. W. (2010). Bullying, cyberbullying, and suicide. Archives
of Suicide Research, 14(3), 206-221.
Hoff, D. L., & Mitchell, S. N. (2009). Cyberbullying: Causes, effects, and rem-
edies. Journal of Educational Administration, 47, 652-665. doi:10.1108/
09578230910981107
Hong, J. S., Lee, J., Espelage, D. L., Hunter, S. C., Patton, D. U., & Rivers, T., Jr.
(2016). Understanding the correlates of face-to-face and cyberbullying victim-
ization among U.S. adolescents: A social-ecological analysis. Violence and
Victims, 31, 638-663. doi:10.1891/0886-6708.VV-D-15-00014
Jang, H., Song, J., & Kim, R. (2014). Does the offline bully-victimization influence
cyberbullying behavior among youths? Application of general strain theory.
Computers in Human Behavior, 31, 85-93. doi:10.1016/j.chb.2013.10.007
Jones, S. E., Pezzi, C., Rodriguez-Lainz, A., & Whittle, L. (2016). Health risk behav-
iors by length of time in the United States among high school students in five
sites. Journal of Immigrant and Minority Health, 18, 150-160. doi:10.1007/
s10903-014-0151-3
Kaufman, L., & Rousseeuw, P. J. (1990). Wiley Series in Probability and Mathemat­
ical Statistics. Finding groups in data: An introduction to cluster analysis. New
York, NY: Wiley-Blackwell.
Wood and Graham 19

Keith, S. (2017). How do traditional bullying and cyberbullying victimization affect


fear and coping among students? An application of general strain theory. American
Journal of Criminal Justice, 43, 67-84. doi:10.1007/s12103-017-9411-9
Kowalski, R. M., & Limber, S. P. (2013). Psychological, physical, and academic cor-
relates of cyberbullying and traditional bullying. Journal of Adolescent Health,
53(Suppl. 1), S13-S20. doi:10.1016/j.jadohealth.2012.09.018
Low, S., & Espelage, D. (2013). Differentiating cyber bullying perpetration from non-
physical bullying: Commonalities across race, individual, and family predictors.
Psychology of Violence, 3(1), 39-52. doi:10.1037/a0030308
Martino, S. C., Tucker, J. S., Ryan, G., Wenzel, S. L., Golinelli, D., & Munjas, B.
(2011). Increased substance use and risky sexual behavior among migratory
homeless youth: Exploring the role of social network composition. Journal of
Youth and Adolescence, 40, 1634-1648. doi:10.1007/s10964-011-9646-6
Menesini, E., Nocentini, A., & Calussi, P. (2011). The measurement of cyberbul-
lying: Dimensional structure and relative item severity and discrimination.
Cyberpsychology, Behavior, and Social Networking, 14, 267-274. doi:10.1089/
cyber.2010.0002
Olweus, D. (1993). Bullying at school: What we know and what we can do
(Understanding children’s worlds). Oxford, UK: Blackwell Publishers.
Palermiti, A. L., Servidio, R., Bartolo, M. G., & Costabile, A. (2017). Cyberbullying
and self-esteem: An Italian study. Computers in Human Behavior, 69, 136-141.
doi:10.1016/j.chb.2016.12.026
Patchin, J. W., & Hinduja, S. (2010). Cyberbullying and self-esteem. Journal of
School Health, 80, 614-621.
Patchin, J. W., & Hinduja, S. (2011). Traditional and nontraditional bullying among
youth a test of general strain theory. Youth & Society, 43, 727-751. doi:10.1177
/0044118X10366951
Payne, A. A., & Hutzell, K. L. (2015). Old wine, new bottle? Comparing interper-
sonal bullying and cyberbullying victimization. Youth & Society, 49, 1149-1178.
doi:10.1177/0044118X15617401
Pereyra, S. B., & Bean, R. A. (2017). Latino adolescent substance use: A mediating
model of inter-parental conflict, deviant peer associations, and parenting. Children
and Youth Services Review, 76, 154-162. doi:10.1016/j.childyouth.2017.03.001
Resett, S., & Gamez-Guadix, M. (2017). Traditional bullying and cyberbullying:
Differences in emotional problems, and personality. Are cyberbullies more
Machiavellians? Journal of Adolescence, 61, 113-116. doi:10.1016/j.adoles-
cence.2017.09.013
Rice, E., Petering, R., Rhoades, H., Winetrobe, H., Goldbach, J., Plant, A., . . . Kordic,
T. (2015). Cyberbullying perpetration and victimization among middle-school
students. American Journal of Public Health, 105(3), e66-e72.
Stadler, M., Aust, M., Becker, N., Niepel, C., & Greiff, S. (2016). Choosing between
what you want now and what you want most: Self-control explains academic
achievement beyond cognitive ability. Personality and Individual Differences,
94, 168-172. doi:10.1016/j.paid.2016.01.029
20 Youth & Society 00(0)

Sticca, F., & Perren, S. (2013). Is cyberbullying worse than traditional bullying?
Examining the differential roles of medium, publicity, and anonymity for the
perceived severity of bullying. Journal of Youth and Adolescence, 42, 739-750.
doi:10.1007/s10964-012-9867-3
Tokunaga, R. S. (2010). Following you home from school: A critical review and
synthesis of research on cyberbullying victimization. Computers in Human
Behavior, 26(3), 277-287.
Waasdorp, T. E., & Bradshaw, C. P. (2015). The overlap between cyberbullying and
traditional bullying. Journal of Adolescent Health, 56, 483-488. doi:10.1016/j.
jadohealth.2014.12.002
Wilson, J. Q., & Petersilia, J. (Eds.). (2011). Crime and public policy (2nd ed.). New
York, NY: Oxford University Press.
Ybarra, M. L., Boyd, D., Korchmaros, J. D., & Oppenheim, J. K. (2012). Defining
and measuring cyberbullying within the larger context of bullying victimization.
Journal of Adolescent Health, 51, 53-58. doi:10.1016/j.jadohealth.2011.12.031

Author Biographies
Frank R. Wood Jr. is a doctoral student in the Department of Sociology and Criminal
Justice at Old Dominion University. His research is at the intersection of race, sexual
orientation, and deviance.
Roderick Graham is an assistant professor in the Department of Sociology and
Criminal Justice at Old Dominion University. His research explores behavioral pat-
terns in the digital environment, including political movements, cybercrime, and
racial inequality.

Вам также может понравиться