Ten-Year Trends in Bullying and Related Attitudes Among 4th- to 12th-Graders
BACKGROUND AND OBJECTIVES: Bullying is a significant public health concern, and it has received considerable attention from the media and policymakers over the past decade, which has led some to believe that it is increasing. However, there are limited surveillance data on bullying to inform our understanding of such trends over the course of multiple years. The current study examined the prevalence of bullying and related behaviors between 2005 and 2014 and explored whether any such changes varied across schools or as a function of school-level covariates.
METHODS: Youth self-reports of 13 indicators of bullying and related behaviors were collected from 246 306 students in 109 Maryland schools across 10 years. The data were weighted to reflect the school populations and were analyzed by using longitudinal hierarchical linear modeling to examine changes over time.
RESULTS: The covariate-adjusted models indicated a significant improvement over bullying and related concerns in 10 out of 13 indicators (including a decrease in bullying and victimization) for in-person forms (ie, physical, verbal, relational) and cyberbullying. Results also showed an increase in the perceptions that adults do enough to stop bullying and students’ feelings of safety and belonging at school.
CONCLUSIONS: Prevalence of bullying and related behaviors generally decreased over this 10-year period with the most recent years showing the greatest improvements in school climate and reductions in bullying. Additional research is needed to identify factors that contributed to this declining trend.
- FARM —
- free and reduced-priced meal
- HLM —
- hierarchical linear modeling
- ICC —
- intraclass correlation coefficient
What’s Known on This Subject:
Bullying is a significant public health concern receiving considerable attention from media and policymakers, leading some to believe that the rates are increasing. However, there are limited surveillance data on bullying available to understand such trends over multiple years.
What This Study Adds:
The current study included 13 indicators assessed over a 10-year period, including separate forms of bullying (relational, verbal, cyber) and school climate (safety and belonging). Results show significant improvements on bullying and related concerns especially in the most recent years.
Bullying has received considerable media attention over the past decade and has led some people to believe that bullying is on the rise.1 Research has also documented the negative impacts of bullying on mental health and behavioral, academic, and health outcomes,2,3 and all states have now adopted laws that address bullying.4 Yet, there are limited surveillance data on bullying to examine trends over multiple years. The current study examined the prevalence of bullying and related behaviors and attitudes over 10 years (2005–2014) using a population-based sample of students in grades 4 to 12. The extent to which any such changes varied as a function of school-level covariates was also examined.
Bullying is the form of victimization most frequently experienced by school-aged children.5–8 Since a landmark, large-scale surveillance study documented that 30% of youth are involved in bullying,6 there has been a significant increase in the research on various aspects of bullying, including prevalence, risk factors, and correlates.1,9–12 Multiple studies have documented both short- and long-term problematic effects and correlates of bullying, thereby demonstrating its significance as a public health concern.1,6,9,11,12 However, few studies have examined prevalence over multiple years; moreover, virtually none of the large, population-based studies have included longitudinal data or an in-depth examination of the prevalence across several indicators and characteristics of bullying.1 Despite these and other limitations regarding the availability of surveillance data on bullying,1,13 there have been some efforts to formulate conclusions regarding trends over time.1 For example, Finkelhor summarized bullying trends from 5 US national surveys by suggesting that bullying has decreased since the 1990s with less decline since 2007.14 The annual report of the Indicators of School Crime and Safety8 indicated that in 2013, 22% of youth ages 12 to 18 had been bullied, which was down from 28% in 2012.
Nevertheless, there are a number of limitations associated with the extant population-based data sources. For example, several studies have used measures of “bullying,” which was not well defined on the survey or did not meet the Centers for Disease Control and Prevention’s13 definition of bullying (ie, intentional aggressive behavior that occurs in the context of a power differential and is repeated or has the potential to be repeated). Moreover, previous studies have focused almost exclusively on victimization with limited exploration of perpetration,14 multiple forms of bullying (eg, physical, verbal, indirect, electronic), witnessing, or other bullying-related behaviors or attitudes (eg, beliefs about aggressive retaliation).8 Another limitation is the historical focus on a narrow age range, typically middle and/or early high school, with a lack of trend data spanning elementary through high school.8 This has precluded the field from formulating strong conclusions regarding potential developmental differences in bullying trends.7,8,14–16 Most exploration of trends in rates of bullying has been based upon t tests comparing the current rates to a single, earlier time point.8 Finally, few population-based studies have adjusted for school-level covariates despite research suggesting that school factors are potential risk factors for bullying.6,17,18 In summary, although much of the extant population-based survey research examining trends in bullying over time suggests a decline, there are several limitations caused by the previous studies’ large focus on secondary-school students and reliance on a relatively narrow band of indicators across just a few sequential years of data.8
The overarching goal of this study was to address the above-mentioned gaps in bullying research by examining trends in 13 indicators of bullying and related attitudes and behaviors in a population-based sample of youth in grades 4 to 12 across a decade. The timing of this study (2005–2014) is also important given the unprecedented increase in attention to this public health issue in recent years.1 Consistent with extant studies,1,8,14,16 slight reductions in the rates of bullying over time were expected (with the possible exception of cyberbullying, which smaller-scale studies suggest may be increasing19 due in part to greater availability of smartphones).20 We also explored whether the variation in bullying prevalence was functionally associated with school-level covariates21 such as school size, urbanicity, and school level (eg, secondary versus elementary schools).5,6,8,15
Participants were enrolled in grades 4 to 12 within a large public school district in Maryland; 109 schools were included in this analysis, of which 78 were elementary, 19 were middle, and 12 were high schools (see Tables 1 and 2). Alternative and special education settings were excluded because of small sample sizes and differences in the school structures and student populations.
Across the 10 years, there were a total of 246 306 respondents; see Tables 1 and 2 for year-specific data and Table 3 for student demographic characteristics. Student self-report surveys were completed anonymously and therefore could not be linked over time (see weighting procedure below). The student participation rate was ∼74% across all schools. For additional information on the data collection, see ref 5.
An online survey system was developed by researchers at the Johns Hopkins Center for the Prevention of Youth Violence in collaboration with school-based partners (see refs 5,22,23). The content for the survey was based on the Olweus Bully/Victim Questionnaire24 and the World Health Organization’s bullying survey.6
Participants in grades 4 and 5 completed an abbreviated version of the full survey to reduce the time and reading comprehension burdens on the younger children. Otherwise, the items completed by fourth- and fifth-graders and sixth- to twelfth-graders were identical. The survey included a definition of bullying that was consistent with the World Health Organization6 and the Centers for Disease Control and Prevention13: “A person is bullied when he or she is exposed, repeatedly and over time, to negative actions on the part of one or more other persons. Bullying often occurs in situations where there is a power or status difference. Bullying includes actions like threatening, teasing, name-calling, ignoring, rumor spreading, sending hurtful emails and text messages, and leaving someone out on purpose.”25,26 For additional information regarding the validity and reliability of self-reported measures of bullying using this approach, see refs 13,24.
Prevalence and Forms of Bullying
The primary indicators of interest included whether a student had been a frequent victim of bullying and had frequently perpetrated bullying (ie, 2 or more experiences) in the past month.24,25 A multiresponse item asked respondents to indicate whether they had experienced as a victim (yes/no) specific forms of bullying within the past month5,22,26 including 2 forms of physical bullying (pushing and slapping), 1 form of verbal bullying (threats), 1 form of relational bullying (spreading rumors), and cyberbullying (e-mail and blogs6,24). Participants were also asked, “During the past 30 days (month), how often have you seen someone else being bullied” (recoded to seen 1 or more times  and not seen ; witnessed bullying).
Attitudes toward aggressive retaliation (“It is okay to hit others if they hit me first”27), the overall perception of bullying as a problem (“Bullying is a problem at this school”), and the perceived role of adults in preventing bullying (“Adults are doing enough to prevent or stop bullying”) were also assessed. Participants responded on a 4-point Likert scale from strongly disagree to strongly agree. For the current analysis, these items were dichotomized into strongly disagree/disagree (0) and agree/strongly agree (1).
Perceptions of School Climate
A final set of items assessed students’ perceptions of belonging28 (“I feel like I belong at this school”) and safety (“I feel safe at this school”). Participants used the same 4-point Likert scale as above, which was also dichotomized, with (1) indicating a more favorable perception of the school.
School-level covariates from 2004 to 2005 (ie, the year preceding data collection) were publicly available data collected from the Maryland State Department of Education and included the school enrollment (ie, number of students), student-teacher ratio, attendance rate (ie, percent of students present in the school for at least half of the average school day during the school year), mobility rate (ie, number of entrants and withdrawals divided by the total enrollment), and the percent of students receiving free and reduced-priced meals (FARMs). The school level (ie, elementary, high) was also included.
Each year, the school district administered the online survey over the course of a 3- to 4-week window by using a passive parental-consent process. School staff (ie, teachers, guidance counselors, or school psychologists) administered the survey to students after a written procedure. Language arts classrooms were selected at random after a written procedure developed by the researchers to survey students in grades 4 to 12; the school staff members identified to administer the survey read aloud the instructions and bullying definition and emphasized the anonymous nature of the survey. Data were anonymously self-reported through a password-protected Web site on school computers that was only accessible during school hours. Limited participant demographic information was collected to ensure anonymity. The district provided the nonidentifiable data to the researchers for analysis. This study was approved by the researchers’ institutional review board (for additional detail on methods, see refs 5,17,22,23).
The data from years and schools were weighted29 to reflect the school populations by using grade, sex, and race/ethnicity data provided by students. Specifically, at each time point, survey weights were calculated for each student within his or her school using a raking ratio estimation procedure.30 The purpose of the weights was to reduce noncoverage biases.30 The weighted data were then aggregated to the school level at each time point to examine school-level changes in prevalence of the 13 bullying-related indicators over time. Percentages were calculated using the weighted count of students who endorsed each item divided by the total student weights in the school.
Hierarchical Linear Modeling
Repeated measures, 2-level growth models in hierarchical linear modeling (HLM) version 731 were fit for each of the 13 bullying indicators. Level 1 represented the 10 years of repeated observations for each of the school-level percentages, and Level 2 represented the school.
Fully unconditional (ie, unadjusted) models, with only a variable representing time as the predicting covariate, were fit for each outcome, allowing for an examination of the initial between-school variability as well as the unadjusted intercept (ie, prevalence) and slope (ie, growth) estimates. Covariate-adjusted models included school type (elementary, high), enrollment, attendance, student mobility, student-teacher ratio, and percentage of students receiving FARMs to predict the intercept and slope of time at Level 2. Grand-mean centering was employed for all continuous covariates whereas the dichotomous school-type variables were uncentered.32 Intraclass correlation coefficients (ICCs) and Cohen’s d effect size estimates were calculated as indicators of the change across time.
Weighted Descriptive Findings
See Table 4 for the weighted average responses and effect size of the changes across time for each of the 13 indicators. The rates of self-reported, frequent-bullying victimization ranged from 13.4% to 28.8% across the decade. With regard to bullying perpetration, the rates ranged from 7.0% to 21.3%. See Table 4 for ranges for the other indicators.
Covariate-Adjusted HLM Models
See Table 5 for both unadjusted and covariate-adjusted models.
There were 10 statistically significant slope effects in the unadjusted models, thereby indicating statistically significant changes over time. Specifically, being victimized (overall, β = −.02, P < .001), as well as being victimized by being pushed (β = −.03, P < .001), hit (β = −.03, P < .001), threatened (β = −.02, P < .001), cyberbullied (β = −.01, P < .001), and having rumors spread (β = −.02, P < .001) all decreased significantly over the 10 years. In addition, the prevalence of perpetrating bullying decreased significantly (β = −.02, P < .001); the same was true for witnessing bullying (β = −.04, P < .001). In terms of attitudes, significantly fewer students reported aggressive retaliatory attitudes over the 10 years (β = −.01, P < .001). Although perceptions that adults do enough to stop bullying showed significant improvement in the unadjusted models, it was no longer significant in the covariate-adjusted model. Finally, a statistically greater proportion of students reported feeling safe at school over time (β = .02, P < .001).
Associations With Covariates
The most consistently significant school-level covariate associated with the outcomes was school level. With regard to the intercept, elementary schools reported fewer bullying concerns compared with middle and high schools. With regard to the slope, elementary schools showed significantly less decline in reports of witnessing bullying over time compared with middle schools (β = .015, P < .01). Reports of feeling safe in elementary schools had significantly less improvement over time compared with middle schools (β = −.014, P < .01) whereas high schoolers reported significantly greater improvement in feelings of safety compared with middle schoolers (β = .02, P < .01). High schoolers demonstrated greater declines in witnessing bullying across time compared with middle schoolers (β = −.02, P < .05). See Table 5 for additional associations with covariates and the ICCs illustrating the extent of correlation among each variable over time.
This study examined the prevalence rates of 13 bullying-related indicators in a population-based sample over 10 years. Survey data from nearly one-quarter of a million students indicated that bullying has remained a prevalent, although declining, experience for school-aged youth. Specifically, 13.4% to 28.8% of students reported experiencing bullying in the past month, and approximately half of the students reported witnessing bullying. These estimates are consistent with recent bullying-prevalence reports,8 and they add to the current literature with the inclusion of a younger sample of youth. Despite these fairly high prevalence rates, the covariate-adjusted results for 10 of the 13 indicators suggested things may be getting better, as indicated by a reduction in bullying prevalence and related attitudes. The effect sizes of the change were in the small-to-moderate range (ranged 0.04–0.67) when comparing the first and last years’ data.
Some forms of victimization were commonly and consistently experienced across years (eg, relational). The prevalence of cyberbullying was consistently <10% and is comparable to the National Center for Education Statistics and the Bureau of Justice Statistics8 data. Based on previous research,19 it was hypothesized that cyberbullying might increase, but consistent with the other forms of bullying, cyberbullying also decreased. However, given the rapid change of technology and new social media platforms used by youth and increasingly at younger ages, the nature and quality of cyberbullying may change; therefore, future studies should examine cyberbullying in greater detail (eg, broader definitions).33 Physical, verbal, and relational bullying experiences dropped ∼2% each year to below 10% in the most recent year. Rates of perpetration reduced by 1% to 2% per year and dropped below 10% in recent years. Furthermore, witnessing bullying also decreased significantly across the decade.
Approximately 80% of students reported feeling safe and like they belong at school across the decade. Interestingly, ratings of safety, but not belonging, improved significantly over time. There were clear differences regarding bullying and climate across school levels. For example, high schools showed the most improvement across time; although this is promising, bullying peaks during middle school,5 so additional supports may be necessary during these school years. Although the yearly changes for all of the outcomes were small, some of these changes were fairly substantial across 10 years, as indicated by the effect size estimates comparing the first and last years (average d = 0.325). Notably, the most recent years evinced the greatest improvements in school climate and reductions in bullying. This could be due to the increase in bullying policies4 over the past decade as well as the simultaneous increase in effective evidence-based programming aimed at reducing school-based bullying.2,34 These factors could be associated with further decreases in bullying in years to come,34 so it will be important to continue to surveil these indicators over time.
The significant associations between school-level covariates and the change in bullying indicators were limited and not consistent across outcomes. Given the small magnitude of the covariate estimates and the large number of outcomes examined, readers should be cautious in the interpretation of the trends regarding school-level covariates.
Although this study possesses strengths such as the population-level sampling, the large sample size, the broad age range surveyed, and the longitudinal nature of the data, some limitations should be noted. For example, student data across years were not linked, given the anonymous nature of the survey, which limited the analyses that could be conducted. The anonymity was, however, an essential part of maximizing response rates and may have contributed to students being more candid in their responses; given the sensitive nature of the data collected, anonymity decreases the chances of response bias related to social desirability, which could otherwise be a threat to validity. Although the data collection and sampling procedures were consistent across the 10 waves of data collection, the district leadership has placed greater emphasis on the collection of data in recent years; this may have contributed to a slight uptick in the number of students participating in the survey in the last year of the data collection. However, the weighting procedure allows for generalizability of the sample to the full population of students within these schools. The use of sampling weights was a strength of this study, although we were somewhat limited in the number of variables we had to use as weights; therefore, we relied largely on basic demographic information that was available at both the student and school levels. This is a common approach in large-scale survey research.30
Although this study addresses the prevalence of bullying-related indicators over a decade, it does not shed light on what accounts for these changes. Information related to particular school efforts, programs, or initiatives to which reductions in bullying or related attitudes could be attributed was not assessed and, therefore, could not be examined. Future research is needed to systematically examine the implementation of antibullying programs and policies and the impact of these efforts.
The current findings suggest that there have been some improvements in the rates of bullying between 2005 and 2014. Although promising, it is important to emphasize that a large proportion of youth are still experiencing bullying, and the current prevalence rates continue to be of great concern.1 Nevertheless, these findings do contradict the public’s (mis)perception that bullying is increasing.1,35 With greater attention given to bullying in recent years, it is encouraging to see that there may be some improvements.
The authors’ time analyzing data for this paper was covered in part by grants from the US Department of Education, the William T. Grant Foundation, and the National Institute of Justice to Dr Catherine Bradshaw.
- Accepted March 9, 2017.
- Address correspondence to Tracy Evian Waasdorp, PhD, MEd, Department of Mental Health, Johns Hopkins University Bloomberg School of Public Health, 415 N. Washington St, Baltimore, MD 21231. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: The researchers’ time conducting secondary data analysis for this paper was supported in part by grants from the US Department of Education, the William T. Grant Foundation, and the National Institute of Justice.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
COMPANION PAPER: A companion to this article can be found online at www.pediatrics.org/cgi/doi/10.1542/peds.2017-0504.
- Committee on the Biological and Psychosocial Effects of Peer Victimization: Lessons for Bullying Prevention
- Board on Children, Youth, and Families
- Committee on Law and Justice
- Division of Behavioral and Social Sciences and Education
- Health and Medicine Division
- National Academies of Sciences, Engineering, and Medicine
- Swearer SM,
- Espelage DL,
- Vaillancourt T,
- Hymel S
- Bradshaw CP,
- Sawyer AL,
- O’Brennan LM
- Zhang J,
- Musu-Gillette L,
- Oudekerk B
- Nansel TR,
- Craig W,
- Overpeck MD,
- Saluja G,
- Ruan WJ; Health Behaviour in School-Aged Children Bullying Analyses Working Group
- Gladden RM,
- Vivolo-Kantor AM,
- Hamburger ME,
- Lumpkin CD
- Finkelhor D
- Jones LM,
- Mitchell KJ,
- Finkelhor D
- Lenhart A
- Bradshaw CP,
- Debnam KJ,
- Martin L,
- Gill R
- Olweus D
- Huesmann R,
- Guerra N,
- Miller L,
- Zelli A
- ↵Stata Statistical Software: Release 12.0 [computer program]. College Station, TX: StataCorp LP; 2011
- Battaglia MP,
- Izrael D,
- Hoaglin DC,
- Frankel MR
- ↵Hierarchical Linear and Nonlinear Modeling (HLM7) [computer program]. Lincolnwood, IL: Scientific Software International, Inc; 2011
- Luke DA
- Waasdorp TE,
- Horowitz-Johnson Z,
- Leff SS
- ↵Intel Security. Cyberbullying triples according to new McAfee “2014 teens and the screen study”. Available at: www.mcafee.com/us/about/news/2014/q2/20140603-01.aspx. Accessed April 13, 2016
- Copyright © 2017 by the American Academy of Pediatrics