Meta-Analysis of School-basd Bullying Prevention Programs
Content
School Psychology Review, 2012, Volume 41, No. 1, pp. 47–65
A Meta-Analysis of School-Based Bullying Prevention Programs’ Effects on Bystander Intervention Behavior Joshua R. Polanin Loyola University Chicago Dorothy L. Espelage University of Illinois Urbana—Champaign Therese D. Pigott Loyola University Chicago
meta-analysis synthesi synthesized zed bullying prevention programs’ effec Abstract. This meta-analysis tivenes tive nesss at inc increa reasing sing byst bystande anderr inte interve rventio ntion n in bull bullying ying situ situatio ations. ns. Evid Evidenc encee from 12 schoolschool-based based programs, programs, involvin involving g 12,874 students, indicated that overall the programs were successful (Hedges’s g .20, 95% confidence interval [CI] .11 to .29, p .001), with larger effects for high school (HS) samples compared to kindergarten through eighth-grade (K-8) student samples (HS effect size [ES] secondary ary synthesis from eight of the studies 0.43, K-8 ES 0.14; p .05). A second that reported empathy for the victim revealed treatment effectiveness that was positive but not signific significantly antly different different from zero ( g .05, 95% CI .07 to .17, p .45). .45). Nevert Nevertheless, heless, this metameta-analysi analysiss indicate indicated d that progra programs ms increa increased sed bystanderr interve bystande intervention ntion both on a practic practical al and statisti statistically cally significant significant level. These results suggest that researchers and school administrators should consider implementing programs that focus on bystander intervention behavior supplementary to bullying prevention progra programs. ms.
Bullying perpetration often occurs when bystan bys tander derss are pre presen sentt (Ha (Hawki wkins, ns, Pep Pepler ler,, & Craig, 2001; Lagerspetz, Bjorkqvist, Bertz, & King, 1982). In fact, some research has indicated that more than 80% of the time an observer ser ver wit witnes nesses ses vic victim timiza izatio tion n (O’ (O’Con Connel nell, l, Pepler, & Craig, 1999). Despite the presence of wi witn tnes esse sess an and d by byst stan ande ders rs,, ne near arly ly 1 in 3 children report victimization by a bully in the past 2 months (Frey, Hirschstein, Edstrom, & Snell, 2009; Nansel et al., 2001; Wang, Ian-
notti, & Nansel, 2009). Consequentially, bullyin ly ing g oc occu curs rs wi with th an au audi dien ence ce of me memb mber erss who pla play y mul multip tiple le rol roles es (Sa (Salma lmaval valli, li, Lag Lagererspetz, spe tz, Bjo Bjorkq rkqvis vist, t, Ost Osterm erman, an, & Kau Kaukia kianen nen,, 1996) and often fail to intervene on behalf of the victim with regularity. These bullying incident cid entss hav havee las lastin ting g neg negati ative ve eff effect ectss on the bully, victim, and bystanders (Olweus, 2002; Sweare Swe arer, r, Esp Espela elage, ge, Vil Villan lancou court, rt, & Hym Hymel, el, 2010; Sweeting, Young, West, & Der, 2006; Stevens, Oost, & Bourdeaudhuij, 2004).
Correspond Corres pondenc encee reg regard arding ing this art articl iclee shou should ld be add addres ressed sed to Josh Joshua ua R. Pola Polanin nin,, Loyo Loyola la Univ Univers ersity ity Chicago, 820 N. Michigan Ave., Suite 1022C, Chicago, IL 60611; e-mail: [email protected] Copyright 2012 by the National Association of School Psychologists, ISSN 0279-6015
47
School Psychology Review, 2012, Volume 41, No. 1
The past 20 years have seen a burgeoning of bullying prevention programs (Ferguson, Miguel, Kilburn, & Sanchez, 2007; Ryan & Sm Smit ith, h, 20 2009 09;; Tt Ttofi ofi & Fa Farr rrin ingt gton on,, 20 2009 09,, 2011). Researchers, school administrators, and teachers have used myriad designs, theories, and techniques in an attempt to mitigate the preval pre valenc encee of bul bullyi lying ng (As (Astor tor,, Mey Meyer, er, Ben Ben-benishty, Marachi, & Rosemond, 2005). Ttofi and Far Farrin ringto gton’s n’s (20 (2011) 11) rec recent ent lar largege-sca scale le meta-analysis of over 90 studies found that the majority of these programs have been successful at slowing the rate of bullying. Although successful bullying programs remain important accomplishments, Ttofi and Farrin Far ringto gton n (20 (2011) 11) fou found nd tha thatt few pro progra grams ms specifically target the behavior of bystanders (i.e.,, an indiv (i.e. individual idual who witne witnesses sses bullying). bullying). As such, prevention programs deemphasize a population that constitutes between 60% and 70% of primary or secondary school students (Glew, Fan, Katon, Rivara, & Kerntic, 2005; Rivers, Poteat, Noret, & Ashurst, 2009). This program oversight is unfortunate because observational research has found that when bystan st ande ders rs in inte terv rven enee on be beha half lf of th thee vi vict ctim im,, they the y suc succes cessfu sfully lly aba abate te vic victim timiza izatio tion n mor moree than 50% of the time (Craig, Pepler, & Atlas, 2000; O’Connell et al., 1999). Suppor Sup ported ted by the kno knowle wledge dge tha thatt bystanders can successfully intervene on behalf of the victim, a small amount of literature has focused recently on increasing this behavior.
Bullying in the Schools Definition Olweus (1973 Olweus (1973)) first described bullying as “m “mob obbi bing ng”” wh wher eree a gr grou oup p or in indi divi vidu dual al teases or harasses another individual. As such, early research focused solely on the physical aspects of school environment (e.g., teacher– student ratio), but found little connection to perpetration or victimization (Swearer et al., 2010). Recently, Frey et al. (2009) described bullyi bul lying ng as a soc social ial con constr struct uct tha thatt dis disrup rupts ts social connections among students. Ross and Horner Hor ner (20 (2009) 09) sum summar marize ized d the ple pletho thora ra of definitions: Common definitions of bullying involve repeated acts of aggression, intimidations, or coercion against a victim who is weaker in terms of physical size, psychological or social power, or other factors that result in a notable power differential. (p. 748)
The bully construct has received much rese re sear arch ch an and d ho host stss of de defin finit itio ions ns re rema main in.. Taken together, together, the bully generally involves an individual or group who incites physical or emot em otio iona nall ab abus usee on an anot othe herr in indi divi vidu dual al or group. Although other research exists on Internet and workplace bullying (Mishna, Cook, Saini, Wu, & MacFadden, 2010), this review focuses on school bullying.
Prevalence Preva lence and Negat Negative ive Effec Effects ts
These programs explicitly emphasize the importance porta nce of bysta bystander nder inter interventi vention on behav behavior ior and measure this construct. Given these conditions, the purpose of this meta-analysis is to synthesize synth esize schoo school-bas l-based ed bully bullying ing preve prevention ntion programs’ progr ams’ effec effectiven tiveness ess to chang changee bysta bystander nder intervention behavior. We also aggregated the program’s influence on empathy for the victim as a se seco cond ndar ary y sy synt nthe hesi siss be beca caus usee it ha hass re re-ceived cei ved rec recent ent inv invest estiga igatio tion n (Gi (Gini, ni, Alb Albier iero, o, Benelli, & Altoe`, `, 2007). The following summari ma rize zess th thee re rele leva vant nt li lite tera ratu ture re,, pr prov ovid ides es a comprehens compr ehensive ive exami examinatio nation n of the synth synthesis esis
School bullying is not a problem local to the Un the Unit ited ed St Stat ates es;; ra rath ther er,, it is re reco cogn gniz ized ed worldwide (Espelage & Swearer, 2003). Report po rtss fr from om Eu Euro rope pe to No Nort rth h Am Amer eric icaa ha have ve indicated that anywhere from 1% to 50% of students had been bullied or victimized within the last 2 months (Wang et al., 2009). Some observations have shown that as many as 30% of students were involved in bullying as either the bully or victim (Frey et al., 2009). Hymel and Swearer (2010) recently reported that 35% of stu studen dents ts ind indica icated ted bei being ng bul bullie lied d at lea least st once in the last 2 months with as many as 11%
proce process ss and quant quantitati itative ve analysis analysis analy sis and outcomes, outco mes, and elucidates moderator publication bias results. Suggestions for future research and policy are also provided.
of those bulliedMoremore than 2 orsampled 3 times reported in the lastbeing 2 months. over, some research has found that bullying roles remain relatively stable across time. In a
48
A Meta-Analysis of School-Based Bullying
sample of 516 middle school students, Espelage, Bosworth, and Simon (2001) found that individual indiv idualss who perpe perpetrate trated d bully bullying ing conti continnued to across multiple years. This finding has been replicated from other settings and populationss (McDo lation (McDougal, ugal, Hymel, & Vaill Vaillianco iancourt, urt, 2009; 200 9; Sch Scholt olte, e, Eng Engels els,, Ove Overbe rbeek, ek, Kem Kemp, p, & Haselager, Hasel ager, 2007; Sourander, Sourander, Helst Helstela, ela, Helenius, & Piha, 2000). Althou Alt hough gh bul bullyi lying’ ng’ss per pervas vasive ivenes nesss is cause for concern, the negative consequences relate rel ated d to bul bullyi lying ng rem remain ain tan tantam tamoun ount. t. Resear se arch cher erss ha have ve in indi dica cate ted d th that at bu bull llyi ying ng ha hass been bee n lin linked ked to ang anger er and mis miscon conduc ductt (Bo (Bossworth, Espelage, & Simon, 1999), criminal or delinquent behavior (Olweus, 2002), and suicida ci dall id idea eati tion onss (K (Kal alti tial alaa-He Hein ino, o, Ri Rimp mpel ela, a, Marttunen, Rimpela, & Rantanen, 1999). Victimization, on the other hand, has been linked to poo poorr phy physic sical al hea health lth (Ri (Rigby gby,, 199 1999), 9), low self-esteem (Rigby & Slee, 1993), depression (Sweeting et al., 2006), anxiety (Craig, 1994), and school avoidance (Kochenderfer & Ladd, 1996). Indeed, the negative effects of bullying and being bullied are persistent and problematic.
Bystander Definition, Roles, and Negative Effects
pening), assistant (e.g., follower of the bully), defender (e.g., being supportive of the victim), or out outsid sider er (e. (e.g., g., rem remain aining ing awa away y fro from m the bullying situation; Salmivalli et al., 1996). For the purposes purposes of thi thiss rev review iew,, we defi defined ned the bystander generally as any student who witnessed bullying episode, withthe the wit operationalizin ali zing g a cha charac racter terist istic ic bei being ng witnes nessin sing g pres pr esen ence ce of bu bull llyi ying ng,, re rega gard rdle less ss of ot othe herr characteristics. Despit Des pitee the these se lit litera eratur turee dis discor cordan dances ces,, few dis disagr agree ee abo about ut the adv advers ersee eff effect ectss wit wit-nessing bullying can have on the bystander. Bystanders felt significantly more uncomfortable in bullying situations compared to bullies (Stevens et al., 2004), and reported feelings of anxiety and insecurity (Rigby & Slee, 1993). This anxiety due to witnessing bullying has been linked to aggressive retaliation (MusherEizenman et al., 2004), and the fear of being bullied often prevented bystanders from seeking adult help (Unnever & Cornell, 2003). A rece re cent nt la larg rgee-sc scal alee st stud udy y co cond nduc ucte ted d in th thee United Kingdom found that compared to perpetrators, bystanders were at elevated risk for nonclinical outcomes (i.e., interpersonal sensitivity), and compared to victims bystanders were more likely to have elevated levels of substance abuse (Rivers et al., 2009).
A signi significant ficant proportion proportion of indiv individual idualss within school systems are considered individuals who are bystanders of bullying (Glew et
Evidence of the Bystanders’ Effects on Bullying
al., 20 al., 2005 05). ). Tw Twem emlo low, w, Fo Fona nagy gy,, an and d Sa Sacc cco o (2004) (20 04) defi defined ned a bys bystan tander der as an ind indivi ividua duall who lacks participation in bullying scenarios as either the bully or victim. The bystander may actively intervene to stop the bully, encourage the bully to continue, or view bullying passiv pas sively ely;; bys bystan tander derss can be eit either her boy boyss or girl gi rlss (C (Cow owie ie,, 20 2000 00;; Sm Smit ith, h, Tw Twem emlo low, w, & Hoover, 1998). Ther Th eree ar aree sp spec ecifi ificc ro role less th that at th thee by by-stander can demonstrate. Some authors refer to the bystander as a passerby, observer, witness, or participant (Salmivalli, Kaukianinen,
Individuals who are bystanders remain present more than 80% of bullying situations (O’Connell et al., 1999), and therefore some resear res earch ch has foc focuse used d on a soc social ial-ec -ecolo ologic gical al model of bullying prevention and intervention (Frey et al., 2009; Swearer & Espelage, 2004). The social presence and pervasiveness of the bystander fosters myriad opportunities to intervene. terve ne. For examp example, le, bysta bystanders nders suppo supported rted victim vic timss by rep report orting ing bul bullie liess to adu adults lts whe when n participating in a setting specifically designed to change bullying behavior patterns through bystanders (Sharp, Sellors, & Cowie, 1994).
& 2005; Twemlow et to al.,sustaining 2004); othersVoeten, described roles in relation or preventing the bullying behavior such as reinforcer (e.g., laughing or seeing what is hap-
Ross and Horner (2009) intervention recently implemented a school-wide bullying program that resulted in a decrease in reinforcing bystande sta nderr beh behavi avior or and bul bullyi lying ng per perpet petrat ration ion 49
School Psychology Review, 2012, Volume 41, No. 1
overall. Moreover, interventions that focused on dealing with conflict through peers instead of direct interventions with adults led to positive effects (Cowie & Hutson, 2006), and an individual’s willingness to intervene in bullying in g si situ tuat atio ions ns wa wass in inve vers rsel ely y re rela late ted d to th thee amount amo unt of pee peer-g r-grou roup p bul bullyi lying ng per perpet petrat ration ion (Espelage, Green, & Polanin, 2011).
Bystander Intervention Program Characteristics To date, best practice guidelines to promote effective bystander intervention behaviors remain undefined because research findings varied widely with regard to their implementat men tation ion foc focuse usess and app approa roache ches. s. Sev Severa erall mediums for interventions have been studied to tea teach ch chi childr ldren en abo about ut bys bystan tander der beh behavi avior, or, includ inc luding ing cla classr ssroom oom-ba -based sed dra drama ma (Me (Merre rrell, ll, 2004), media such as videotaped reenactments (McLaughlin, 2009; Schumacher, 2007), and individualized computer-adaptive software to track students’ progress within social scenarios and pro provid vided ed fee feedba dback ck on eff effect ective ive bystander behavior (Evers, Prochaska, Van Marter, Johnson, & Prochaska, 2007). However, all of these programs focus on bystander behavi ha vior or pe perh rhap aps, s, be beca caus usee th ther eree se seem emss to be some support for targeting peer-group behaviors to mitigate individual bullying (Salmivalli et al., 1996) 1996).. PeerPeer-group group intervention interventionss often encourage encou rage bysta bystander nder inter interventi vention on (Andr (Andreou, eou, Dida Di dask skal alou ou,, & Vl Vlac acho hou, u, 20 2008 08;; Fr Frey ey et al al., ., 2009 20 09;; St Stev even enss et al al., ., 20 2000 00)) or en enha hanc ncee by by-stande sta nderr emp empath athy y for the vic victim tim (Gi (Gini ni et al. al.,, 2007; Nickerson, Mele, & Princiotta, 2008). However, few studies have examined the effects of peer-group interventions on previctim empath emp athy y (Me (Merre rrell, ll, Gue Gueldn ldner, er, Ros Ross, s, & Isa Isava, va, 2008) 200 8) and the there re are res result ulting ing gui guidel deline iness to promote this behavior.
Previous Meta-Analyses on Bullying Prevention Programs A number of recent quantitative metaanalyses (Ferguson et; al., 2007; Farring ri ngto ton, n, 20 2009 09,, 20 2011 11; Merr Me rrel elll etTtofi al., al ., &20 2008 08;; Smith, Schneider, Smith, & Ananiadou, 2004) and qua qualit litati ative ve sys system temati aticc rev review iewss (Ry (Ryan an & 50
Smith, 2009) have been conducted regarding bullyi bul lying ng and vic victim timiza izatio tion n int interv ervent ention ion and prevention programs. However, none of these meta me ta-a -ana naly lyse sess fo focu cuse sed d sp spec ecifi ifica call lly y on by by-stander intervention constructs. Merrell et al.’s (2008) review included three studies that measured “intervene “intervene to stop bullying behavior,” behavior,” which resulted in a mean effect size of 0.17, but this was a secondary analysis of a small number of studies. Therefore, the goal of the present study was to conduct a meta-analysis thatt wou tha would ld dir direct ectly ly add addres resss bys bystan tander der int interervention behavior and empathy attitudes. Given Giv en thi thiss goa goal, l, two pri primar mary y res resear earch ch questions are addressed: 1. What What is th thee av aver erag agee tr trea eatm tmen entt ef effe fect ct,, across the current literature, of bullying prevention preve ntion programs on bysta bystander nder intervention behavior? 2. What study study charact characteristi eristics cs produced produced the the largest treatment effect? A se seco cond ndar ary y re rese sear arch ch qu ques esti tion on ad ad-dressed bystander empathy for the victim: 3. What What is th thee av aver erag agee tr trea eatm tmen entt ef effe fect ct,, across the current literature, of bullying prevention programs on bystander empathy for the victim?
Method Search Strategy We used a comprehensive search to retrieve articles from the international research lite li tera ratu ture re wi with thin in th thee la last st 30 ye year arss (1 (198 980 0– 2010). We searched primarily five online databases: tabas es: Disse Dissertati rtation on Abstr Abstracts acts Inter Internatio national, nal, Educ Ed ucat atio ion n Re Reso sour urce cess In Info form rmat atio ion n Ce Cent nter er (ERIC), PsycINFO, Medline, and Science Direct re ct.. Co Comb mbin inat atio ions ns of th thee fo foll llow owin ing g te term rmss weree use wer used: d: “by “bysta stande nderr or par partic ticipa ipant nt or defender or other, ” “bully or victim, ” “school, school program, or program, ” “prevention or intervention, ” “aggression, ” and “not higher education or not cyber-bully.” To ensure that the ide identi ntified fied stu studie diess foc focuse used d on bys bystan tander der behavior as the primary goal, these terms were searched in the abstract of the study. In addi-
A Meta-Analysis of School-Based Bullying
tion on,, we se sear arch ched ed th thee bi bibl blio iogr grap aphi hies es of al alll ti articles selected for relevant studies. The search retrieved 360 total articles, but only 83 were unique and compared to the criteria listed in the next section. Of those 83, 53 were deem deemed ed irrelevant, irrelevant, 13 did not addre address ss
Furthermore, Furthermor e, we inclu included ded only studies that used a treatment-control research design. These designs included true experimental randomly assigned groups, nonrandom quasi-experi pe rime ment ntal al de desi sign gns, s, an and d no nonr nran ando doml mly y as as-signed matched group. We also included all
the intervene construct, 6 failed to include a control group, and 1 was a repeat of a previous study. Finally, we corresponded with a number of experts in the field to ensure inclusion of all relev relevant ant artic articles. les. This corre correspond spondence ence prod pr oduc uced ed 1 re rele leva vant nt st stud udy. y. He Henc nce, e, we re re-viewed and included 11 studies total.
control group types; these included wait list, treatmenttreat ment-as-us as-usual, ual, and “stra “straw-man w-man.” .” However, eve r, sin single gle-gr -group oup pre pre/po /postst-tes testt (e. (e.g., g., gai gain n scores) and cohort designs were excluded.
Criteria for Considering Studies for Review The pre presen sentt stu study dy foc focuse used d on sch school ool-based interventions that emphasized changing the bystander’s intervention behavior. To assess se ss th thee ef effe fect ctss of th thes esee pr prog ogra rams ms,, we co colllected peer-reviewed studies published or conducted from 1980 to 2010, based solely within a school system and intended purposefully to modify bystander intervention behavior. Subsequently, we excluded studies that focused on changi cha nging ng bul bullyi lying ng beh behavi aviors ors pri primar marily ily and collected a bystander measure only as a secondary procedure. The review included interventions from the United States and Europe, but we lim limite ited d inc inclus lusion ion to Eng Englis lish-w h-writ ritten ten studies. We reviewed studies that included participants from the kindergarten through 12thgrad gr adee po popu pula lati tion on,, bu butt in inte terv rven enti tion onss wi with th school-aged children based outside the school sett se ttin ing g we were re ex excl clud uded ed.. In ad addi diti tion on,, we at at-tempted to collect studies that included “atrisk” students and the general population, but none no ne of th thee st stud udie iess di dist stin ingu guis ishe hed d be betw twee een n these populations. It should be mentioned that one stu study dy att attemp empted ted to dec decons onstru truct ct the bystan st ande derr in into to se seve vera rall ty type pess of by byst stan ande ders rs to observe treatment effects (Evers et al., 2007). Although the deconstruction was informative, it wa wass th thee on only ly st stud udy y to im impl plem emen entt su such ch a procedure. As such, for that study we used the aver av erag agee in inte terv rven enti tion on ef effe fect ctss ac acro ross ss al alll by by-stander types.
Outcomes Studies must have included a bystander intervention measure. We operationalized this outcome as a measure that assessed the contribution of the bystander to a bullying situation (Frey et al., 2005, 2009). Therefore, we includ inc luded ed stu studie diess tha thatt mea measur sured ed int intent ention ion to interv int ervene ene,, int intent ention ion to sto stop p bul bullyi lying, ng, dir direct ect interv int ervent ention ion,, or con conver versel sely, y, dif difficu ficulty lty in respondi spo nding ng ass assert ertive ively ly to a bul bullyi lying ng sit situat uation ion.. For example, Andreou et al. (2008) included items that assessed students’ intention to intervene on a 5-point Likert scale. We included items ite ms tha thatt con concer cerned ned stu studen dents’ ts’ int intent ention ion to “seek teacher’s help,” “react against bullying,” and “support the victims of bullying” (p. 241). Tabl Ta blee 1 pr prov ovid ides es th thee me meas asur ures es us used ed,, th thee study’ stu dy’ss sta stated ted con constr struct uct,, and the num number ber of items combined to create the measure. In addition to the intervention outcome, we collected results of the program’s effects on changes in attitudes of empathy toward the victim. We operationalized the empathy outcome as a measure that indicated empathy for the victim. For instance, Stevens et al. (2000) used an empathy measure that included “feeling sad about students who are bullied” and “unpleasantness when another student is being bullied” (p. 26). Measures that included items similar to this construct constituted an empathy scale. We should also mention that this outc ou tcom omee co cons nsti titu tute ted d a se seco cond ndar ary y ou outc tcom omee measure and thus was not a criterion for synthesis the sis inc inclus lusion ion.. Stu Studie diess tha thatt inc includ luded ed a bystande sta nderr int interv ervent ention ion beh behavi avior or mea measur suree but failed to include a measure of empathy were included. 51
School Psychology Review, 2012, Volume 41, No. 1
) 9 . 5 1 , 1 . 0 1 . (
) 6 . 7 4 , 1 . 1 8 . (
d e t r o u s N a e M
) 9 . 1 1 , 0 . 0 2 . (
) 4 . 6 6 4 , . 2 7 . (
) 2 2 . , 5 1 0 . 1 . (
) 4 2 . , 1 2 1 . 0 . (
) 9 . 4 1 1 , . 1 0 . (
) 3 8 . , 1 2 2 . 4 . (
) 6 2 . , 3 1 0 . 2 . (
Q
Q
E
E
E
Q
Q
) I S C E % N 9 I 5 (
t n n o e i t m n e e r v u r e s a t n e I M
g n i d n l t u a f s p y l e b H
y t i l i b i r e s d n o n p a s t s e y r B
g n r i e d d n n e a f t e s d y B
s t i a r t r e d n a t s y B
e c e e r G
s e t a t e S l p i t l S u U M
s a s n a K
n o t g n i h s a W
d n a l n i F
o i h O
y l a t I
J
J
J
J
J
D
J
u ) 8 o 0 e r 0 d 2 n ( A
) 7 0 2 ( s r e v E
) y 9 0 g a 0 ( n 2 o F
) 5 0 2 ( y e r F
) 1 1 0 2 ( a n r a K
n i l h g ) 9 u a 0 0 2 L c ( M
i n ) 3 i 0 s e 0 2 n e ( M
s t i a r t r e d n e f e D d ) ) ; e ) r ) ) 2 ) 2 t r 2 r o 1 1 6 1 h ( r e t ( ( ( ( o e d a t h g t p t r r r r c i e n e u e e e e r l i s e h p h h h a R ) ) 2 c a c c c c e L 1 t 1 a ( m B s a a a a ( ( o o e F e e e e T C T T T R N r ; ; d g - i t o e e t g i n r n l g f c n n l v a e e l e t i s n n a r l e i a l n c S a w n a i h l e e d i d t d e y i l h t o c l ; e f P d t r o ; e i s ) k o i n o s a p m o i B n z o n ; u s i r i a ) E l s m a o ) S l s P R M M i o s l b ; i r t e b i s d e t r ; ; ; ; a i d o t ; d ; n a g c e e c s a s s s s n e c ) s g u t a n s n g s ; e o y o i o s s s c t t e ; t a i g r r d u t n t l e e i n e n t e m M e i n o i e r t i n fi d o i M r P r u c a c i n n a n a s a n t d n i e n d n v p e e o i p h i a e r n e d e l e e h e h o y o s w o e M d m r a h n w r i r a r t r p i M r e i a C r a a a A fl c a u a l a a o o m a e B n P r w ( r m ( h e I c P ( y s c A w e b w p P w w ( w c R t A B P A A A A y e t l l n i l n t c t o i u o - n e i g t T p B i m A g t n t s i n n c u d i e e n E l m d e e o e v v y a u R L t n e p i r l v r t c s l s e t e S e s i g i d a c a u e r r l r P p e o n r B B i f V R f P i r e i u u A e f t P C B C S K E B n n h h t h h h t h t t t t d e g 6 n n a 2 d e l a b a 6 d e b a n a 5 6 r 8 i r r 1 t x x a - b b u i i u r t r d d U r h h h t t r r h e b t G S 4 U h 6 U 3 R 3 u t 4 M t 9 M S 6 ) e l ) 6 ) ) ) a 8 ) 0 ) 8 ) 1 6 3 1 6 0 1 1 3 3 1 7 1 9 N M 1 0 1 5 4 4 4 6 7 4 5 4 9 5 2 5 ( ( ( ( ( ( ( 8 % ( n o i t a c o L e p y T ) P D o ( y d u t S
52
) 1 . 8 3 , 1 . 5 0 . (
) 1 . 9 0 , 1 . 8 3 . (
n g i s e D
s i s y l a n A a t e M e h t n i d e 1 s u e s l e b i a d T t u S e h t f o s c i t s i r e t c a r a h C
. M . N
) 7 0 . 3 2 . , 0 4 . (
) S I E C M % E 5 9 (
o t n e r o n e i e d t v n r n e e a t t t n i n s i y B
g n i d n a e t v s i y s s a b P
) s e u n i t n o c 1 e l b a T (
A Meta-Analysis of School-Based Bullying
) S I E C M % E 5 9 (
d e r t u o s a N e M
) 3 4 . , 9 5 2 . 1 . (
) 8 1 . , 5 8 0 . 2 . (
) 3 3 . , 4 6 0 1 . . (
d e r t u o s a N e M
) I S C E % N 9 I 5 (
) 6 3 . 2 , 0 7 6 . 1 . (
) 7 5 . , 3 9 4 . 2 ( .
) 9 2 . , 6 7 0 1 . ( .
) 9 5 . , 9 9 3 . 1 ( .
) 4 3 5 . , 2 . 5 0 1 ( .
E
E
E
E
E
n g i s e D t n n e o i t m n e e r v u r e s a t n e I M
d e u n i t n o C 1 e l b a T
; r ) o h t a t g t i n l i e c L a ( F s c i t s m i r a r t e g c o r a P r a h C e l t i T m a r g o r P d e g n i a t r t e G S ) e l a N M % ( n o i t a c o L e p y T ) P D o ( y d u t S
o o o o o t t t t t e e e e n n n n n e r r r r r n n n n o o o o o n e t e e e e e e e e e i v d t i v d t i v d t i t d i v d v n r n r n r n r n r n n n n n e e e e e e e e e e a a a a a t t t t t t t t t t t t t t t n i n s i n i n s i n i n s i n i n s i n i n s i y y y y y B B B B B r e h c r a ) e 2 s e ( R d e M d o M ; s s e n e r a w A
r e h c r a ) e 1 s e ( R d e M d o M ; s s e n e r a w A o o e t d i h c g V m a i n g a o r l y i n r g s p ’ y o p l u l r W A B l u P 5 B n a r 9 b U
h t
2 d e 1 x - i 9 M
h t
h t
) 1 ( r e h c a T e
) 1 ( r e h c a T e
d e M e d v o i t i M n ; g g o n c - i n l i a a i r c t o S
d e M e d v o i t i M n ; g g o n c - i n l i a a i r c t o S
n g i o n t i y n l l e u v r b e - t i n t I n A
n g i o n t i y n l l e u v r b e - t i n t I n A
) 2 1 ( r e h c a T e o t n h c e r a y s P P ; ; n s o s t e i a n c e r u a d w e A t c e p s e R t c e p x E
d d t t r e 1 t r e d e 8 o t x - o o 1 o 5 i - N p N p e e 4 M R 9 R
h t h t
h t
h t
h t
) 6 0 5 3 (
5 ) 2 2 8 4 (
) 1 * 0 0 3 5 (
) 1 * 0 0 4 5 (
3 ) 6 0 7 5 1 (
k r o Y w e N
a i n a v l y s n n e P
m u i g l e B
m u i g l e B
s a x e T
D
M
J
J
B
”
) 4 l l 0 e 0 r 2 r e ( M
r e h c 7 a ) m 0 u 0 2 ( h c S
” B “ 0 s ) 0 n e 0 2 v e ( t S
r e ) 4 k 0 a t i 0 ( h 2 W
A “ 0 s ) 0 n e 0 2 v e ( t S
t e h f t f g E n e y L h ; t a g p n i m n E i a r T t S n E e r P a E M ; e z t i n S e r a t c e P f ; f e E l a n m i o t t n n e e c v r r e t p e n I d e t u p S m E i N I * ; s ; p r e t u o p a r g h l C o r k o t o o n B C c
B C ; s ; s i p s e u h o T r g s t ’ n r e e t m s a t a M e r T
M ; T n l o ; i a t a t n t e r e m s i s i r D e p x E D ; e E l c l ; i t a r t A n e l a i m n r r u e o p J x e i J s ; a u n Q o i t a c i l Q b ; u s P h f t o n o e m t a n D i t s e t P t s D o P o o t : s e . e e t m i z o i T S N
53
School Psychology Review, 2012, Volume 41, No. 1
As shown in Table 1, 10 of the 83 unique articles met the inclusion criteria and were included in the meta-analysis. We discontin con tinued ued the lit litera eratur turee sea search rch on May 20, 2010, but added 1 article brought to our attention 3 months later, which brought the total
Independentt findin Independen findings. gs. A par paramo amount unt meta-analytic assumption is independence of findings. Cooper (2010) discussed several occurrences that constitute nonindependence and their effects on subsequent findings. Thus, we conducted several common procedures to en-
number to 11 studies.
sure independent findings. Tofrom ensure thatinteronly one effect size was derived each vention, we used only the first treatment outcome reported for studies that reported multiple post-treatment outcomes (see Andreou et al.,, 200 al. 2008). 8). If stu studie diess imp implem lement ented ed int interv ervenention ti onss wi with th tw two o gr grou oups ps bu butt on only ly on onee co cont ntro roll group, gro up, we syn synthe thesiz sized ed the tre treatm atment ent eff effect ectss prior to calculating the study effect size (see McLaughlin, 2009). Finally, if one author implemented an intervention and published multipl ti plee ar arti ticl cles es on th thee sa same me sa samp mple le,, th then en we reviewed only the first article published (see Frey et al., 2005, 2009).
Coding Study details, appropriate program, and sample information were coded directly into an EXC EXCEL EL (20 (2010) 10) dat databa abase. se. Thi Thiss inc includ luded ed public pub licati ation on yea year, r, pub public licati ation on typ type, e, fun fundin ding g provided, provi ded, country of origi origin n and publi publicatio cation, n, program location, treatment and control sample cha charac racter terist istics ics (e. (e.g., g., age age,, gen gender der,, rac race, e, SES,, dis SES disabi abilit lities ies), ), pro progra gram m cha charac racter terist istics ics (e.g., length of time, intervention details), and progr program am athy facilitato facil r. Ine addit addition, ion, interventio n and emp empath y itator. outcom out come measur mea sures es interv wereeention wer trantra nscribed. By coding directly into the EXCEL database, we eliminated errors that might have occurred during the normal transcription phase (Lipsey & Wilson, 2001). The first author coded all 12 studies, but one independent rater coded a randomly selected portion of studies (5) for reliability purposes. The two raters agreed 92% of the time. The coders came to an agreement for all discrepancies prior to completion.
Analysis Quantitative synthesis, or meta-analysis, is a statistical technique that combines related research studies to estimate an overall treatmentt eff men effect ect (Co (Coope oper, r, Hed Hedges ges,, & Val Valent entine ine,, 2009; 200 9; Gla Glass, ss, 197 1976; 6; Hed Hedges ges & Olk Olkin, in, 198 1985). 5). Often, and in the case of the present review, meta-a met a-anal nalysi ysiss agg aggreg regate atess tre treatm atment ent eff effect ect sizes to assess an intervention’s effectiveness. Thee pu Th purp rpos osee of a me meta ta-a -ana naly lysi sis, s, th then en,, is to generalize findings across multiple treatment and setting types, participants, and times (Matt & Cook, 2009). conducted analyses using SPSS (2010) andWe Comprehensive Meta-Analysiss (Bo ysi (Boren renste stein, in, Hed Hedges ges,, Hig Higgin gins, s, & Rot Rothhstein, 2005) software. 54
Effect siz Effect size e met metric rics. s. The The maj majori ority ty of effect sizes calculated used a continuous scale. As such, the appropriate effect size metric was the stand standardiz ardized ed mean diffe difference rence (Equation 1):
d
X G1 X G2
(1)) (1
S p
where the numerator is the mean difference en ce be betw twee een n tr trea eatm tmen entt an and d co cont ntro roll gr grou oup p posttests, and the denominator is the pooled standa sta ndard rd dev deviat iation ion for the int interv ervent ention ion and comparison groups. Further, all d metrics metrics were biass cor bia correc rected ted usi using ng Hed Hedges ges’s ’s (19 (1981) 81) sma small ll sample correction (g). This correction as well as th thee sa samp mpli ling ng va vari rian ance ce is re repr pres esen ente ted d by Equations 2 and 3:
g 1
Var g
3
4 N 9
nG1 nG2 nG1nG2
*d
g2
2nG1 nG2
(2)) (2
(3 (3))
where N is the total sam sample ple,, d is th thee original standardized mean difference, and nG1
A Meta-Analysis of School-Based Bullying
representt the treat treatment ment and contr control ol and nG2 represen group sample sizes, respectively. In addition, we calculated logged odds ratio effect sizes for two studies that used a catego cat egoric rical al out outcom comee mea measur suree (Ev (Evers ers et al. al.,, 2007; Merrell, 2004). Both measures were ob-
was mos mostt app approp ropria riate. te. We fur furthe therr ass assume umed d that an underlying distribution of effect sizes was plausible; thus our goal, given the random effects framework, was to estimate the distribution’s mean and confidence interval. To estimate a random effects mean and
ser servat vation sing of fre treatm tre ent contro con troll aut partic par ticii-s pants pan ts ions during dur free eatment period per iodand times. tim es. The author hors observed how many times treatment children intervened (or intended to intervene) in a bullying lyi ng sit situat uation ion com compar pared ed to chi childr ldren en in the control group. Standard odds ratio calculations were first used (Sanchez-Meca et al., 2003). We then converted the logged odds ratio into a standardized mean difference as outlined by Lipsey & Wilson (2001) (Equations 4 and 5):
co con nfid fiden ence cetreatment intter in erva val, l, weof each calc ca lcul ulat ateed The thee th weighted effect study. weighted weigh ted effec effectt estim estimation ation synth synthesize esized d both within-study error variance and common betweentwe en-stu study dy var varian iance. ce. We rep repres resent ented ed thi thiss weight calculation as Equation 7:
d
SE
3 * lnOR
3 * SE 2lnOR 3
(4 )
(5 )
where ln( where ln(OR) OR) rep repres resent entss the ori origin ginal al logged odds ratio and SEln(OR) represents the original sampling variance.
Missing data. Only one of the studies failed to provide appropriate descriptive statistics. Schumacher (2007) provided only a t statis sta tistic tic,, as wel welll as sam sample ple siz sizes, es, that com com-pared the treatment and control groups. Lipsey and Wil Wilson son (20 (2001) 01) pro provid vided ed an app approp ropria riate te conversion (Equation 6):
W i
1
V i T 2
(7)) (7
where W represents represents the i th study weight, V i indicates indicates the withi within-stu n-study dy error varia variance, nce, 2 and T represents the between-study variance. We used these weights then to estimate the combined treatment effect. This can be represented by Equation 8:
M
W * g W i
i
(8)) (8
i
where M represents represents the combined effect, W i represents represents the ith st stud udy y we weig ight ht,, an and d gi indicates indic ates Hedge Hedges’s s’s effec effectt size g for for th thee ith study. Further, we calculated confidence intervals and p values by taking the square root of the inverse of the sum of the weights.
Random effects model. We assumed that the treatments were derived from a ran-
Moderator analy Moderator analysis. sis. A cr crit itic ical al ne next xt step to the investigation of effect size distribution but ion is mod modera erator tor ana analys lysis. is. We sta starte rted d by calculating the homogeneity statistic Q. This statistic provided information about the distribution of effect sizes, and a large test statistic (i.e., (i. e., rej reject ecting ing the nul nulll hyp hypoth othesi esiss of stu study dy homogeneity) indicated that moderator analyses wer weree app approp ropria riate te (Ra (Raude udenbu nbush, sh, 200 2009). 9). Given Giv en thi thiss sta statis tistic tical al con confirm firmati ation, on, we use used d procedures analogous to analysis of variance (ANOVA), where one attempts to model ef-
dom of theGiven literature lacked a common sample effect size. this but assumption, Borens re nste tein in,, He Hedg dges es,, Hi Higg ggin ins, s, an and d Ro Roth thst stei ein n (2010) posited that the random effects model
fectfect-size sizestudy-level heterogene heter ogeneity ity assoc associated iated with categorical variables. Further, because of the small number of studies that constituted the review, we calculated the variance compo-
ES sm t
n1 n2 n1n2
(6 )
where t represent representss the t statisti statisticc the study provided, n1 represents the sample size of the treatment group, and n2 represents the sample size of the control group.
55
School Psychology Review, 2012, Volume 41, No. 1
nent nt ac acro ross ss al alll gr grou oups ps ra rath ther er th than an wi with thin in ne groups as is generally conducted (Hedges & Vevea, 1998). Lipsey (2009) discussed three types of independen indep endentt varia variables bles comm common on to metameta-anaanalytic practice: extrinsic variables, method vari-
To avoid and interpret the overestimation of the random-effects estimate, we used the nonparametric trim and fill procedure (Duval & Tweedie, 2000) to assess the sensitivity of results to publication bias. This procedure estimates the number of publications theoret-
ables, and substantive variables. Extrinsic and method variables relate to the study’s disseminat in atio ion n (i (i.e .e., ., pu publ blis ishe hed d or un unpu publ blis ishe hed) d) or methodolog metho dological ical const constraint raintss (i.e (i.e., ., rando randomized mized or nonrandomized). Substantive variables, on the other hand, should be regarded as variables of interest and generally include characteristics of the population or treatment. For this review rev iew,, we cod coded ed sub substa stanti ntive ve ind indepe epende ndent nt variables to reflect the participant’s age, length of treatment, and treatment type (e.g., individual or group). Finally, we conducted a moderator anal-
ically missing because of funnel plot asymmetry, and then recalculates the random-effects mean and confidence interval to include the impute imp uted d mis missin sing g stu studie dies. s. We als also o rep report orted ed Rose Ro sent ntha hal’ l’ss fa fail il-s -saf afee N (1 (197 979) 9) an and d Eg Egge ger, r, Davey Smith, Schnieder, & Minder’s (1997) regression coefficient.
ys ysis is pr proc oced edur uree an anal alog ogou ous s to regr re gres essi sion on,, weighted meta-regression. This statistical procedure allows for the simultaneous estimation of study-level effects, but shares the problems of typical regression (Cooper, 2010). For this review rev iew,, we mod modele eled d two ind indepe epende ndent nt var variiables abl es sim simult ultane aneous ously, ly, tre treatm atment ent pop popula ulatio tion n (categorical variable) and the percent of males in the treatment group.
viewed includeand 7 published journal articles, 1 book chapter, 3 unpublished papers (two dissertations and one master’s thesis). Seven of the 11 studies were conducted within the United States and the other 4 were conducted in Belgium, Finland, Greece, and Italy, respectively tiv ely.. All stu studie diess wer weree com comple pleted ted bet betwee ween n 2000 and 2010. One article contributed two effect sizes becaus bec ausee it inc includ luded ed two mut mutual ually ly exc exclus lusive ive interventi inter ventions ons from two separ separate ate popul population ationss (Stevens et al., 2000). Therefore, we synthesize si zed d a to tota tall of 12 in inte terv rven enti tion ons. s. Ea Each ch of the 12 interventions included a treatment and control group; 4 of the 12 programs used quasi-experimental design and the other 8 used a rand ra ndom omiz ized ed ex expe peri rime ment ntal al de desi sign gn.. A to tota tall of 12 12,8 ,874 74 st stud uden ents ts pa part rtic icip ipat ated ed in th thee 12 interventions.
Sensitivity Sensitivit y analy analysis. sis. Two Two of th thee re re-view vi ewed ed st stud udie iess co cont ntri ribu bute ted d a me meas asur uree th that at used an odds ratio. To ensure that study findings were not biased by including these measures, we conducted a sensitivity analysis. The analysis consisted of removing the 2 studies (Evers et al., 2007; Merrell, 2004) and recalculating the weighted effect size. We hypothesized that no difference would be found between the two types of measurements.
Publication bias. Publication bias remained an important consideration during the lite li tera ratu ture re se sear arch ch an and d an anal alys yses es.. Ro Rose sent ntha hall (1979) introduced the “file-drawer problem,” which stated that studies with small or nonsignificant effect sizes tended to remain unpublished. Toed combat problem, included unpubl unp ublish ished works wor ksthis from fro m thr three ee we disser dis sertat tation ionss and one master’s thesis (Rothstein, Sutton, & Borenstein, 2005). 56
Results Meta-Analysis Literature Table 1 provides characteristics for each study included in the review. The studies re-
Outcome Effect Sizes Bystander intervention outcome. As delineated in Lipsey and Wilson (2001), we estimated the random effects weighted mean by using Equation 8. The results revealed a statistically significant positive weighted aver
p age (g er .20, .001, 95% CI inc .11 to .29). In oth other words, wor ds, the treatm tre atment ent increa reased sed bystander intervention behavior 20% of one standard dar d dev deviat iation ion mor moree tha than n ind indivi ividua duals ls in the
A Meta-Analysis of School-Based Bullying
Table 2 Summary Statistics for Bystander Intervention Effect Sizes Study Name
Standard Error
Hedges’ g
Andreou Evers Fonagy Fre y Karna McLaughlin Mensini Merrell Schumacher Stevens “A” Stevens “B” Whitaker Overall
.01 .46 .05 .11 .14 .21 .03 .60 .43 .06 .39 .25 .20
.10 .09 .08 .07 .02 .32 .12 .90 .07 .12 .10 .05 .04
Lower Limit
Empathy outco Empathy outcome. me. Of th thee 12 in inte terrventions used to calculate the bystander intervention outcome weighted average, 8 included a measure on victim empathy. As previously conducted, we used a random effects model to estima est imate te the wei weight ghted ed tre treatm atment ent mea mean. n. The results revealed a very small, nonstatistically significant signi ficant result (g .05, p .38) with a confide con fidence nce int interv erval al tha thatt inc includ luded ed zer zero o (95 (95% % CI .07 .0 7 to .1 .17) 7).. Th Thee sm smal alll nu numb mber er of
Z-Value
.19 .64 .22 .24 .19 .83 .26 2.24 .57 .29 .59 .34 .28
control group. Table 2 provided a forest plot of the random-effects model’s relevant statistics.
.20 .27 .11 .02 .10 .42 .21 1.17 .29 .17 .19 .15 .11
Upper Limit
.04 4.88 .62 1.68 6.51 .65 .22 .66 6.08 .53 3.89 5.16 4.54
p-Value
.97 .01 .54 .09 .01 .52 .82 .51 .01 .60 .01 .01 .01
studiess tha studie thatt inc includ luded ed a mea measur suree of emp empath athy y may ma y no nott pr prov ovid idee en enou ough gh po powe werr to de dete tect ct a small effect. Therefore, the results of this analysiss sho ysi should uld be con consid sidere ered d inc inconc onclus lusive ive (se (seee Table 3).
Moderator Analysis We categorized the effect sizes into severall rel era releva evant nt gro groups ups and con conduc ducted ted the ran ran-dom-effects ANOVA-like analysis (Table 4). A statis statistical tically ly signi significant ficant Q value indic indicated ated appropriate heterogeneity between studies that
Table 3 Summary Statistics for Empathy Effect Sizes Study Name Andreou Fonagy Fre y Karna McLaughlin Schumacher Stevens “A” Stevens “B” Overall
Hedges’ g
.19 .23 .18 .15 .17 .29 .05 .14 .05
Standard Error .10 .09 .07 .02 .32 .07 .12 .10 .06
Lower Limit
Upper Limit
.38 .40 .05 .10 .80 .15 .28 .06 .07
.01 .07 .31 .19 .45 .43 .18 .33 .17
Z-Value 1.90 2.75 2.72 6.62 .05 4.13 .40 1.35 .74
p-Value
.06 .01 .01 .01 .59 .01 .69 .18 .46
57
School Psychology Review, 2012, Volume 41, No. 1
Table 4 Moderator & Meta-regression Analysis of Bystander Intervention Effect Sizes Moderator Population 3rd–8th Grade 9th–12th Grade Location United States Europe Treatment Treatm ent length 1–2 months 6–12 months Parent Component Yes No Facilitatora Teacher Other
Assignment Non-Random Random Publication Publica tion type Peer-review Non-peer review Meta-Regression Percent Perce nt Male HS (1 Yes)
K
Hedges’ g
8 4
.14 .43
.11, .18 .33, .52
7 5
.26 .13
.14, .38 .01, .26
.17
5 7
.31 .16
.16, .45 .06, .25
.09
5 7
.19 .20
.08, .31 .07, .34
.92
7 4
.15 .43
.09, .22 .30, .56
.01
4 8
.17 .21
8 4
.16 .32
.01 .25
S..E. S .009 .13
95% C.I.
p-value
.04
.01, .35 .10, .31
.74
.06, .25 .17, .48
.07
Z-score 1.04 1.98
p-value .15 .02
p .05; a-Menesini et al. (2003) did not indicate the facilitator; HS High School sample.
measured the bystander intervention construct grams located in the United States (US) did (Q 39.81, df 11, p .001; I 2 72.36). not differ significantly from those located in However, because of the relatively small num- Europe (EU; US ES 0.26, EU ES 0.13, ber of studies that included an empathy mea- p .17). sure, we chose not to condu conduct ct moderator analAnothe Ano therr sub substa stanti ntive ve mod modera erator tor,, tre treatatyses to protect against findings of chance. mentt len men length gth,, fai failed led to pro produc ducee sig signifi nifican cantly tly The results of these analyses were de- gr greeat ater er tr trea eatm tmen entt eff ffec ects ts (1 (1– –2 mo mon nth thss composed into substantive and methodologi- ES 0.31, 6–12 months ES 0.16; p .09). cal characteristics. Samples that consisted of Similarly, treatments that included a parental high school students only generated a signifi- component (e.g., parent guides, parent training cantly can tly gre greate aterr tre treatm atment ent eff effect ect (ES 0.43, sessions) failed to influence the treatment sigCI .33 .33 to .5 .52) 2) co comp mpar ared ed to sa samp mple less of nificantly compared to programs without the primary schools only (ES 0.14, CI .11 to .18). We further assessed sample differences by conducting an analysis to evaluate location diff di ffer eren ence ces. s. Th Thee re resu sult ltss re reve veal aled ed th that at pr proo58
component (parent included ES 0.19, parent excluded ES 0.20, p .92). We al also so gr grou oupe ped d th thee st stud udie iess by wh who o facilitated the treatment programs. The teach-
A Meta-Analysis of School-Based Bullying
ers implemented a significant portion of the programs for 7 of the 12 programs. Four other programs were facilitated by the researcher, a counselor, or in one case (Evers et al., 2007), with computer software. One study failed to indicate who implemented the program (Me-
To ensure that the overall effect size was not upw upward ardly ly bia biased sed by inc includ luding ing dif differ ferent ent measur mea sureme ements nts,, we con conduc ducted ted a sen sensit sitivi ivity ty analysis removing studies that used a dichotomous omo us out outcom come. e. The res result ultss of the ana analys lysis is found fou nd tha thatt the ove overal ralll wei weight ghted ed eff effect ect siz sizee
nesini,, Cod nesini Codeca ecasa, sa, Ben Benell elli, i, & Cow Cowie, ie, 200 2003). 3). The results of this moderator analysis revealed significantly greater treatment effects for programs that implemented the program with facili ci lita tato tors rs ot othe herr th than an th thee te teac ache herr (t (tea each cher er ES 0.15, other ES 0.43, p .01). However, serious caution should be given to this finding because two of the four programs that used researchers as facilitators had the smallest sample sizes, and therefore this could be a reflection of imprecision or biased effects becaus ca usee of sm smal alll sa samp mple less (L (Lev evin ine, e, As Asad ada, a, & Carpenter, 2009).
0.20, decreased (original ESeffect modified ES slightly 0.18). The overall remained statistically and practically significant.
Finally, we conducted moderator analyses with two metho methodolog dological ical groupings. groupings. We first observed mean group differences between randomly assigned (RA) and nonrandomly assigned treatments (NRA) groups. The results indicated that there were no statistically significan nifi cantt dif differ ferenc ences es (RA ES 0.21, 0.21, NRA ES 0.17, p .74). We also estimated group difference diffe rencess betwe between en peerpeer-revie reviewed wed (PR) and nonpeer-reviewed studies (NPR). The results of this calculation calculation revea revealed led that nonpeer-renonpeer-reviewed studies did not produce a greater treatment effect (PR ES 0.16, NPR ES 0.32,
cally significant (see Figure 1). We also used Rosenthal’s fail-safe N procedure; the results of thi thiss cal calcul culati ations ons ind indica icated ted tha thatt 236 nul nulll studies would be required to result in a nonsignificant signi ficant findin finding. g. Egger Egger’s ’s regre regression ssion inter inter-cept coefficient calculation also produced nonsignificant results (0 0.57, p .26). Taken together, we concluded that the review’s results were not affected significantly by publication bias.
p .07).
In addition to the ANOVA-like modeling, we conducted a weighted regression analyses. This analysis allowed us to estimate the effects effec ts of sever several al predi predictors ctors simul simultaneo taneously. usly. We hypothesized that the percentage of males in the tre treatm atment ent int interv ervent ention ionss and pro progra grams ms conducted in a high school would be significant ca ntly ly re rela late ted d to th thee tr trea eatm tmen entt ef effe fect ct (i (i.e .e., ., Hedges’s g). The results of this analysis again revealed that, after controlling for the percentagee of ma ag male less in th thee tr trea eatm tmen entt gr grou oup, p, hi high gh school samples produced a greater treatment
sis was to exa examin minee the treatmen treatmentt eff effect ectss of bullying prevention programs on bystander intervention terve ntion behavior. behavior. Empat Empathy hy for the victi victim m was also synthesized as a secondary outcome, but was not of primary purpose for the current review. In total, we reviewed 11 studies (12 effect sizes) from the United States and Europe that included 12,874 children. Using meta-analytic techniques, the results revealed that the intervention behavior of bystan bys tander derss inc increa reased sed (i. (i.e., e., bys bystan tander derss ind indiicated greater intervention behavior in bullying situ si tuat atio ions ns)) co comp mpar ared ed to co cont ntro roll gr grou oups ps (g
ef effe fect ctl co comp mpar ared edons to (mi midd or el elem tary ry ddle le enta school schoo interventi inter ventions 0.25, Z emen 1.98, p .05). These results bolstered the previous findings.
.2 .20) 0).. Th Thee re resu sult ltss of a se seco cond ndar ary y an anal alys ysis is reveal rev ealed ed tha thatt int interv ervent ention ion pro progra grams ms did not have a similar effect on empathy for the victim (g .05), but this finding should be viewed as
Publication Bias
We applied Duval and Tweedie’s (2000) trim and fill procedure to address publication bias. This procedure revealed that one negativee res tiv result ult was mis missin sing g fro from m the bys bystan tander der intervention outcomes. However, the imputed missing values would only slightly change the overal ove ralll fixe fixed d eff effect ect siz sizee (ES 0.20); 0.20); more important, it remained practically and statisti-
Discussion
The purpose of this quantitative synthe-
59
School Psychology Review, 2012, Volume 41, No. 1
Funnel Plot of Standard Error by Hedges’s g 0.0
0.2
r o r r E d r a d n a t S
0.4
0.6
0.8
1.0 -2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
Hedges’s g
Figure Figur e 1. Tri Trim m and fill fu funne nnell pl plot ot of int inter erve vent ntio ion n ef effe fect ct si size zes. s. Thi Thiss fig figure ure illustrates illustra tes each effect size relative to its standar standard d error; the shaded dot indicates an imputed effect size. inconclusive because of the small number of studies that reported this outcome and its secondary nature. These overall results mirrored the findings of a previously conducted small synthesis. Merrell et al. (2008) synthesized bullying prevention programs to investigate the effects of the programs on bullying perpetration, but also al so in incl clud uded ed se seve vera rall se seco cond ndar ary y me meas asur ures es
aged children (Williford et al., 2011). These results may indicate that bystander intervention behavior is a developmental process and programs may not influence younger students as intended. The purpose of meta-analytic research is to ge gene nera rali lize ze fin findi ding ngss ac acro ross ss po popu pula lati tion ons, s, treatm tre atment ents, s, out outcom comes, es, and des design ignss (Ma (Matt tt & Cook, 2009). Altho Although ugh this synthesis aggre-
(e.g., bystander intervention behavior, empathy). thy ). The aut author horss rep report orted ed sma small ll but sig signifi nifi-cant treatment effects for bystander intervention behavior (k 3, g .17) and nonsignificant negative effects with regard to empathy (k 3, g .10). .10 ). Tak Taken en tog togeth ether, er, the these se replications results provided evidence against mono-operation bias and thus greater validity (Shadish, Cook, & Campbell, 2002). Moderator Moder ator analyses also revea revealed led several er al fin findi ding ngss of in inte tere rest st.. Re Resu sult ltss of bo both th ANOVA-like and weighted-regression analyses rev reveal ealed ed tha thatt the tre treatm atment ent eff effect ectss wer weree
gated a smaller number of studies, its findings rendered generalizability of bullying prevention program’s effects on the bystander intervention construct. A few factors bolster this belief. The populations assessed varied across ages, age s, loc locati ations ons,, and tre treatm atment ents. s. The lar larges gestt effects were found for high school only samples; however, an overall significant treatment effe ef fect ct wa wass al also so fo foun und d fo forr th thee to tota tall sa samp mple le.. With regard to location, no significant differences were found between U.S. and European samples. In addition, the studies employed a
greater for high school only samples. This is somewhat somew hat surpr surprising ising because some schol scholars ars have postulated that bullying prevention programs are more effective for middle school-
wide va wide vari riet ety y of tr trea eatm tmen entt pr prog ogra rams ms th that at proved prove d effica efficacious cious.. Moder Moderator ator analy analyses ses also suggested that study design, publication type, and par parent ental al com compon ponent entss pro produc duced ed sim simila ilarr
60
A Meta-Analysis of School-Based Bullying
findings across studies. These results constitute tu ted d a te test st of ef effe fect ctss ho hold ldin ing g “a “acr cros osss pr preesumed irrelevancies” (Shadish et al., 2002, p. 455) and increases the findings’ external validity. Further investigation and evaluation is certai cer tainly nly req requir uired, ed, but the these se res result ultss sho should uld
variable. This becomes especially clear with regard reg ard to the stu studie diess gro groupe uped d by tre treatm atment ent length. As alluded to previously, a majority of the studies grouped in the 1–2 month category were the smallest of studies. Smaller studies tend to produce larger and more unstable ef-
cautiously suggest program generalizability.
fect sizes (Levine al., 2009), and therefore the grouping couldetreflect this phenomenon. Fourth, although we made efforts to collectt all pri lec primar mary y stu studie diess tha thatt foc focuse used d on bystander intervention behavior, it is quite possible that studies failed to be included. New material mater ial publi published shed posts postsearch earch,, missp misspecifie ecified d search terms, or simple human error could all cause inadvertent omission of extant literature. As such, we must temper our inferences with regard to the extrapolation of this information.
Limitations Severa Seve rall li limi mita tati tion onss sh shou ould ld be no note ted. d. First, this meta-analysis included only 11 studiess an ie and d 12 ef effe fect ct si size zes. s. Al Alth thou ough gh we to took ok precautions to ensure unbiased effect sizes and findings, a great deal of caution should be used when whe n int interp erpret reting ing the find finding ings. s. The find finding ingss from a small collection of studies, no matter the statistical statistical techn technique ique or numb number er of stude students nts surveyed, should not enact immediate policy and pra practi ctical cal cha change nges. s. Thi Thiss bec become omess esp espeecially clear when one considers the effect of studie stu diess acr across oss tim time. e. Rec Recent ent res resear earch ch in the field of meta-analysis publication bias has indicated that as programs increase in size and fidelity, effect sizes tend to decrease (Trikalinos & Ioannidis, 2005). Therefore, we plan to update these results periodically to observe the effect of time. Second, because of the nature of metaanalys ana lyses, es, cau causal sal inf infere erence ncess sho should uld be sta stated ted cautio cau tiousl usly. y. Qua Quanti ntitat tative ive met meta-a a-anal nalysi ysis, s, although thoug h stati statistica stically lly sophi sophistica sticated ted and impor impor-tant ta nt,, re rema main inss es esse sent ntia iall lly y an ob obse serv rvat atio iona nall
Implications for Future Policy and Practice This me This meta ta-a -ana naly lysi siss sh shou ould ld he help lp ca cauutiou ti ousl sly y to sh shif iftt th thee em emph phas asis is of po poli licy cy an and d practice. The results of this meta-analysis revealed two implications for policy. First, state and national bullying legislation should implement and evaluate programs that address bullying behaviors as a group process. Prevention fram fr amew ewor orks ks an and d pr prog ogra rams ms th that at at atte temp mptt to abate bullying within schools are increasingly emphas emp hasizi izing ng cha change ngess in sch school ool cli climat matee tha thatt desist reinforcing bystander behavior or bul-
study stud y (L (Lip ipse sey y & Wi Wils lson on,, 20 2001 01;; Co Coop oper er & Hedges, 2009). On the other hand, the sample of studies we synthesized contained only those that used a treatment and control group, and these research designs constitute the most efficient measure of treatment effect (Shadish et al.,, 200 al. 2002). 2). The Theref refore ore,, the syn synthe thesis sis of the these se results should partially reflect the nature of the primary studies. Thir Th ird, d, be beca caus usee of th thee lo low w nu numb mber er of studies per factor, the ANOVA-like moderator analys ana lyses es gen genera erated ted rel relati ativel vely y low sta statis tistic tical al power pow er and the res result ultss sho should uld be int interp erpret reted ed
lying perpetration (Cohen, 2006). The results of th this is st stud udy y su supp ppor ortt th thes esee ef effo fort rtss to ra rais isee awareness about the participant roles, to encourage active and prosocial behavior, and to provide opportunities to role-play and practice bystander intervention in vivo. Second, the results of this meta-analysis reveal rev ealed ed tha thatt bul bullyi lying ng pre preven ventio tion n pro progra grams ms might mig ht be eff effect ective ive at enc encour ouragi aging ng pre presoc social ial bystander bysta nder inter interventi vention on when the frame framework, work, program, progr am, and/o and/orr curri curriculum culum expli explicitly citly targe targett bystander bysta nder attitudes and beha behaviors viors.. It is simpl simply y nott su no suffi ffici cien entt to on only ly de defin finee pr pros osoc ocia iall by by--
with caution (Hedges & Pigott, 2004). Moreover ov er,, th thee lo low w nu numb mber er of st stud udie iess pe perr gr grou oup p coul co uld d ea easi sily ly ca capi pita tali lize ze on ch chan ance ce,, or th thee groupings may reflect some other unforeseen
stande stan derr be beha havi vior ors, s, su such ch as wa walk lk aw away ay,, ge gett help, or stand up to those engaged in bullying. Policy Pol icy mus mustt enc encour ourage age the ado adopti ption on of pro pro-grams gra ms and int interv ervent ention ionss tha thatt shi shift ft att attitu itudes des 61
School Psychology Review, 2012, Volume 41, No. 1
supportive of intervention (willingness to interven ter vene) e) and beh behavi aviors ors thr throug ough h a con consis sisten tentt message about intervention and ample support from adults and administrators. Future Research
This research suggests important future projects. First, primary research should focus on designing programs, implementing change, and mea measur suring ing the bys bystan tander der con constr struct uct.. As mentioned previously, researchers should focus on changing the behavior of the bystander. Second, further work is required to evaluate the effects of bystander behavior on bullying. This meta-analysis merely demonstrated that explicitly stated bystander programs have the ability to increase bystander intervention behavior hav ior.. How Howeve ever, r, fut future ure res resear earch ch mus mustt con con-tinu ti nuee to as asse sess ss ho how w by byst stan ande ders rs im impl plem emen entt
Cohen, J. (2006). Social, emotional, ethical, and academic education: Creating a climate for learning, participation in democ democracy, racy, and wellwell-being being.. Harvard Educa Educa-tional Review, 76, 201–237. 76, 201–237. Cooper, H. (2010). Research (2010). Research synthesis and meta-analysis (4th ed.). Thousand Oaks, CA: Sage Publications. Cooper, H., & Hedges, L.V. (2009). Potentials and limitations. In H. Cooper, L.V. Hedges, & J. C. Valentine (Eds.), The (Eds.), The handbook of research synthesis and meta (pp. 562–571). New York: Sage Publication. analysis (pp. analysis Cooper, H., Hedges, L. V., & Valentine, J. C. (2009). The handbook of research synthesis (2nd synthesis (2nd ed.). New York: Russell Sage Foundation. Cowie, Cow ie, H. (20 (2000) 00).. Bys Bystan tandin ding g or sta standi nding ng by: Gen Gender der issues iss ues in cop coping ing wit with h bul bullyi lying ng in Eng Englis lish h sch school ools. s. Aggressive Behavior, 26, 26, 85–97. 85–97. Cowie, H., & Hutson, N. (2006). Peer support: A strategy to help bystanders challenge school bullying. Pastoral bullying. Pastoral Care in Education, 23(2), 23(2), 40–44. Craig, W. M. (1994). The relationship among bullying, victimizat victi mization, ion, depre depression ssion,, anxie anxiety, ty, and aggre aggression ssion in elementary eleme ntary schoo schooll childr children. en. Persona Personall Indiv Individual idual Dif ferences, 24 24(1), (1), 123. Craig, W. M., Pepler, D., & Atlas, R. (2000). Observations of bullying in the playground and in the classroom. School Psychology International, 21(2), room. 21(2), 22–36.
*Andreou, *Andre ou, E., Did Didask askalo alou, u, E., & Vla Vlacho chou, u, A. (20 (2008) 08).. Outcomes Outco mes of a curri curriculum culum-base -based d anti-b anti-bullyin ullying g inter inter-vention program on students’ attitudes and behavior. Emotional & Behavioural Difficulties, 13, 13, 235–248. 235–248. Astor, R. A., Meyer, H. A., Benbenishty, R., Marachi, R., & Rosem Rosemond, ond, M. (2005 (2005). ). Schoo Schooll safet safety y inter interventio ventions: ns: Best practices and progra programs. ms. Children Children and Schoo Schools, ls, 27 (1), (1), 17. Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstei st ein, n, H. R. (2 (200 005) 5).. Comprehensive Comprehensive meta-analysis, Version Versi on 2. 2. Englewood, NJ: Biostat. Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H. R. (2010). A basic introduction to fixed-effect and ran random dom-ef -effec fects ts mod models els for met meta-a a-anal nalysi ysis. s. Re-
Duval, S., & Tweedie, R. (2000). A nonparametric “trim and fill” method of accounting for publication bias in meta-analysis. Journal meta-analysis. Journal of the American Statistical Association, 95(449), 95(449), 89–98. Egger, M., Davey Smith, G., Schnieder, M., & Minder, C. (1997). (1997 ). Bias in metameta-analy analysis sis detected by a simpl simple, e, graphical test. British test. British Medical Journal, 315, 629–634. 315, 629–634. Espelage, D. L., Bosworth, K., & Simon, T. R. (2001). Short-term stability and prospective correlates of bullying in middle middle-scho -school ol stude students: nts: An exami examinatio nation n of potential demographic, psychosocial, and environmental factors. Violence factors. Violence and Victims, 16, 411–426. 16, 411–426. Espelage, D. L., & Swearer, S. M. (2003). Research on school sch ool bul bullyi lying ng and vic victim timiza izatio tion: n: Wha Whatt hav havee we learned and where do we go from here? School Psychology Review, 32, 365–385. 32, 365–385. Espelage, D. L., Green, H. D., & Polanin, J. R. (in press). Willingness to intervene in bullying episodes among middlee schoo middl schooll stude students: nts: Indiv Individual idual and peerpeer-group group influences. Journal of Early Adolescence. fluences. *Evers, K., Prochaska, J., Van Marter, D., Johnson, J., & Prochaska, Proch aska, J. (2007 (2007). ). Trans Transtheor theoretica etical-bas l-based ed bully bullying ing prevention effectiveness trials in middle schools and high schools. [references]. Educational Research, 49, 397–414. Ferguson, C. J., Miguel, C. S., Kilburn, J. C., & Sanchez, P. (2007 (2007). ). The effec effectiven tiveness ess of school school-base -based d antianti-bulbullying programs: A meta-analytic review. Criminal review. Criminal Justice Review, 32, 401–414. 32, 401–414. *Fonagy, P., Twemlow, S. W., Vernberg, E. M., Nelson, J. M., Dill, E. J., Little, T. D., et al. (2009). A cluster randomized controlled trial of child-focused psychiatric consultation and a school systems-focused intervention to reduce aggression. Journal aggression. Journal of Child Psychology and Psychiatry, 50, 607–616. 50, 607–616.
search Synthesis Methods, 1,Simon, 97–111. Bosworth, K., Espelage, D., &1, 97–111. T. (1999). Factors associ ass ociate ated d wit with h bul bullyi lying ng beh behavi avior or in mid middle dle sch school ool students. The The Jo Jour urna nall of Ea Earl rlyy Ad Adol oles esce cenc nce, e, 19 19,, 341–352.
Frey, K.(2009). S., Hirschstein, K., Edstrom, L. V.,bullying, & Snell, J. L. ObservedM. reductions in school nonbullying aggression, and destructive bystander behavior: A longitudinal evaluation. Journal of Educational Psychology, 101, 101, 466–481.
these processes direct effects active bullying. Third,and thethe results from the on empathy review revealed inconclusive findings. Future research is required to elucidate the effects of prevention programs on this outcome. Finally, continued quantitative syntheses that focus on bystan bys tander der beh behavi avior or are req requir uired. ed. Cer Certai tainly nly only 11 studies from the last 10 years cannot accura acc uratel tely y des descri cribe be the sco scope pe of thi thiss iss issue. ue. Futu Fu ture re me meta ta-a -ana naly lyse sess sh shou ould ld in inco corp rpor orat ate, e, hopefully, new relevant literature. Footnote
*Article used in meta-analysis. References
62
A Meta-Analysis of School-Based Bullying
*Frey, K. S., Hirschstein, M. K., Snell, J. L., Edstrom, L., Van Sch Schoia oiack, ck, L., Mac MacKen Kenzie zie,, E. P., & Bro Broder derick ick,, C. J. (2005). Reducing playground bullying and supporting beliefs: An experimental trial of the steps to respect respe ct progr program. am. Developm Development ental al Psy Psycho cholog logy, y, 41, 479–491. Glass, G. V. (1976). Primary, secondary, and meta-analysis of research. research. Educational Researcher, 5(10), 5 (10), 3–8. 3–8. Glew Gl ew, , G. M. M., Fan, Fa n, M.,, Bullying, M. Kato Ka ton, n, W. W., , Ri Riva vara ra,, F. adjustP.,, & P. Kermic, M., A. (2005). psychosocial ment, and academic performance in elementary school. Pediatric Adolescent Medicine, 159, 1026–1031. 159, 1026–1031. Gini, G., Albie Albiero, ro, P., Benel Benelli, li, B., & Altoe `, G. (2007 `, (2007). ). Does empathy predict adolescents’ adolescents’ bullying and defending behavior? behavior? Aggressive Behavior, 33, 33, 467 467– – 476 476.. Hawk Ha wkin ins, s, D. L. L.,, Pe Pepl pler er,, D. J. J.,, Cr Crai aig, g, W. M. (2 (200 001) 1).. Naturalistic observations of peer interventions in bullying. Social Development, 10, 543–555. lying. 10, 543–555. Hedges, L. V. (1981). Distribution theory for glass’s estimator of effect size and related estimators. Journal estimators. Journal of Educational Statistics, 6, 6, 107–128. 107–128. Hedges, L.V., & Olkin, I. (1985). (1985). Statistical methods for meta-analysis. Orlando, FL: Academic Press. meta-analysis. Hedg He dges es,, L. V. V.,, & Pi Pigo gott tt,, T. D. (2 (200 004) 4).. Th Thee po powe werr of statistica stati sticall tests for moder moderators ators in metameta-analy analysis. sis. Psychological Methods, 9, 9, 426–444. Hedges, L. V., & Vevea, J. L. (1998). Fixed- and randomeffects models in meta-analysis. Psychological Methods, 3, 486–504. 3, 486–504. Howell, D. C. (2010). (2010). Statistical methods for psychology (7th ed.). Pacific Grove, CA: Duxbury. Hymel, S., & Swearer, S. (2010). Bullying: An age-old problem that needs new solutions. Education.com. solutions. Education.com. Re Retrieved triev ed from http:/ http://www. /www.educa education.c tion.com/re om/referenc ference/ e/ article/bullying-about-power-and-abuse-of-power/ Kaltiala-Heino, R., Rimpela, M., Marttunen, M., Rimpela, A., & Rantanen, P. (1999). Bullying, depression, and suicidal suici dal ideati ideation on in Finnis Finnish h adole adolescent scents: s: Schoo Schooll survey. British Medical Journal, 319, 348–351. vey. 319, 348–351. *Karna, A., Voeten, M., Little, T. D., Poskiparta, E., Kal jonen, jone n, A., & Salm Salmival ivalli, li, C. (201 (2011). 1). A larg large-sc e-scale ale eval evaluuation ati on of the KiV KiVaa an antib tibull ullyin ying g pr progr ogram am:: Gr Grade adess 4 – 6. Child Development, 82, 82, 311–330. Kochenderfer, B. J., & Ladd, G. W. (1996). Peer victimization: Manifestations and relations to school adjustment in kindergarten. Journal of School Psychology, 34, 267–283. 34, 267–283. Lagerspetz, K. J., Bjorkqvist, K., Bertz, M., & King, E. (1982). (1982 ). Group aggression aggression among school children in three schoo schools. ls. Scandinavia Scandinavian n Journ Journal al of Psych Psychology ology,, 23, 45–53. 23, 45–53. Levine, T., Asada, K. J., & Carpenter, C. (2009). Sample sizes and effect sizes are negatively correlated in metaanalyses: Evidence and implications of a publication bias again against st nonsi nonsignific gnificant ant findin findings. gs. Communication 286–302. Monographs, 76, 76, 286–302. Lipsey, Lipse y, M. W. (2009 (2009). ). Identi Identifying fying interesting interesting varia variables bles and an d an anal alys ysis is op oppo port rtun unit itie ies. s. In H. Co Coop oper er,, L. V. Hedges Hed ges,, & J. C. Val Valent entine ine (Ed (Eds.) s.),, The handbook of (pp. 147–1 147–158). 58). research resea rch synthe synthesis sis and metameta-analy analysis sis (pp. New York: Sage Publication. Lipsey, M. W., & Wilson, D. B. (2001). Practical metaanalysis (2nd (2nd ed ed.) .).. Th Thou ousa sand nd Oa Oaks ks,, CA CA:: Sa Sage ge Publications.
Matt, G. E., & Cook, T. D. (2009). Threats to the validity of generalized inferences. In H. Cooper, L. V. Hedges, & J. C. Valentine (Eds.), The handbook of research synthesis and meta-analysis (pp. meta-analysis (pp. 538–557). New York: Sage Publication. McDoug McD ougal, al, P., Hym Hymel, el, S., & Vai Vailla llanco ncourt urt,, T. (20 (2009) 09).. What happens over time to those who bully and those who are victi victimized mized?? Education.com. Retrieved Retrieved from http://www.education.com/reference/article/Ref_ What_Happens_Over/ *McLaughlin, L. P. (2009). The (2009). The effect of cognitive behavioral therapy and cognit cognitive ive behav behavioral ioral therapy plus media on the reduction of bullying and victimization and the increase of empathy and bystander response in a bully prevention program for urban sixth-grade students. Unpublished dents. Unpublished doctoral dissertation, University of Toledo, Digital Dissertations. *Menes *Me nesini ini,, E., Cod Codeca ecasa, sa, E., Ben Benell elli, i, B., & Cow Cowie, ie, H. (2003). Enhancing children’s responsibility to take action against bullying. bullying. Aggressive Behavior, 29, 1–14. *Merrell, *Merr ell, R. (2004 (2004). ). The impact of drama intervention intervention program on response of the bystander to bullying situations. Unpublished doctoral dissertation, Universituations. sity of Rochester, Digital Dissertations. Merrell, K. W., Gueldner, B. A., Ross, S. W., & Isava, D. M. (2008). How effective are school bullying intervention programs? A meta-analysis of intervention research. School Psychology Quarterly, search. Quarterly, 23, 23, 26–42. 26–42. Mishna, F., Cook, C., Saini, M., Wu, M. J., & MacFadden, R. (2010). Interventions to prevent and reduce cyber abuse abu se of you youth. th. Research Research on Socia Sociall Work Practice, 1(1), 1–10. Musher-Eizeman, D. R., Boxer, P., Danner, S., Dubow, E. F., Goldstein, S. E., & Herretick, D. M. I. (2004). Social-cognitive mediators of the relation of environmentall and emoti menta emotion on regul regulation ation factors to childr children’s en’s aggression. Aggressive Behavior, 30, 389–408. aggression. Nansel, T. R., Overpeck, M., Pilla, R. S., Ruan, W. J., Simons-Mor Simon s-Morton, ton, B., & Schei Scheidt, dt, P. (2001 (2001). ). Bully Bullying ing behaviors behav iors among US youth youth:: Preva Prevalence lence and assoc associaiation with psychosocial adjustment. Journal of American Medical Association, 285, 2094–2100. 285, 2094–2100. Nicker Nic kerson son,, A. B., Mel Mele, e, D., & Pri Princi nciott otta, a, D. (20 (2008) 08).. Attachment and empathy as predictors of roles as defenders or outsiders in bullying interactions. Journal interactions. Journal of School Psychology, 46, 687–703. 46, 687–703. O’Connell, P., Pepler, D., & Craig, W. (1999). Peer involvem vol vement ent in bul bullyi lying: ng: Ins Insigh ights ts and cha challe llenge ngess for intervention. Journal of Adolescence, intervention. Adolescence , 22, 437–452. Olweus, D. (1973). Personality and aggression. In J. K. Cole & D. D. Jensen (Eds.), Nebraska Symposium on (pp. 260–321). Lincoln: University of Ne Motivation (pp. Motivation braska Press. Olweus Olw eus,, D. (20 (2002) 02).. Bullyin Bullying g at sch school ool.. Malden, Malden, MA: Blackwell Publishers. Raudenbush, S. W. (2009). Analyzing effect sizes: Random-effects models. In H. Cooper, L. V. Hedges, & J. C. Valentine (Eds.), The (Eds.), The handbook of research synthesiss and metathesi meta-analy analysis sis (pp. 295–314). 295–314). New York: Sage Publication. Rigby, Rig by, K. (19 (1999) 99).. Pee Peerr vic victim timiza izatio tion n at sch school ool and the health healt h of secon secondary dary school stude students. nts. The Jou Journa rnall of School Psychology, 69, 95–104. 69, 95–104. Rigby, K., & Slee, P. T. (1993). Dimensions of interpersonal relation among Australian children and implica-
63
School Psychology Review, 2012, Volume 41, No. 1
tions for psychological well-being. British Journal of Educational Psychology, 133, 133, 33–42. Rivers, I., Poteat, V. P., Noret, N., & Ashurst, N. (2009). Observing bullying at school: The mental health implications of witness status. status. School Psychology Quarterly, 24, 211–223. 24, 211–223. Rosent Ros enthal hal,, R. (19 (1979) 79).. The “fil “filee dra drawer wer pro proble blem” m” and tolerance for null results. Psychological Bulletin, 86, 638–641. Ross, S. W., & Horner, R. H. (2009). Bullying prevention in positi positive ve behav behavior ior suppo support. rt. Journal of Applied Behavior Analysis, 42, 747–759. 42, 747–759. Rothstein, H. R., Sutton, A. J., & Borenstein, M. (2005). Publication bias in meta-analysis. West Sussex, England: John Wiley & Sons Ltd. Ryan, W., & Smith, D. J. (2009). Antibullying programs in schoo schools: ls: How effective are evalu evaluation ation practices? practices? Prevention Science, 10, 248–262. 10, 248–262. Salmivalli, C., Lagerspetz, K., Bjorkqvist, K., Osterman, K.,, & Ka K. Kauk ukia iane nen, n, A. (1 (199 996) 6).. Bu Bull llyi ying ng as a gr grou oup p process: Participant roles and their relations to social status within the group. Aggressive group. Aggressive Behavior, 22, 1–12. 22, 1–12. Salmivalli, Salmi valli, C., Kauki Kaukianine aninen, n, A., & Voete Voeten, n, M. (2005 (2005). ). Anti-bully Antibullying ing interv interventio ention: n: Imple Implementa mentation tion and outcome. British come. British Journ Journal al of Educa Educationa tionall Psych Psychology ology,, 75, 465. Sanch Sanchez-Me ez-Meca, ca, J., Marin Marin-Mart -Martinez, inez, F., for & dichotomized Chacon-Mos Chaco n-Mos-coso, S. (2003). Effect-size indices outcomes in meta-analysis. Psychological meta-analysis. Psychological Methods, 8, 448–462. Scholt Sch olte, e, R. H. J., Eng Engels els,, R. C. M. E., Ove Overbe rbeek, ek, G., Kemp, R. A. T., & Haselager, G. J. T. (2007). Stability in bullying and victimization and its association with social adjustment in childhood and adolescence. Jouradolescence. Journal of Abnormal Child Psychology, 35, 217–228. *Schumacher, P. J. (2007). To what extent will a sixtysecond sec ond vid video eo on bul bullyi lying ng pro produc duced ed by hig high h sch school ool students increase students’ awareness of bullying and change their attitudes to reduce acceptance of bullying in a high school environment? Unpublished master’s theesi th sis, s, Univ Un iver ersi sity ty of Pen Pe nns nsyl ylva vani nia, a, Dig Di git ital al Dissertations. Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-experimental designs. New York: Houghton Mifflin. Sharp, S., Sellors, A., & Cowie, H. (1994). Time to listen: Setting up a peer counseling service to help tackle the problem probl em of bully bullying ing in schoo schools. ls. Pastoral Pastoral Care Education, 12, 3–6. 12, 3–6. Smith, J. D., Twemlow, S. W., & Hoover, D. W. (1999). Bullie Bul lies, s, vic victim tims, s, and bys bystan tander ders: s: A met method hod of inschooll inter schoo interventi vention on and possi possible ble paren parental tal contri contribubutions. Child tions. Child Psychiatry and Human Development, 30, 29–37. Smith, J. D., Schneider, B. H., Smith, P. K., & Ananiadou, K. (2004 (2004). ). The effec effectivene tiveness ss of whole whole-scho -school ol antib antibulullying programs: programs: A synth synthesis esis of evalua evaluation tion research. School Psychology Review, Review, 33, 33, 547–560. 547–560. Souran Sou rander der,, A., Hel Helste stela, la, L., Hel Heleni enius, us, H., & Pih Piha, a, J. (2000). Persistence of bullying from childhood to adolescence— olesc ence—A A longi longitudin tudinal al 8-yea 8-yearr follo follow-up w-up study study.. Child Abuse & Neglect, 24, 873–881.
64
SPSS. (2010 (2010). ). Statistic Statistical al packa package ge for socia sociall scien sciences, ces, Version 19.0.0. Chicago, IL: IBM. Sutton Sut ton,, A. J. (20 (2005) 05).. Evi Eviden dence ce con concer cernin ning g the con consesequence que ncess of pub public licati ation on and related related bia biases ses.. In H. R. Rothstein, A. J. Sutton, & M. Borenstein (Eds.), Publication licat ion bias in metameta-analy analysis sis (pp. (pp. 175–1 175–192). 92). West Sussex, England: Wiley & Sons. *Stevens, V., Oost, P. V., & Bourdeaudhuij, I. D. (2000). The effectsattitudes of an anti-bullying intervention on peers’ and behaviour. Journalprogramme of Adolescence,, 23, cence 23, 21–34. 21–34. Sweare Swe arer, r, S. M., Espelage Espelage,, D. L., Vaillanc Vaillancour ourt, t, T., & Hymel, Hym el, S. (20 (2010) 10).. Wha Whatt can be don donee abo about ut sch school ool bullying? Linking research to educational practice. Edpractice. Educational Researcher, 39, 38–47. Sweeting, H., Young, R., West, P., & Der, G. (2006). Peer victimization and depression in early-mid adolescence: A longitudinal study. British Journal of Educational Psychology, 76, 577–594. 76, 577–594. Trikalinos, T. A., & Ioannidis, J. P. A. (2005). Assessing the evolution of effect sizes over time. In H. R. Rothstein, A. J. Sutton, & M. Borenstein (Eds.), Publication (Eds.), Publication bias in metameta-analy analysis. sis. West Sus Sussex sex,, Eng Englan land: d: Joh John n Wiley & Sons Ltd. Ttofi, M. M., & Farrington, D. P. (2009). What works in preventing bullying: Effective elements of anti-bullying programmes. Journal programmes. Journal of Aggression, 1, 13–24. Ttofi, M. M., & Farrington, D. P. (2011). Effectiveness of school-based programs to reduce bullying: A systematic and meta-analytic review. Journal review. Journal of Experimental Criminology, 7, 27–56. 7, 27–56. Twemlow, S. W., Fonagy, P., & Sacco, F. C. (2004). The role of bysta bystander nder in social architectur architecturee of bullying and violence viole nce in schoo schools ls and commu communities nities.. Annals Annals of the , 215–232. New York Academy of Sciences Sciences, 1036, 1036, 215–232. Unnever, J. D., & Cornell, D. G. (2003). The culture of bullying in middle school. Journal school. Journal of School Violence, 2, 5–27. 2, 5–27. Wang, J., Iannotti, R. J., & Nansel, T. R. (2009). School bullying among adolescents in the United States: Physical, verbal, relat relational, ional, cyber. Journal Journal of Adole Adolescent scent Health, 45, 45, 368–375. 368–375. *Whitaker, D. J., Rosenbluth, B., Valle, L. A., & Sanchez, E. (2004). Expect respect: A school-based intervention to promote awareness and effective responses to bullying and sexual harassment. In D. L. Espelage & S. M. Swearer (Eds.), Bullying (Eds.), Bullying in American schools (1st schools (1st ed., pp.. 32 pp 327– 7–35 350) 0).. Ma Mahw hwah ah,, NJ NJ:: La Lawr wren ence ce Er Erlb lbau aum m Associates. Willif Wil liford ord,, A., Bou Boulto lton, n, A., Nol Noland and,, B., Lit Little tle,, T. D., Karna, A., & Salmivalli, C. (2011). Effects of the KiVa anti-bullying program on adolescents’ depression, anxiety, iet y, and per percep ceptio tion n of pee peers. rs. Journal Journal of Abnor Abnormal mal Retrieved from http://www.springChild Psychology. Psychology. erlink.com/content/ 2532380182216828/
Date Received: February 1, 2011 Date Accepted: September 7, 2011 Action Editor: Joseph Betts
A Meta-Analysis of School-Based Bullying
Joshua hua R. Pol Polani anin n is a res resear earch ch met methodo hodology logy doc doctora torall stu studen dentt at Loy Loyola ola Univ Univers ersity ity Jos Chicago. His interests include methodological improvements to meta-analysis and hierarchical linear modeling. He currently serves as the methodologist for a large, schoolbased bullying prevention program and as the managing editor of the Campbell Collaboration’s Methods Group. Dorothy L. Espelage is a professor in the Department of Educational Psychology at the University Univers ity of Illinois Illinois,, Urbana Urbana-Cham -Champaign. paign. Her resea research rch progra programs ms include investigations investigations of bullying, sexual harassment, and dating violence among adolescents for almost two decades. She is engaged in a large randomized clinical trial of a school-based bullying prevention program. Therese D. Pigott is a professor at the School of Education, Loyola University Chicago. Her research interests are statistical methods for meta-analysis and methods for handling missing data in statistical analysis. She is a co-chair and editor of the Methods Group of the Campbell Collaboration, an international organization that produces systematic reviews of social interventions.
65
Copyright of School Psychology Review is the property of National Association of School Psychologists and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
