Tonja R. Nansel, Mary Overpeck, Ramani S. Pilla, W. June Ruan, Bruce Simons-Morton, Peter Scheidt. Bullying Behaviors Among US YouthPrevalence and Association With Psychosocial Adjustment. JAMA. 2001;285(16):2094–2100. doi:10.1001/jama.285.16.2094
Author Affiliations: National Institute of Child Health and Human Development, Bethesda, Md. Dr Overpeck is now with the Maternal and Child Health Bureau, Health Resources and Services Administration, Rockville, Md; Dr Pilla is now with the University of Illinois at Chicago.
Context Although violence among US youth is a current major concern, bullying
is infrequently addressed and no national data on the prevalence of bullying
Objectives To measure the prevalence of bullying behaviors among US youth and to
determine the association of bullying and being bullied with indicators of
psychosocial adjustment, including problem behavior, school adjustment, social/emotional
adjustment, and parenting.
Design, Setting, and Participants Analysis of data from a representative sample of 15 686 students
in grades 6 through 10 in public and private schools throughout the United
States who completed the World Health Organization's Health Behaviour in School-aged
Children survey during the spring of 1998.
Main Outcome Measure Self-report of involvement in bullying and being bullied by others.
Results A total of 29.9% of the sample reported moderate or frequent involvement
in bullying, as a bully (13.0%), one who was bullied (10.6%), or both (6.3%).
Males were more likely than females to be both perpetrators and targets of
bullying. The frequency of bullying was higher among 6th- through 8th-grade
students than among 9th- and 10th-grade students. Perpetrating and experiencing
bullying were associated with poorer psychosocial adjustment (P<.001); however, different patterns of association occurred among
bullies, those bullied, and those who both bullied others and were bullied
Conclusions The prevalence of bullying among US youth is substantial. Given the
concurrent behavioral and emotional difficulties associated with bullying,
as well as the potential long-term negative outcomes for these youth, the
issue of bullying merits serious attention, both for future research and preventive
Bullying among school-aged youth is increasingly being recognized as
an important problem affecting well-being and social functioning. While a
certain amount of conflict and harassment is typical of youth peer relations,
bullying presents a potentially more serious threat to healthy youth development.
The definition of bullying is widely agreed on in literature on bullying.1- 4 Bullying
is a specific type of aggression in which (1) the behavior is intended to
harm or disturb, (2) the behavior occurs repeatedly over time, and (3) there
is an imbalance of power, with a more powerful person or group attacking a
less powerful one. This asymmetry of power may be physical or psychological,
and the aggressive behavior may be verbal (eg, name-calling, threats), physical
(eg, hitting), or psychological (eg, rumors, shunning/exclusion).
The majority of research on bullying has been conducted in Europe and
Australia. Considerable variability among countries in the prevalence of bullying
has been reported. In an international survey of adolescent health-related
behaviors, the percentage of students who reported being bullied at least
once during the current term ranged from a low of 15% to 20% in some countries
to a high of 70% in others.5,6
Of particular concern is frequent bullying, typically defined as bullying
that occurs once a week or more. The prevalence of frequent bullying reported
internationally ranges from a low of 1.9% among 1 Irish sample to a high of
19% in a Malta study.1,7- 12
Bullying takes many forms, and findings about the types of bullying
that occur are fairly similar across countries. A British study involving
23 schools found that direct verbal aggression was the most common form of
bullying, occurring with similar frequency in both sexes.13
Direct physical aggression was more common among boys, while indirect forms
were more common among girls. Similarly, in a study of several middle schools
in Rome, the most common types of bullying reported by boys were threats,
physical harm, rejection, and name-calling.14
The most common forms for girls were name-calling, teasing, rumors, rejection,
and taking of personal belongings.
Research examining characteristics of youth involved in bullying has
consistently found that both bullies and those bullied demonstrate poorer
psychosocial functioning than their noninvolved peers. Youth who bully others
tend to demonstrate higher levels of conduct problems and dislike of school,
whereas youth who are bullied generally show higher levels of insecurity,
anxiety, depression, loneliness, unhappiness, physical and mental symptoms,
and low self-esteem.1- 4,8,15- 25
Males who are bullied also tend to be physically weaker than males in general.2 The few studies that have examined the characteristics
of youth who both bully and are bullied found that these individuals exhibit
the poorest psychosocial functioning overall.15,17,19,26
The current research provides a foundation for an understanding of the
bullying problem. However, it is insufficient to guide intervention and policy
development. Moreover, little is known specifically about bullying among US
youth.6 In one county-wide middle school survey,
24.1% of youth reported bullying others at least once in the past semester26; it is not known whether this is characteristic of
the rest of the nation.
The purpose of this study was to report the prevalence of bullying in
a nationally representative sample of US youth in grades 6 through 10, along
with information on differences in the prevalence of bullying by sex, grade,
and race. In addition, the relationships among bullying, being bullied, and
psychosocial adjustment are explored for 3 distinct groups: bullies only,
those bullied only, and those who both bully and are bullied.
The National Institute of Child Health and Human Development supported
a nationally representative survey of US youth in grades 6 through 10 during
spring of 1998. The survey, entitled the Health Behaviour of School-aged Children
(HBSC), was part of a collaborative, cross-national research project involving
30 countries and coordinated by the World Health Organization.27
The US survey was approved by the National Institute of Child Health and Human
Development Institutional Review Board and was carried out by Macro International
Inc (Calverton, Md). Both parental and student consent were solicited.
The US sampling universe consisted of all public, Catholic, and other
private school students in grades 6 through 10, or their equivalent, excluding
schools with enrollment of fewer than 14 students. The sample design used
a stratified 2-stage cluster of classes. The sample selection was stratified
by racial/ethnic status to provide an oversample of black and Hispanic students.
The sample was also stratified by geographic region and counties' metropolitan
statistical area status (largest urban areas/not largest urban areas) with
probability proportional to total enrollment in eligible grades of the primary
sampling units. Sample size was determined on the criteria of making estimates
for all US students in grades 6 through 10 with a precision of 3% at a 95%
confidence level, and for minority students with a precision of 5% at a 95%
An 83% participation rate was achieved. The school-based sample design,
using 1 class period for completion of the questionnaire, precluded ability
to compare respondent characteristics with those of nonparticipants. Responding
students in sampled classes were excluded if they were out of the target range
for grade or if age was outside of the 99th percentile for grade (n = 440
students), or if either grade or age were unknown (n = 39 students), yielding
an analytic sample of 15 686 students.
Measures were obtained from a self-report questionnaire containing 102
questions about health behavior and relevant demographic variables. Items
were based on both theoretical hypotheses related to the social context of
adolescents and measurements that had been validated in other studies or previous
WHO-HBSC surveys.27 Measures were pretested.
Bullying. Questions about bullying were preceded with the following explanation.10,28Here are some questions
about bullying. We say a student is BEING BULLIED when another student, or
a group of students, say or do nasty and unpleasant things to him or her.
It is also bullying when a student is teased repeatedly in a way he or she
doesn't like. But it is NOT BULLYING when two students of about the same strength
quarrel or fight.
Participation in bullying was assessed by 2 parallel questions that
asked respondents to report the frequency with which they bullied others in
school and away from school during the current term. Similarly, being bullied
was assessed by 2 parallel questions asking respondents to report the frequency
with which they were bullied in school and away from school during the current
term. Because the analytic focus of the current study was the relationship
of bullying behaviors to overall psychosocial adjustment, frequencies of bullying
behaviors in and out of school were combined for all analyses. Response categories
were "I haven't . . . ," "once or twice," "sometimes," "about once a week,"
and "several times a week." An analysis of the response distribution revealed
fewer subjects in the fourth category than the fifth, a deviation from the
expected skewed pattern. Hence, the latter 2 response options were collapsed.
Additional questions asked respondents to report the frequency with which
they were bullied in each of 5 ways—belittled about religion/race, belittled
about looks/speech, hit/slapped/pushed, subject of rumors or lies, and subject
of sexual comments/gestures.
Psychosocial Adjustment. Measures of psychosocial adjustment included questions about problem
behaviors, social/emotional well-being, and parental influences. Alcohol use
was measured by 3 items assessing frequency of alcohol consumption. The frequency
of smoking, fighting, and truancy were assessed by 1 item each. Academic achievement
was assessed by an item querying perceived school performance. Three items
(α = .70) queried the frequency of feeling lonely, feeling left out,
and being alone because others at school did not want to spend time with the
person. One item assessed ease of making friends. Three items (α = .72)
were used to assess relationship with classmates: "enjoy being together,"
"are kind and helpful," and "accept me." School climate was measured by 7
items (α = .82) related to the respondent's perception of the school
and teachers. Three items measured parental involvement in school (α
= .82), and 1 item assessed respondents' perceptions about their parents'
attitudes toward teen drinking.
Statistical sample weights were developed to adjust the minority oversampling
and to obtain student totals by grade comparable to population grade estimates
from the US National Center for Education Statistics. Weighted data analyses
were conducted using SUDAAN software.29 Descriptive
statistics were conducted using SUDAAN to obtain percentage distributions
and confidence intervals (CIs) based on the weighted data, with SEs adjusted
for the sample design. All CIs are shown at the 95% level.
To examine the relationship between psychosocial adjustment and bullying/being
bullied, students were classified as noninvolved, bullies only, those bullied
only, or both bully and bullied coincidentally, and a separate model was fit
for each outcome. Students who were neither bullies nor bullied served as
the reference group. Each outcome had 4 ordinal levels based on frequency
of the behavior—never, once or twice, sometimes, and once a week or
more. The proportional odds model30 was used
to examine the relationship between a range of psychosocial adjustment constructs
and each of the outcomes. Inherent in this model is the proportional odds
assumption, which states that the cumulative odds ratio for any 2 values of
the covariate is constant across response categories. Its interpretation is
that the odds of being in category ≤κ is exp[β′(x1 − x2)] times higher at the covariate vector x = x1 than at x = x2, where the parameter vector β contains
the regression coefficients for the covariate x. A cumulative logit function
was used to estimate the model parameters via the generalized estimating equations.31 The dependence of responses within clusters was specified
using an exchangable working correlation structure. To account for the dependence
between outcomes in estimating the variances, robust variance estimates were
used for the estimated parameters. The MULTILOG procedure of SUDAAN was used
to fit the proportional odds model with exchangable correlation structure.
Each model was first fit using the full sample, and then refit using 4 subsamples
stratified by sex and education level (middle school vs high school).
Overall, 10.6% of the sample reported bullying others "sometimes" (moderate
bullying) and 8.8% admitted to bullying others once a week or more (frequent
bullying), providing a national estimate of 2 027 254 youth involved
in moderate bullying and 1 681 030 youth in frequent bullying (Table 1). Experiencing bullying was reported
with similar frequency, with 8.5% bullied "sometimes" and 8.4% bullied once
a week or more, for a national estimate of 1 634 095 students bullied
with moderate frequency and 1 611 809 bullied frequently (Table 2). A sizable number of students
reported both bullying others and being bullied themselves. Of the total sample,
29.9% (an estimated 5 736 417 youth) reported some type of involvement
in moderate or frequent bullying, as a bully (13.0%), a target of bullying
(10.6%), or both (6.3%).
Demographic variation in the frequency of bullying was observed. Males
both bullied others and were bullied significantly more often than females.
Bullying occurred most frequently in 6th through 8th grade. Hispanic youth
reported marginally higher involvement in moderate and frequent bullying of
others, whereas black youth reported being bullied with significantly less
frequency overall. No significant differences in the frequency of being bullied
were observed among youth from urban, suburban, town, and rural areas (χ29 = 11.72, P = .24). However, small
differences were observed in the frequency of bullying others (χ29 = 19.13, P = .03): 2% to 3% fewer
suburban youth reported participation in moderate bullying, and 3% to 5% more
rural youth reported ever bullying than youth from town, suburban, and urban
areas (data not shown).
Table 3 presents the frequency
with which those bullied reported being bullied in each of 5 specific ways.
Being bullied through belittling one's looks or speech was common for both
sexes. Males reported being bullied by being hit, slapped, or pushed more
frequently than did females. Females more frequently reported being bullied
through rumors or sexual comments. Being bullied through negative statements
about one's religion or race occurred with the lowest frequency for both sexes.
Results of the analyses of the relationship among indicators of psychosocial
adjustment and bullying/being bullied using the proportional odds model are
presented in Table 4a. The overall
model for each of the outcomes was significant (P<.001).
All main effects were significant in at least 1 of the models. Table 4 also shows the estimated odds ratios for each psychosocial
adjustment construct in the model (adjusting for all other constructs in the
model), indicating the odds of having a greater frequency of the outcome variable
compared with the reference group.
Bullies, those bullied, and individuals reporting both bullying and
being bullied all demonstrated poorer psychosocial adjustment than noninvolved
youth; however, differences in the pattern of maladjustment among the groups
were observed. Fighting was positively associated with all 3 outcomes. Alcohol
use was positively associated with bullying and negatively associated with
being bullied. Smoking and poorer academic achievement were associated with
both bullying and coincident bullying/being bullied; poorer perceived school
climate was related only to bullying. Poorer relationships with classmates
and increased loneliness, on the other hand, were associated with both being
bullied and coincident bullying/being bullied. Ability to make friends was
negatively related to being bullied and positively related to bullying. A
permissive parental attitude toward teen drinking was associated only with
coincident bullying/being bullied, while increased parental involvement in
school was related to both being bullied and coincident bullying/being bullied.
Results from the analyses of the 4 sex/age subgroups (data not shown)
yielded findings similar to the model based on the full sample. No notable
differences among groups were observed for fighting, academic achievement,
perceived school climate, and relationship with classmates. However, differences
by sex and age were observed for several variables. While smoking was positively
associated with bullying and coincident bullying/being bullied among all groups,
the magnitude of the relationship was greater for middle school youth than
high school youth. Middle school males also showed a positive relationship
between loneliness and bullying; this was not the case for any of the other
groups. Among high school youth, bullying/being bullied was positively related
to alcohol consumption; this relationship was not observed among middle school
youth. High school females, on the other hand, did not demonstrate a significant
relationship between poorer friendship-making and being bullied, whereas the
other groups did. In addition, permissive parental attitude toward teen drinking
was associated with bullying/being bullied for all groups except high school
females. Finally, greater parental involvement in school was related to being
bullied and bullying/being bullied for males (both middle and high school)
but not females. It was related to bullying for high school males only.
This study indicates that bullying is a serious problem for US youth.
Consistent with previous studies,1,7,8,11,12
bullying was reported as more prevalent among males than females and occurred
with greater frequency among middle school–aged youth than high school–aged
youth. For males, both physical and verbal bullying were common, while for
females, verbal bullying (both taunting and sexual comments) and rumors were
more common. However, verbal bullying through derogatory statements about
one's religion or race occurred infrequently for both sexes. This finding
may reflect stronger social norms among adolescents against such behavior.
That is, it may be more socially acceptable for a youth to taunt peers about
their appearance than to make derogatory racial statements.
Both bullying and being bullied were associated with poorer psychosocial
adjustment; however, there were notable differences among those bullied, bullies,
and those reporting both behaviors. Those bullied demonstrated poorer social
and emotional adjustment, reporting greater difficulty making friends, poorer
relationships with classmates, and greater loneliness. Youth who are socially
isolated and lack social skills may be more likely targets for being bullied.
This is consonant with the finding by Hoover and colleagues32,33
that the most frequent reason cited by youth for persons being bullied is
that they "didn't fit in." At the same time, youth who are bullied may well
be avoided by other youth, for fear of being bullied themselves or losing
social status among their peers. Considering the high degree of relationship
observed, it is likely that both processes occur. Being bullied was also associated
with greater parental involvement in school, which may reflect parents' awareness
of their child's difficulties. Conversely, parental involvement may be related
to a lower level of independence among these youth, potentially making them
more vulnerable to being bullied. Interestingly, being bullied was associated
with less frequency of alcohol use and had a nonlinear relationship with smoking.
This is not altogether surprising, given Farrington's34,35
finding that socially inept youth were less likely to be involved in delinquency
than other youth.
Persons who bullied others were more likely to be involved in other
problem behaviors such as drinking alcohol and smoking. They showed poorer
school adjustment, both in terms of academic achievement and perceived school
climate. Yet they reported greater ease of making friends, indicating that
bullies are not socially isolated. Considering their greater involvement in
other problem behaviors, it is likely that these youth have friends who endorse
bullying and other problem behaviors, and who may be involved in bullying
Those youth who reported both bullying and being bullied demonstrated
poorer adjustment across both social/emotional dimensions and problem behaviors.
Considering the combination of social isolation, lack of success in school,
and involvement in problem behaviors, youth who both bully others and are
bullied may represent an especially high-risk group. It is not known whether
these youth were first bullied and then imitated the bullying behavior they
experienced or whether they were bullies who then received retaliation. Current
understanding tends to support the former explanation. Olweus2
describes a small subset of bullied youth he terms "provocative victims,"
individuals who demonstrate both anxious and aggressive behavior patterns
and who are known for starting fights and engaging in disruptive behavior.
Pellegrini and colleagues36,37
further discuss the "aggressive victim," defined as youth who respond to bullying
with reactive aggression. These youth do not tend to use aggression in a proactive
or instrumental manner, but rather are aggressive in retaliatory circumstances.
The patterns of relationships between bullying/being bullied and psychosocial
adjustment observed in this study were similar across age and sex groups,
providing support for the stability of the findings. The differences that
emerged may be useful for those conducting research or developing interventions
targeting specific populations. For example, the stronger relationship between
bullying and smoking observed among middle school youth may reflect an association
of bullying with deviance; as smoking becomes more normative in the older
youth, it is less associated with bullying. The lack of a relationship between
being bullied and poorer friendship-making among high school females could
indicate that by this age, females are more apt to find a peer group in which
they "fit," even though the peer group may consist of youth of similar social
Several limitations of the study should be noted. The HBSC is a broadly
focused survey regarding the health behaviors of middle– and high school–aged
youth. As such, more in-depth information, such as might be obtained from
an intervention study addressing bullying, are not available. This study includes
middle– and high school–aged youth but does not address elementary
school youth. The data are cross-sectional, and as such, the direction of
relationships among the variables cannot be determined. Another limitation
is the reliance on self-report for measurement of bullying. While self-report
is a common and accepted method of measuring bullying, individual perceptions
of bullying nevertheless may vary. To minimize subjectivity, students were
provided with a detailed definition of bullying along with examples.
While research on the long-term consequences of bullying is minimal,
the studies that have been conducted show negative effects into adulthood.
Olweus39 found former bullies to have a 4-fold
increase in criminal behavior at the age of 24 years, with 60% of former bullies
having at least 1 conviction and 35% to 40% having 3 or more convictions.
Their earlier pattern of achieving desired goals through bullying likely inhibited
the learning of more socially acceptable ways of negotiating with others.
Conversely, individuals formerly bullied were found to have higher levels
of depression and poorer self-esteem at the age of 23 years, despite the fact
that, as adults, they were no more harassed or socially isolated than comparison
adults.40 Those who have been bullied may view
such treatment as evidence that they are inadequate and worthless and may
internalize these perceptions. No study has assessed the long-term outcomes
for those who both bully others and are bullied. Given their initial poorer
adjustment status, it is possible that they fare worse than either bullies
or those bullied.
While this study provides important data on the prevalence and psychosocial
correlates of bullying among US youth, further research is needed. Of particular
importance would be prospective studies addressing factors that lead to bullying,
as well as studies on the long-term consequences of bullying and being bullied.
Longitudinal studies also would be valuable in better understanding the nature
of those who bully and are bullied.
The prevalence of bullying observed in this study suggests the importance
of preventive intervention research targeting bullying behaviors. Effective
prevention will require a solid understanding of the social and environmental
factors that facilitate and inhibit bullying and peer aggression. This knowledge
could then be used to create school and social environments that promote healthy
peer interactions and intolerance of bullying. School-based interventions
have demonstrated positive outcomes in Norway and England,40- 43
with reductions in bullying of 30% to 50%. These interventions focused on
changes within the school and classroom climate to increase awareness about
bullying, increase teacher and parent involvement and supervision, form clear
rules and strong social norms against bullying, and provide support and protection
for individuals bullied. This type of approach has not been tested in the