19
The Relationship Between Cultural IndividualismCollectivism and Student Aggression Across 62 Countries Silvia Bergmüller* BIFIEBundesinstitut für Bildungsforschung, Innovation & Entwicklung des österreichischen Schulwesens, Salzburg, Austria .......................................... .......................................... This study examined the relationship between countriesdominant cultural values (i.e., individualism and collectivism) and (a) school principalsperceptions of aggressive student behavior and (b) studentsselfreports of being aggressively victimized in school. Data on student aggression and victimization were collected across 62 countries in nationally representative samples of fourth and eighth graders (N ¼ 428,566) and their principals (N ¼ 15,043) by the Trends in International Mathematics and Science Study (TIMSS) 2007. Students were asked about three forms of aggressive victimization: physical, verbal, and relational; principals about two forms of aggressive student behavior: physical and verbal. Countrylevel regression analyses revealed that the level of cultural individualism, according to the individualism index (IDV) by Hofstede, Hofstede, and Minkov (2010), was not signicantly related to either form of studentreported victimization. However, school principals reported aggressive student behavior more often the more individualist, and hence less collectivist, their countrys culture. This relation was evident in the principalsreports on 4th and 8th grade studentsaggressive behavior for both physical and verbal aggression. Multilevel analyses revealed that cultural individualism was still a powerful predictor of principalreported aggressive student behavior after controlling for school and country characteristics. The discussion outlines reasons why principalsreports of aggressive student behavior are probably more valid indicators of student aggression than student selfreports of victimization, thereby supporting the hypothesis of culturedependency of aggression. Aggr. Behav. 39:182200, 2013. © 2013 Wiley Periodicals, Inc. .......................................... .......................................... Keywords: student aggression; culture; individualism; collectivism; TIMSS INTRODUCTION Aggressionany behavior intended to harm another person (DeWall, Anderson, & Bushman, 2011)has been described as a universal feature of human relations (e.g., Brown, 2004). More specically, student aggres- sion has been found to be a global phenomenon that affects one of the core institutions of modern society to some degree in virtually all nationstates(Akiba, LeTendre, Baker, & Goeslin, 2002, p. 830). However, actual rates of student aggression differ greatly across countries and cultures (Akiba et al., 2002; Crystal et al., 1994). These great variations may be explained by culture dependent reinforcement or punishment of aggressive behavior. Chen and French (2008) proposed that peers and adults perceive and evaluate specic socioemotional characteristics in ways consistent with cultural belief and value systems in their society. Due to these evaluations, adults and peers express different attitudes (e.g., acceptance or rejection) toward children who display certain characteristics in social interactions. Accordingly, different societies provide different con- texts for social learning experiences. Bandura (1973) mentioned three important learning processes of aggres- sive behavior: imitation, direct operant reinforcement, and vicarious reinforcement through observational learning. Therefore, children in societies that value aggression highly may learn to behave aggressively because they receive direct reinforcement for their aggressive behavior more often, they have more aggressive models they can imitate, and they observe other persons being reinforced for aggressing more often. * Correspondence to: Silvia Bergmüller, BIFIEBundesinstitut für Bildungsforschung, Innovation & Entwicklung des österreichischen Schulwesens, Alpenstraße 121, 5020 Salzburg, Austria. Email: s.bergmueller@bifie. Received 11 August 2011; Accepted 14 January 2013 Published online 12 March 2013 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/ab.21472 AGGRESSIVE BEHAVIOR Volume 39, pages 182200 (2013) © 2013 Wiley Periodicals, Inc.

The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

  • Upload
    silvia

  • View
    215

  • Download
    1

Embed Size (px)

Citation preview

Page 1: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

The Relationship Between Cultural Individualism–

Collectivism and Student Aggression Across62 CountriesSilvia Bergmüller*

BIFIE—Bundesinstitut für Bildungsforschung, Innovation & Entwicklung des österreichischen Schulwesens, Salzburg,Austria

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .This study examined the relationship between countries’ dominant cultural values (i.e., individualism and collectivism) and(a) school principals’ perceptions of aggressive student behavior and (b) students’ self‐reports of being aggressively victimized inschool. Data on student aggression and victimization were collected across 62 countries in nationally representative samples offourth and eighth graders (N ¼ 428,566) and their principals (N ¼ 15,043) by the Trends in International Mathematics andScience Study (TIMSS) 2007. Students were asked about three forms of aggressive victimization: physical, verbal, and relational;principals about two forms of aggressive student behavior: physical and verbal. Country‐level regression analyses revealed thatthe level of cultural individualism, according to the individualism index (IDV) by Hofstede, Hofstede, andMinkov (2010), was notsignificantly related to either form of student‐reported victimization. However, school principals reported aggressive studentbehavior more often the more individualist, and hence less collectivist, their country’s culture. This relation was evident in theprincipals’ reports on 4th and 8th grade students’ aggressive behavior for both physical and verbal aggression. Multilevelanalyses revealed that cultural individualism was still a powerful predictor of principal‐reported aggressive student behaviorafter controlling for school and country characteristics. The discussion outlines reasons why principals’ reports of aggressivestudent behavior are probably more valid indicators of student aggression than student self‐reports of victimization, therebysupporting the hypothesis of culture‐dependency of aggression. Aggr. Behav. 39:182–200, 2013. © 2013 Wiley Periodicals, Inc.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Keywords: student aggression; culture; individualism; collectivism; TIMSS

INTRODUCTION

Aggression—any behavior intended to harm anotherperson (DeWall, Anderson, & Bushman, 2011)—hasbeen described as a universal feature of human relations(e.g., Brown, 2004). More specifically, student aggres-sion has been found to be “a global phenomenon thataffects one of the core institutions of modern society tosome degree in virtually all nation‐states” (Akiba,LeTendre, Baker, & Goeslin, 2002, p. 830). However,actual rates of student aggression differ greatly acrosscountries and cultures (Akiba et al., 2002; Crystal et al.,1994). These great variations may be explained byculture dependent reinforcement or punishment ofaggressive behavior. Chen and French (2008) proposedthat peers and adults perceive and evaluate specific socio‐emotional characteristics in ways consistent with culturalbelief and value systems in their society. Due tothese evaluations, adults and peers express differentattitudes (e.g., acceptance or rejection) toward childrenwho display certain characteristics in social interactions.Accordingly, different societies provide different con-

texts for social learning experiences. Bandura (1973)mentioned three important learning processes of aggres-sive behavior: imitation, direct operant reinforcement,and vicarious reinforcement through observationallearning. Therefore, children in societies that valueaggression highly may learn to behave aggressivelybecause they receive direct reinforcement for theiraggressive behavior more often, they have moreaggressive models they can imitate, and they observeother persons being reinforced for aggressingmore often.

*Correspondence to: Silvia Bergmüller, BIFIE—Bundesinstitut fürBildungsforschung, Innovation & Entwicklung des österreichischenSchulwesens, Alpenstraße 121, 5020 Salzburg, Austria. E‐mail:s.bergmueller@bifie.

Received 11 August 2011; Accepted 14 January 2013

Published online 12 March 2013 in Wiley Online Library(wileyonlinelibrary.com).DOI 10.1002/ab.21472

AGGRESSIVE BEHAVIORVolume 39, pages 182–200 (2013)

© 2013 Wiley Periodicals, Inc.

Page 2: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

Victimization—the experience of being a target ofaggressive behavior (Hawker & Boulton, 2000)—is animportant indicator of aggressive behavior. To explainstudent victimization in a cultural context, Benbenishtyand Astor (2008) proposed a “model of social‐ecologicalinfluences on student victimization”. This model ishighly influenced by Bronfenbrenner’s (1979) ecologicaldevelopmental theory and presents student victimizationwithin nested contexts. The center of the model is theschool context, and distal layers include family,neighborhood, and culture. The authors emphasize thatthe external contexts, in which a school is embedded,interact with internal school and student characteristics toinfluence levels of victimization in schools. It has alreadybeen shown that aggression rates differ greatly betweenschools within a country and that these school differencesare associated with characteristics of schools and theirstudent population, like school type, size of schoolcommunity, or percentage of male students (Bergmüller&Wiesner, 2009, 2012; Fuchs, 2008). Altogether, schoolcharacteristics are powerful predictors of studentaggression. Therefore, when searching for predictors ofstudent aggression at the individual student or the countrylevel, characteristics of the school context need to becontrolled for. However, in the search for country‐levelpredictors of student aggression, the school context hasyet to be considered.

Individualism and Collectivism

Individualism–collectivism is a cultural dimension thathas been widely employed for delineating differencesbetween cultures (e.g., Hofstede, Hofstede, &Minkov, 2010; Triandis & Suh, 2002). The individual-ism–collectivism dimension characterizes the degree towhich people in a society are integrated into groups(Hofstede, 2001). Individualism is characterized by theassumption that individuals are independent of oneanother; there is a high emphasis on the needs and desiresof the individual and linkages between members of thesociety are relatively weak. Collectivism, however, ischaracterized by the assumption that individuals aredependent on one another; and places great emphasis onthe needs and values of the group rather than those of theindividual. In collectivistic societies, people are expectedto subordinate their personal goals and ambitions to thoseof the group and harmony with the environment is highlyvalued (e.g., Oyserman, Coon, & Kemmelmeier, 2002;Hofstede et al., 2010).An enormous number of studies have revealed

systematic variations in cognitive, emotional, and socialfunctioning between participants from individualistsocieties compared to those from collectivist societies(Kim, Triandis, Kagitçibasi, Choi, & Yoon, 1994;Oyserman et al., 2002, for reviews), thus demonstrating

the applicability of the individualism–collectivismframework in cross‐cultural comparison.There has been some dispute however whether

individualism and collectivism characterize societies orcountries (as proposed by Hofstede et al., 2010) orwhether individuals may also be characterized by thesevariables (Oyserman et al., 2002 provided a review ofboth). Oyserman et al. showed that there were cleardifferences between individuals of the same society (e.g.,United States) and demonstrated that individualism andcollectivism were two separate (uncorrelated) variables.Hofstede et al. (2010) however proposed that while

people within a country might differ in their culturalorientation there are some stable values that characterizea society on the whole and that these cultural valuesindeed differ between societies. Hofstede et al. (2010)and Triandis and Suh (2002) suggested that, on thesociety level, individualism and collectivismmay be seenas the two poles of the same dimension.Hofstede et al. provided individualism scores for a

large number of countries. These scores were obtained byquestionnaires from employees of the samemultinationalcorporation in different nations. The questionnaires wereinitially designed to measure work satisfaction, however,when analyzing the obtained data, Hofstede (1984)identified four different cultural dimensions: Powerdistance, uncertainty avoidance, individualism, andmasculinity. The questionnaires were originally (1968/1972) administered to over 116,000 workers in 40 coun-tries. Currently, the Hofstede scales cover 76 countries(see Hofstede et al., 2010). Based on the survey data,Hofstede derived index scores for each country on eachscale.Hofstede’s work can be criticized for twomain reasons:

First, he did not explicitly design the questions tomeasure cultural values but only developed his theorypost hoc when analyzing the data. Second, he did notassess the data in country‐representative samples, whichclearly limits the validity and reliability of his scales.However, various studies have replicated the Hofstededimensions (see Hofstede, 2001 and Hofstedeet al., 2010, for overviews). In particular, the individual-ism–collectivism dimension has been well validated (seealso Cheng et al., 2013).Despite being aware of its limitations, I drew on

Hofstede’s individualism scale in the present study. First,because it has been validated thoroughly, second becauseit is considered state of the art in current high‐qualitypsychological research (e.g., Cheng et al., 2013), andthird because it has been applied in the cross‐culturalresearch on aggression (e.g., Bergeron & Schneider,2005) this article aims to be comparable with.As already mentioned, Hofstede et al. (2010) specifi-

cally designed the individualism index (IDV) scale to

Aggr. Behav.

Student Aggression and Culture 183

Page 3: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

compare countries and not individuals or unrepresenta-tive groups of individuals. Therefore when using theHofstede scales to compare the characteristics ofindividuals across different societies, these character-istics need to be measured in country‐representativesamples and aggregated at the country level.

Aggression in Cultural Context

Based on the individualism–collectivism framework,Hofstede (2001) pointed out cultural differences in theacceptance of conflict. For collectivist societies, heemphasized the importance of always maintainingharmony and avoiding confrontations; for individualistsocieties however, he emphasized that confrontationswere seen as normal and speaking one’s mind isconsidered a characteristic of an honest person. Hofstedealso conveyed these principles to school life: “In thecollectivist classroom, the virtues of harmony andmaintaining face reign supreme. Confrontations andconflicts should be avoided or at least should beformulated so as not to hurt anyone” (Hofstedeet al., 2010, p. 118). In individualist classrooms however,“confrontations and open discussions of conflict aresometimes considered salutary, and face consciousness isweak or nonexistent. Good educators are supposed toreinforce the student’s self‐esteem” (Hofstede, 2001, p.235).Because of their strong emphasis on the avoidance of

conflict and the maintenance of social harmony, it seemsreasonable to expect that aggression is less common andless accepted in collectivistic societies. Looking back toBandura’s social learning theory and to Chen andFrench’s contextual‐developmental perspective, themechanisms by which cultural individualism–collectiv-ism affect aggressive behavior may be uncovered. First,children in collectivist societies are supposed to beintegrated into social learning environments in whichthey are less reinforced for aggressive behavior andobserve aggressive behavior less often than children inindividualist societies. Second, from these culturallydifferent learning experiences, individual beliefs andattitudes about aggressive behaviors may be shapedwhich finally lead children in collectivist societies tobehave less aggressively than children in individualistsocieties.Cross‐cultural studies have already shown that children

are culture‐specifically reinforced for expressing emo-tions like anger (approved in individualist societies butreproved in collectivist ones; Cole, Tamang, &Shrestha, 2006; Zahn‐Waxler, Friedman, Cole, Mizuta,& Hiruma, 1996) and shy‐inhibited behavior (approvedin collectivist, but reproved in individualist; Chenet al., 1998; Rubin, Burgess, & Coplan, 2002). For theschool context in particular, Chen and French (2008)

pointed out that aggressive, disruptive, and deviantbehavior were strictly prohibited in collectivist Chinesesociety because of their potential threat to groupharmony. For example, Chinese schools regularly engagein public evaluations that require children to evaluatethemselves in the group with respect to whether theirbehavior reaches school standards and whether theirbehavior improves over time.A large number of studies compared the aggression

rates of children and adolescents cross‐culturally. Thesestudies found that aggression varies greatly acrosscultures. For example, Akiba et al. (2002) showed thata fear of physical aggression is prevalent in eighth‐gradestudents among all 37 nations studied. However, the rateof students who fear physical aggression from otherstudents varies considerably, from less than 10% inDenmark and Singapore to more than 70% in Israel.Crystal et al. (1994) compared several indicators ofmaladjustment between US, Japanese, and Taiwanese11th‐graders. The highest levels of aggressive feelingswere found for US students, whereas students from Japanand Taiwan reported a significantly lower frequency ofaggressive feelings. Even when shifting the reporter’sperspective from students to parents, aggressive behavior(e.g., fighting, arguing) was reported to bemore prevalentin youths from the individualist United States than fromcollectivist Thailand (Weisz, Suwanlert, Chaiyasit, &Walter, 1987).A study by Lansford et al. (2012) however did not

provide support for the notion that aggression rates wouldbe culture‐dependent. In this nine‐country study, childrenages 7–10 years (n ¼ 1,410) reported on their physicallyand relationally aggressive behavior. Results showed thatphysical and relational aggression were over‐representedin collectivist Kenya and Jordan and underrepresented incountries as different as China, Columbia, Sweden, andthe United States.It must be noted that even in strongly conflict avoidant

and harmony seeking collectivistic societies there may behigh levels of aggression against members of out‐groups(e.g., Triandis, 2004) and against women by male familymembers (e.g., AFPPD/UNFPA, 2003). Furthermore,studies on collectivist cultures of honor have shown thataggressive retaliation is a likely response when ideas ofhonor are threatened among members of a collectivistsociety (e.g., Nisbett & Cohen, 1996).Although there might be more aggression in collectiv-

ist societies under certain conditions and for certaingroups, overall aggression rates among students arelikely to be higher in more individualist, less collectivistsocieties. In the following, I present the few studiesthat explicitly investigated the relationship betweencultural individualism–collectivism and studentaggression.

Aggr. Behav.

184 Bergmüller

Page 4: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

A study by Forbes, Zhang, Doroszewicz, and Haas(2009) compared college students (n ¼ 716) in threecountries: theUnited States (highly individualist), Poland(intermediate), and China (highly collectivist). Aggres-sion was measured using a self‐report questionnaire onresponses to conflict. Direct and indirect forms ofaggression were most often reported by students in theUnited States. Polish students reported intermediateaggression levels and Chinese students the lowest. Thisled the authors to the conclusion that aggression is higher,the more individualist a nation is. But this conclusionmust be made very cautiously as the sample of countrieswas very small (n ¼ 3) and the samples of studentswithin countries were not representative for the threecountries. This greatly limits any generalization of theresults.The most comprehensive study in this field is a meta‐

analysis by Bergeron and Schneider (2005) that covered28 nations. The authors reviewed 36 studies thatcompared levels of peer‐directed aggression (mostly bystudents) cross‐culturally. The differences betweennations found for aggression were compared to thedifferences of individualism–collectivism for the samenations. Bergeron and Schneider found aggression scoresto be positively related to scores onHofstede’s IDV scale.Although this review tested the relationship betweencultural values and aggression on a relatively broadempirical basis, it has several limitations: First, the largemajority of studies that were included in the review onlycompared measures of aggression between two coun-tries. Second, no detailed results on the relation betweenindividualism–collectivism and aggression were provid-ed, neither for different reporters’ perspectives nor fordifferent forms of aggression. And third, the samples ofthe studies that were included in Bergeron andSchneider’s review were not representative for thesecountries. In this respect, both studies discussed here(Bergeron& Schneider, 2005; Forbes et al., 2009) did notadequately apply Hofstede’s (2001) IDV scale becausethey did not use nationally representative samples tocompare the countries.Furthermore, it is unclear so far whether culture would

still predict aggression if economic or developmentalindicators were controlled for at the country level. Onenotable study by Akiba et al. (2002) did not includemeasures of culture, but investigated the influence ofvarious national variables (gross domestic product,national homicide rate, size of youth population, meanmathematics achievement, math achievement variation,and remedial education) on (a) eighth‐grade student’sfear of physical victimization and (b) students’ reports ofphysical victimization by their fellow students. Multi-variate regression analyses revealed the national varia-tion of mathematics achievement as the most powerful

predictor for both students’ fear of victimization andreports of peer victimization: In countries with greaterachievement differences between high‐achieving andlow‐achieving students the above aggression indicatorswere more frequent. The authors speculate that incountries with high achievement differences low achiev-ers might be labeled as academic failures or regardedthemselves as having limited future opportunities. Thismight cause the frustration that triggers violence inschool.Summing up, research on the relation between cultural

individualism–collectivism and aggression so far islimited in various aspects: (a) in systematically address-ing specific forms of aggression, (b) in systematicallycomparing different age groups, (c) in systematicallycomparing different reporters’ perspectives, (d) in therange of countries (e.g., no Arab or African countrieswere included), (e) in the representativeness of thesamples, and (f) in consistently measuring aggressionacross different countries. Beyond this, it is unclear,whether individualism–collectivism still predicted ag-gression after controlling for school and countrycharacteristics.

The Present Investigation

The present study is the first to test the relation betweenindividualism–collectivism and student aggression usinglarge, representative samples of students and theirprincipals from a broad range of countries. The studyincludes data on physical and verbal aggression reportedby both principals and students along with measures ofrelational aggression reported by students. The samemeasures of aggression were implemented in allcountries participating in TIMSS 2007 and in twogrades: Grade 4 and Grade 8. The variety of countriesgoes beyond existing studies by including a number ofAfrican and Arab countries. This study thereforeovercame the above‐mentioned limitations of precedingstudies.The hypothesis that aggressive student behavior is

reported more often the more individualist, and henceless collectivist, a country’s culture is, was tested by:

1. Country‐level regression analysis examining therelationship between cultural individualism–collec-tivism and (a) different forms of aggressive victimi-zation reported by students and (b) different forms ofaggressive student behavior reported by principals.

2. Multilevel models examining whether the relationshipbetween cultural individualism–collectivism and (a)students’ reports of aggressive victimization and (b)principals’ reports of aggressive student behaviorpersists when school and country characteristics areheld constant.

Aggr. Behav.

Student Aggression and Culture 185

Page 5: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

METHODS

Participants

Every four years the International Association for theEvaluation of Educational Achievement administers theTIMSS tests and questionnaires to fourth‐ and eighth‐grade students worldwide. In 2007, nearly half a millionstudents (n ¼ 428,566, thereof 184,080 at Grade 4 and244,486 at Grade 8) from 62 countries participated. Itshould be noted that for some countries, not the wholecountry but only one or more regions within the countrywere tested. This was the case for Canada (only Alberta,British Columbia, Ontario, and Quebec were included),Spain (only Basque Country was included), and theUnited Arab Emirates (only Dubai was included). TheCanadian provinces were not counted as separatecountries. Even though not all of the other countriesparticipating in TIMSS 2007 were sovereign states (e.g.,England, Scotland, Palestinian National Authority, andHong Kong), these regions were considered to beseparate countries by the international study center(Mullis &Martin, 2008). The present study followed thispolicy.Countries could decide whether they would participate

with only one target population (fourth grade or eighthgrade) or with both populations. In TIMSS 2007, 10countries only participated with fourth grade, 24 onlywith eighth grade, and 28 countries with both grades. InTIMSS, a systematic two‐stage probability proportional‐to‐size sampling technique was used to draw samples thatwere representative of the student populations(Joncas, 2008). At the first stage, schools were sampledwith probability proportional to size. At the second stage,one or more intact classes of students from the targetgrades were sampled.In each country and in each grade at least 3,000

students (M ¼ 4,383, SD ¼ 869.72 for Grade 4;M ¼ 4,445, SD ¼ 842.52 for Grade 8) were tested(except for Basque country: 2,269 students). Thesestudents were enrolled in a total of 15,043 schools (6,805schools for Grade 4 and 8,238 schools for Grade 8). Onaverage 162 schools per country (SD ¼ 34.29) partici-pated in Grade 4 and on average 150 schools per country(SD ¼ 35.76) participated in Grade 8. The sample ofschools exceeded 130 in all countries except for smallcountries or regions with a limited number of schools(Dubai, Malta, Qatar, Hong Kong, Cyprus, and Bahrain).Detailed information about sample sizes of schools andstudents for each country and each gradewas provided byJoncas (2008).Students completed paper‐and‐pencil tests in mathe-

matics and science. Additionally, all sampled studentsand their principals were asked to complete backgroundquestionnaires that assessed various aspects of the

students’ home and school life, including aggressivestudent behavior.

Measures

All measures except for individualism and grossdomestic product were drawn from the TIMSS 2007assessment.1 For all TIMSS questionnaire items exten-sive translation and translation verification ensures cross‐cultural validity. Johansone and Malak (2008) gave adetailed description of the translation process. Informa-tion about questionnaire construction was provided byErberber, Arora, and Preuschoff (2008). The assessmentadministration had to follow a standardized script thatwas defined by the international study center (Johansone& Neuschmidt, 2008). The implementation of interna-tionally defined procedures was controlled by qualitycontrol monitors (Johansone, 2008). All of theseprocedures ensured highly comparable data betweencountries.Aggressive student behavior. Aggressive stu-

dent behavior was measured by questioning principalsand students: Principals were asked about their percep-tions of aggressive student behavior, and students wereasked whether they had been aggressively victimized intheir school by other students. The questions were thesame for Grade 4 and Grade 8 and addressed differentforms of aggressive student behavior. Both principals andstudents were asked about verbal and physical aggres-sion, but only students were asked about relationalaggression, as this form of aggression is less observable.Table I gives an overview of question stems and itemwording along with scale categories. Table I also showsthat the student and principal aggression scales eachincluded only one item for each type of aggression.Although all students and principals sampled were

asked to complete a questionnaire, the items onaggressive student behavior were not completed by allstudents and principals. Notably, no data on studentreported victimization were available for the UnitedStates. Average missing rates across countries werecomputed for in all measures of aggressive studentbehavior (after excluding the United States for student‐reported measures). Results showed very low missingrates for the total sample of less than 5% for all measuresof aggressive student behavior (i.e., for all forms ofaggressive student behavior reported by principals orstudents across both grade cohorts). Consideringcountries separately, missing rates exceeded 15% onlyin three countries/regions (British Columbia, Dubai, andMongolia) for at least one measure of aggressive student

1The complete TIMSS 2007 database can be downloaded from http://timss.bc.edu/timss2007/idb_ug.html.

Aggr. Behav.

186 Bergmüller

Page 6: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

behavior (see estimates denoted by an asterisk inTable II).Table II shows country‐level means and standard

deviations for each item and each grade. Missing valueswere deleted list‐wise to compute these estimates. Thedata show that the reported frequency of different formsof aggressive student behavior varied greatly acrosscountries. The range of physical aggression reported bythe fourth graders’ principals was 1.16 (Georgia andMongolia) to 2.53 (Denmark). The range of physicalaggression reported by eighth graders’ principals was1.19 (Indonesia) to 2.39 (Cyprus). The range of verbalaggression reported by fourth graders’ principals was1.15 (Georgia) to 2.87 (Quebec). The range of verbalaggression reported by eighth graders’ principals was1.18 (Georgia) to 3.15 (Australia). Due to the coding ofthe student data as 0 (no) and 1 (yes), the country meansin Table II represent ratios of self‐identified victims. Theproportion of victims experiencing physical aggression atGrade 4 ranged from 6% (Kazakhstan) to 48% (ChineseTaipei). The proportion of victims experiencing physicalaggression at Grade 8 ranged from 6% (Italy and BasqueCountry) to 39% (Thailand). The range of victims ofverbal aggression at Grade 4 was 8% (Kazakhstan) to45% (Chinese Taipei). The range of victims of verbalaggression at Grade 8 was 6% (Georgia) to 51% (Ghana).The range of victims of relational aggression at Grade 4was 10% (Latvia) to 36% (Ontario). The range of victimsof relational aggression at Grade 8 was 2% (Republic ofKorea) to 33% (Ghana).Table III shows that different forms of aggression from

the same perspective (either principals or students) weregenerally higher intercorrelated than the same forms ofaggression from different perspectives. Table III alsoshows that the same measures of aggressive studentbehavior correlated extremely high between Grade 4 andGrade 8. Notably, principals’ reports of aggressivestudent behavior correlated even higher between thetwo grade cohorts than the students’ reports. This

indicates that countries with high levels of aggressivestudent behavior at Grade 4 also have high levels ofaggressive student behavior at Grade 8 and countries withlow levels of aggressive student behavior at Grade 4 alsohave relatively low levels at Grade 8, especially whenaggressive student behavior was assessed by principals.School size. Principals were asked: “what is the

total school enrollment (number of students) in allgrades?” This was an open‐ended question.Size of school community. Principals were

asked: “how many people live in the city, town, orarea where your school is located?” Originally, this itemwas scored on a 6‐point scale from 1 (More than 500,000people) to 6 (3,000 people or fewer). To make the dataeasier to interpret, the item was reverse‐coded, resultingin a scale in which one stands for the smallestcommunities and six for the largest.Proportion of students from disadvantaged

homes. Principals were asked about the percentage ofstudents in their school who come from economicallydisadvantaged homes. The item was coded on a 4‐pointscale from 1 (0 to 10%) to 4 (more than 50%).Percentage of male students per school. In-

formation about the sex of fourth‐ and eighth‐gradestudents came from the “student–teacher linkage form”

which schools had to send to the TIMSS NationalResearch Coordinators before testing andwhich thenwasintegrated into the database. There were no missingvalues for this variable. In the TIMSS 2007 database thisitem was coded as 1 (female) and 2 (male). For thepresent study, it was recoded into 0 (female) and 1 (male).Therefore, when computing means by school, percen-tages of male students per school resulted.Individualism index (IDV). IDV scores were

taken from Hofstede et al. (2010). Higher scores onthe bipolar IDV scale denote that the country is higher inindividualism, and lower scores indicate that thecountry’s culture is more collectivist in nature. Asindicated in Table IV, scores were spread reasonably

TABLE I. Items for Assessing Aggressive Student Behavior in TIMSS 2007

Questions to principals Questions to students

Question stem How often does each of the following problem behaviorsoccur among fourth‐grade (eighth‐grade) students inyour school?

In school, did any of these things happen during the lastmonth?

Form of aggressionPhysical Physical injury to other students I was hit or hurt by other student(s) (for example, shoving,

hitting, kicking)Verbal Intimidation or verbal abuse of other students I was made fun of or called namesRelational — I was left out of activities by other students

Scale categories 1 ¼ never, 2 ¼ rarely, 3 ¼ monthly, 4 ¼ weekly, 5 ¼daily

0 ¼ noa, 1 ¼ yes

aIn the TIMSS 2007 database, no was coded as 2; to make the data easier to interpret, Code 2 was transformed into Code 0 for the present study.

Aggr. Behav.

Student Aggression and Culture 187

Page 7: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

TABLEII.Agg

ressiveStudentBehaviorReportedbyPrincipalsan

dStudents

Meansan

dStandardDeviation

sbyCou

ntry,Grade,an

dForm

ofAggression

Country

Grade

4Grade

8

Principalsa

Studentsb

Principalsa

Studentsb

Physical

Verbal

Physical

Verbal

Relational

Physical

Verbal

Physical

Verbal

Relational

Algeria

1.39

c(0.59)

1.60

c(0.82)

0.30

(0.46)

0.30

c(0.46)

0.22

(0.42)

1.66

c(0.71)

2.03

c(0.95)

0.18

(0.39)

0.18

(0.38)

0.15

(0.35)

Arm

enia

1.22

c(0.82)

1.29

c(0.83)

0.23

(0.42)

0.19

c(0.40)

0.22

(0.42)

1.25

c(0.87)

1.39

c(0.88)

0.13

(0.33)

0.10

(0.30)

0.16

(0.37)

Australia

1.95

c(0.66)

2.51

c(0.92)

0.40

(0.49)

0.35

c(0.48)

0.33

(0.47)

2.01

c(0.56)

3.15

c(1.00)

0.23

(0.42)

0.30

(0.46)

0.15

(0.35)

Austria

1.71

c(0.78)

2.45

c(1.06)

0.27

(0.44)

0.33

c(0.47)

0.21

(0.40)

NP

NP

Bahrain

NP

NP

1.76

c(0.64)

2.18

c(0.73)

0.31

(0.46)

0.25

(0.43)

0.14

(0.35)

BosniaandHerzegovina

NP

NP

1.66

c(0.65)

1.72

c(0.70)

0.16

(0.37)

0.15

(0.36)

0.09

(0.29)

Botsw

ana

NP

NP

2.16

c(0.85)

2.78

c(1.13)

0.30

(0.46)

0.43

(0.49)

0.26

(0.44)

Bulgaria

NP

NP

1.51

c(0.68)

1.98

c(0.84)

0.22

(0.42)

0.24

(0.42)

0.18

(0.39)

Canada(A

lberta)

1.96

c(0.60)

2.42

c(0.80)

0.38

(0.48)

0.33

c(0.47)

0.33

(0.47)

NP

NP

Canada(British

Colum

bia)

1.75

c(0.57)

2.34

c(0.86)

0.33

(0.47)

0.31

c(0.46)

0.32

(0.47)

1.98

c(0.57)

2.76

c(0.87)

0.22

(0.41)

0.27

(0.45)

0.16

(0.37)

Canada(O

ntario)

2.01

c(0.74)

2.63

c(1.05)

0.40

(0.49)

0.33

c(0.47)

0.36

(0.48)

2.09

c(0.67)

2.76

c(1.07)

0.24

(0.43)

0.28

(0.45)

0.16

(0.37)

Canada(Q

uebec)

2.09

c(0.69)

2.87

c(0.99)

0.36

(0.48)

0.40

c(0.49)

0.27

(0.44)

1.81

c(0.80)

2.96

c(1.07)

0.13

(0.33)

0.22

(0.42)

0.09

(0.28)

Chinese

Taipei

1.57

c(0.60)

1.97

c(0.75)

0.48

(0.50)

0.45

c(0.50)

0.24

(0.43)

1.88

c(0.49)

2.09

c(0.62)

0.24

(0.42)

0.35

(0.48)

0.09

(0.29)

Colum

bia

1.29

c(0.49)

1.83

c(0.85)

0.28

(0.45)

0.40

c(0.49)

0.20

(0.40)

1.54

c(0.69)

2.39

c(1.19)

0.17

(0.37)

0.36

(0.48)

0.10

(0.30)

Cyprus

NP

NP

2.39

c(0.96)

2.86

c(0.98)

0.22

(0.42)

0.33

(0.47)

0.09

(0.29)

Czech

Republic

1.90

c(0.81)

2.11

c(0.83)

0.18

(0.38)

0.43

c(0.50)

0.12

(0.33)

1.76

c(0.62)

2.10

c(0.66)

0.09

(0.29)

0.32

(0.47)

0.08

(0.27)

Denmark

2.53

c(0.81)

2.56

c(1.00)

0.25

(0.43)

0.18

c(0.39)

0.14

(0.34)

NP

NP

Egypt

NP

NP

1.38

c(0.53)

1.84

c(0.97)

0.23

(0.42)

0.24

(0.42)

0.26

(0.44)

ElSalvador

1.52

c(0.76)

1.61

c(0.73)

0.28

(0.45)

0.36

c(0.48)

0.21

(0.41)

1.57

c(0.78)

1.73

c(0.87)

0.13

(0.34)

0.23

(0.42)

0.11

(0.31)

England

2.02

c(0.63)

2.24

c(0.79)

0.43

(0.50)

0.36

c(0.48)

0.30

(0.46)

2.06

c(0.56)

2.60

c(0.92)

0.18

(0.38)

0.26

(0.44)

0.11

(0.32)

Georgia

1.16

c(0.37)

1.15

c(0.36)

0.23

(0.42)

0.15

c(0.36)

0.27

(0.44)

1.29

c(0.45)

1.18

c(0.40)

0.19

(0.39)

0.06

(0.23)

0.06

(0.24)

Germany

2.13

c(0.78)

2.68

c(1.04)

0.24

(0.43)

0.29

c(0.45)

0.18

(0.38)

NP

NP

Ghana

NP

NP

1.55

c(0.59)

2.27

c(1.06)

0.32

(0.47)

0.51

(0.50)

0.33

(0.47)

HongKongSAR

1.55

c(0.55)

1.64

c(0.57)

0.32

(0.47)

0.44

c(0.50)

0.18

(0.39)

1.54

c(0.58)

1.87

c(0.69)

0.15

(0.36)

0.38

(0.49)

0.06

(0.24)

Hungary

1.96

c(0.79)

1.97

c(0.89)

0.34

(0.47)

0.42

c(0.49)

0.30

(0.46)

1.90

c(0.76)

2.07

c(0.86)

0.13

(0.34)

0.23

(0.42)

0.17

(0.37)

Indonesia

NP

NP

1.19

c(0.41)

1.52

c(0.57)

0.33

(0.47)

0.43

(0.50)

0.15

(0.36)

Iran,Islamic

Republic

of1.35

c(0.50)

1.62

c(0.59)

0.26

(0.44)

0.24

c(0.42)

0.26

(0.44)

1.34

c(0.55)

1.97

c(0.77)

0.21

(0.41)

0.25

(0.43)

0.17

(0.37)

Israel

NP

NP

1.75

c(0.65)

2.73

c(0.97)

0.17

(0.38)

0.22

(0.42)

0.08

(0.28)

Italy

1.52

c(0.72)

1.97

c(1.04)

0.22

(0.41)

0.44

c(0.50)

0.24

(0.42)

1.81

c(0.76)

2.38

c(0.89)

0.06

(0.24)

0.26

(0.44)

0.10

(0.30)

Japan

1.58

c(0.73)

1.58

c(0.84)

0.27

(0.45)

0.33

c(0.47)

0.12

(0.33)

1.69

c(0.70)

1.87

c(0.71)

0.19

(0.39)

0.25

(0.43)

0.06

(0.24)

Jordan

NP

NP

1.55

c(0.71)

1.91

c(0.83)

0.19

(0.39)

0.12

(0.33)

0.14

(0.34)

Kazakhstan

1.48

c(0.50)

1.21

c(0.46)

0.06

(0.24)

0.08

c(0.28)

0.11

(0.32)

NP

NP

Korea,Republic

ofNP

NP

1.53

c(0.57)

1.73

c(0.69)

0.12

(0.32)

0.17

(0.37)

0.02

(0.15)

Kuw

ait

2.25

c(1.21)

2.64

c(1.26)

0.34

(0.47)

0.31

c(0.46)

0.26

(0.44)

1.97

c(1.00)

2.33

c(1.03)

0.19

(0.39)

0.15

(0.36)

0.11

(0.32)

Latvia

2.06

c(0.82)

2.25

c(1.01)

0.41

(0.49)

0.32

c(0.47)

0.10

(0.30)

NP

NP

Lebanon

NP

NP

1.66

c(0.62)

2.02

c(0.78)

0.22

(0.41)

0.30

(0.46)

0.27

(0.44)

Lith

uania

1.24

c(0.42)

2.04

c(0.66)

0.26

(0.44)

0.30

c(0.46)

0.20

(0.40)

1.40

c(0.53)

2.36

c(0.87)

0.15

(0.35)

0.29

(0.45)

0.10

(0.29)

Malaysia

NP

NP

1.64

c(0.60)

2.12

c(0.87)

0.09

(0.29)

0.28

(0.45)

0.08

(0.27)

Malta

NP

NP

1.79

c(0.73)

2.97

c(1.12)

0.21

(0.41)

0.26

(0.44)

0.14

(0.34)

continued

Aggr. Behav.

188 Bergmüller

Page 8: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

TABLEII.(Contin

ued)

Country

Grade

4Grade

8

Principalsa

Studentsb

Principalsa

Studentsb

Physical

Verbal

Physical

Verbal

Relational

Physical

Verbal

Physical

Verbal

Relational

Mongolia

1.16

c(0.36)

1.47

c(0.50)

0.37

(0.48)

0.33

c(0.47)

0.19

(0.39)

1.25

c(0.43)

1.85

c(0.79)

0.28

(0.45)

0.32

(0.47)

0.11

(0.32)

Morocco

1.47

c(0.61)

1.78

c(0.66)

0.31

(0.46)

0.32

c(0.47)

0.28

(0.45)

1.73

c(0.73)

2.16

c(0.96)

0.23

(0.42)

0.23

(0.42)

0.18

(0.39)

Netherlands

1.55

c(0.62)

2.23

c(0.83)

0.33

(0.47)

0.26

c(0.44)

0.23

(0.42)

NP

NP

New

Zealand

1.99

c(0.63)

2.37

c(0.80)

0.44

(0.50)

0.42

c(0.49)

0.35

(0.48)

NP

NP

Norway

1.59

c(0.63)

2.24

c(0.89)

0.25

(0.43)

0.18

c(0.39)

0.24

(0.43)

1.72

c(0.50)

2.26

c(0.81)

0.14

(0.35)

0.15

(0.36)

0.10

(0.29)

Oman

NP

NP

1.32

c(0.50)

1.61

c(0.62)

0.24

(0.43)

0.19

(0.39)

0.17

(0.38)

Palestin

ianNational

Auth.

NP

NP

1.99

c(0.96)

2.37

c(1.02)

0.21

(0.41)

0.18

(0.38)

0.17

(0.38)

Qatar

1.66

c(0.92)

2.16

c(1.11)

0.37

(0.48)

0.35

c(0.48)

0.29

(0.45)

1.57

c(0.61)

2.09

c(0.88)

0.22

(0.41)

0.19

(0.39)

0.15

(0.36)

Rom

ania

NP

NP

1.41

c(0.61)

1.86

c(0.75)

0.27

(0.44)

0.28

(0.45)

0.16

(0.36)

Russian

Federation

1.63

c(0.58)

1.25

c(0.52)

0.21

(0.41)

0.27

c(0.44)

0.23

(0.42)

1.48

c(0.51)

1.40

c(0.66)

0.10

(0.30)

0.13

(0.34)

0.13

(0.34)

Saudi

Arabia

NP

NP

1.48

c(0.64)

1.86

c(0.80)

0.20

(0.40)

0.20

(0.40)

0.15

(0.35)

Scotland

1.97

c(0.73)

2.23

c(0.89)

0.39

(0.49)

0.30

c(0.46)

0.25

(0.44)

1.95

c(0.62)

2.69

c(0.89)

0.18

(0.38)

0.25

(0.43)

0.10

(0.30)

SerbiaandMontenegro

NP

NP

1.93

c(0.72)

2.00

c(0.85)

0.14

(0.35)

0.14

(0.35)

0.11

(0.31)

Singapore

1.73

c(0.51)

1.99

c(0.60)

0.35

(0.48)

0.43

c(0.50)

0.19

(0.39)

1.75

c(0.54)

2.19

c(0.64)

0.16

(0.37)

0.30

(0.46)

0.14

(0.35)

SlovakRepublic

1.99

c(1.00)

2.04

c(0.95)

0.30

(0.46)

0.36

c(0.48)

0.17

(0.37)

NP

NP

Slovenia

1.92

c(0.58)

2.11

c(0.75)

0.42

(0.49)

0.31

c(0.46)

0.24

(0.43)

2.17

c(0.73)

2.54

c(0.81)

0.33

(0.47)

0.24

(0.42)

0.13

(0.33)

Spain

(Basque

Country)

NP

NP

1.56

c(0.50)

2.02

c(0.56)

0.06

(0.25)

0.17

(0.37)

0.07

(0.26)

Sweden

1.65

c(0.62)

2.63

c(1.03)

0.12

(0.32)

0.15

c(0.35)

0.12

(0.33)

1.80

c(0.54)

3.02

c(1.10)

0.10

(0.30)

0.14

(0.35)

0.06

(0.25)

SyrianArabRepublic

NP

NP

1.54

c(0.63)

1.82

c(0.82)

0.22

(0.41)

0.18

(0.38)

0.18

(0.38)

Thailand

NP

NP

1.59

c(0.56)

1.71

c(0.63)

0.39

(0.49)

0.29

(0.45)

0.09

(0.29)

Tunisia

1.73

c(0.78)

1.95

c(0.89)

0.39

(0.49)

0.38

c(0.49)

0.31

(0.46)

1.87

c(0.89)

2.20

c(1.05)

0.22

(0.41)

0.29

(0.45)

0.14

(0.35)

Turkey

NP

NP

NP

NP

1.70

c(0.88)

1.67

c(0.89)

0.21

(0.40)

0.33

(0.47)

0.09

(0.29)

UAE(D

ubai)

1.73

c(0.62)

2.00

c(0.77)

0.40

(0.49)

0.37

c(0.48)

0.22

(0.41)

1.67

c(0.62)

1.94

c(0.62)

0.20

(0.40)

0.25

(0.43)

0.14

(0.35)

Ukraine

1.34

c(0.51)

1.41

c(0.52)

0.28

(0.45)

0.24

c(0.43)

0.17

(0.38)

1.32

c(0.47)

1.53

c(0.63)

0.16

(0.37)

0.15

(0.35)

0.08

(0.27)

UnitedStates

1.85

c(0.69)

2.50

c(0.87)

mm

m1.87

c(0.67)

2.75

c(0.94)

mm

mYem

en1.62

c(0.75)

1.85

c(0.94)

0.27

(0.44)

0.26

c(0.44)

0.28

(0.45)

NP

NP

Note.

Standarddeviations

areinparentheses.Auth.,A

uthority;U

AE,U

nitedArabEmirates.S

cale‐range

forp

rincipalansw

ers¼

1–5;Scale‐range

forstudentansw

ers¼

0–1.NP,didnotparticipatewith

this

grade;

m,missing.

a Weightedby

school

weight.

bWeightedby

totalstudentweight.

c Missing

rate

higher

than

15%.

Aggr. Behav.

Student Aggression and Culture 189

Page 9: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

consistently across the entire range of the IDV scale. Inthe present investigation, scores ranged from 13(Colombia) to 91 (United States) with a mean of57.05 (SD ¼ 23.55). Hence, the 62 countries of thepresent investigation include almost the entire range ofvalues considered by Hofstede et al. (2010), whoclassified 76 countries, with individualism scoresranging from 6 to 91. Thus, both individualist andcollectivist countries were represented in the currentsample. Fifteen of the countries in the presentinvestigation, as noted in Table IV, were not representedin the IDV scale. These 15 countries were not included inthose computations that involved the IDV. Regions of acountry were given the index for the whole country (e.g.,Basque was given the index for Spain), with theexception of Quebec, for which a particular regionalIDV score was available.Gross domestic product per capita on pur-

chasing power parity (GDP). Figures provided bythe Human Development Report (United NationsDevelopment Programme, 2009) were used. Theserecords refer to the year 2007—the year of the TIMSSassessment. As Chinese Taipei was not included in theHuman Development Report, 2007 data from The WorldFactbook (Central Intelligence Agency, n.d.) was usedfor this country. GDP for the Palestinian NationalAuthority was not retrieved from either source and isthus set as missing. Again, the regions of a country weregiven the score for the whole country. In the presentinvestigation, GDP ranged from 1,334 (Ghana) to 74,882(Qatar; see Table IV).National variation of mathematics achieve-

ment. Standard deviations of the TIMSS 2007mathematics country means were computed for Grade4 and Grade 8. The standard deviations ranged from

61.35 (Netherlands) to 110.81 (Tunisia) for Grade 4 andfrom 59.25 (Algeria) to 108.74 (Turkey) for Grade 8 (seeTable IV).

Data Analysis

Weights. The TIMSS 2007 User Guide for theInternational Data Base (Foy&Olson, 2009) advises theuse of sample weights to prevent estimates of populationparameters being biased when conducting analyses. TheTIMSS 2007 database contained both student weightsand school weights. The TIMSS total student weightswere used to calculate country results for student‐reported data. The TIMSS school weights were used tocalculate country results for principal‐reported data. Forcountry‐level computations no weights were used.The Organization for Economic Co‐operation and

Development (2009) recommended using normalizedweights for multilevel analyses. Therefore, normalizedschool weights were computed. Normalized schoolweights are a linear transformation of the school weightsso that the sum of weights is equal to the number ofschools in the dataset. The resulting normalized schoolweights were introduced as residual weights in theregression analyses in which “school” was Level 1 and“country” was Level 2.Regression analyses: Ordinary least squares

(OLS) and robust regression. Regression analyseswere performed at the country level to test for the rela-tion between a country’s cultural individualism–collec-tivism and student aggression. Two different regressionmethods were performed: robust regression and OLSregression. Cohen, Cohen, West, and Aiken (2003)pointed out that robust approaches can be thought of as akind of insurance policy and that robust approachesshould perform better thanOLSwhen there are outliers or

TABLE III. Aggressive Student Behavior Reported by Principals and Students: Country‐Level Correlations and Country‐LevelDescriptive Statistics for Grade 4 and Grade 8

Form of aggression by reporter 1 2 3 4 5 M SD Min Max

1. Physical (principals) .85** .76** .32* .20** .12** 1.71 0.31 1.16 2.532. Verbal (principals) .74** .91** .33** .23** .23** 2.03 0.44 1.15 2.873. Physical (students) .04** .02** .70** .59** .55** 0.31 0.09 0.06 0.484. Verbal (students) .17** .28** .45** .79** .24** 0.32 0.09 0.08 0.455. Relational (students) �.02** .05** .49** .34** .64** 0.23 0.07 0.10 0.36M 1.69 2.15 0.20 0.25 0.13SD 0.26 0.46 0.07 0.09 0.06Min 1.19 1.18 0.06 0.06 0.02Max 2.39 3.15 0.39 0.51 0.33

Note. Correlations for Grade 4 countries (n ¼ 42) are presented above the diagonal, correlations for Grade 8 countries (n ¼ 55) are presented below thediagonal. Descriptive statistics for Grade 4 countries are presented in the vertical columns, and descriptive statistics for Grade 8 countries are presented in thehorizontal rows. Intercorrelations betweenGrade 4 andGrade 8 for countries which participatedwith both grades (n ¼ 32) are presented along the diagonal (inboldface). Scale range for principal reports: 1 ¼ never to 5 ¼ daily; scale range for student reports: 0 ¼ no, 1 ¼ yes; Min, minimum; Max, maximum.*P < .05.**P < .01.

Aggr. Behav.

190 Bergmüller

Page 10: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

TABLE IV. Individualism Index Score (Based on Hofstede et al., 2010), Gross Domestic Product (Based on United NationsDevelopment Programme, 2009), and Standard Deviations of TIMSS 2007 Mathematics Achievement for Grade 4 and Grade 8 byCountry

Country IDV score GDP

SD TIMSSmathematicsGrade 4

SD TIMSSmathematicsGrade 8

United States 91 45,592 75.33 76.74Australia 90 34,923 83.31 79.43England 89 35,130 86.04 83.58Scotland 89 35,130 78.93 79.73Canada (Alberta) 80 35,812 66.06 NPCanada (British Columbia) 80 35,812 71.31 72.44Canada (Ontario) 80 35,812 68.00 70.21Hungary 80 18,755 91.16 84.68Netherlands 80 38,694 61.35 NPNew Zealand 79 27,336 86.14 NPItaly 76 30,353 77.03 76.23Denmark 74 36,130 70.83 NPCanada (Quebec) 73 35,812 67.35 68.41Sweden 71 36,712 66.48 70.05Latvia 70 16,377 71.90 NPNorway 69 53,433 76.22 65.66Germany 67 34,401 68.15 NPLithuania 60 17,575 75.76 79.74Malta 59 23,080 NP 91.77Czech Republic 58 24,144 71.46 73.69Austria 55 37,370 67.94 NPIsrael 54 26,315 NP 98.87Slovak Republic 52 20,076 84.94 NPSpain (Basque Country) 51 31,560 NP 68.59Japan 46 33,632 76.08 85.42Morocco 46 4,108 95.27 80.33Iran, Islamic Republic of 41 10,955 83.52 86.09Russian Federation 39 14,690 83.37 83.08Bahrain 38 29,723 NP 83.6Egypt 38 5,349 NP 100.25Kuwait 38 47,812 99.30 78.64Lebanon 38 10,109 NP 74.64Qatar 38 74,882 90.07 93.36Saudi Arabia 38 22,935 NP 76.43United Arab Emirates (Dubai) 38 54,626 89.6 96.18Turkey 37 12,955 NP 108.74Bulgaria 30 11,222 NP 101.6Romania 30 12,369 NP 99.75Slovenia 27 26,753 71.40 71.62Malaysia 26 13,518 NP 79.25Hong Kong SAR 25 42,306 67.13 93.73Serbia and Montenegro 25 10,248 NP 89.45Ghana 20 1,334 NP 91.6Singapore 20 49,704 84.15 92.96Thailand 20 8,135 NP 91.62El Salvador 19 5,804 90.82 72.82Korea, Republic of 18 24,801 NP 92.07Chinese Taipei 17 32,600 69.23 105.51Indonesia 14 3,712 NP 87.34Colombia 13 8,587 90.18 78.94Algeria m 7,740 89.56 59.25Armenia m 5,693 89.52 84.74Bosnia m 7,764 NP 77.80Botswana m 13,604 NP 76.58Cyprus m 24,789 NP 89.32

continued

Aggr. Behav.

Student Aggression and Culture 191

Page 11: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

the residuals have a non‐normal distribution. However, ifthe assumptions for OLS regression are met, robustapproaches are less efficient than OLS regression. Thus,OLS regression is the better choice in this case. To controlfor the violation of assumptions of OLS regression, bothOLS regression and robust regression analyses werecomputed. The iteratively re‐weighted least squaresmethod (O’Leary, 1990) was used to perform robustregression analyses. OLS regression analyses werecomputed using PASW Statistics 18, and robustregression analyses using Statistics R version 2.8.1 (RDevelopment Core Team, 2008).Multilevel analyses. Multilevel data structures

are characterized by observations that are nested withinhigher‐level units or clusters. In the present study,students were nested within schools and schools werenested within countries. Multilevel analysis techniquesare well suited for these data structures because theyappropriately account for nesting and allow researchersto investigate associations at different levels of the datahierarchy (Raudenbush & Bryk, 2002).In the present study, multilevel models were primarily

used to investigate whether a relation between individu-alism and aggressive student behavior persisted aftercontrolling for other potential predictors. As associationsbetween individualism and aggressive student behaviorwere only present for principals’ reports (see ResultsSection), only two‐level models with schools at Level 1and countries at Level 2 were performed. In accordancewith the model of social–ecological influences on studentvictimization (Benbenishty & Astor, 2008), the follow-ing school‐level predictors were included: School size,size of school community, proportion of students fromdisadvantaged homes, and percentage of male studentsper school. At the country level, GDP and variation ofmathematics achievement were introduced as predictors

because these two variables have been identified assignificant predictors of student victimization by Akibaet al. (2002).The method for estimation of variance and covariance

components was full maximum likelihood, using theexpectation–maximization (EM) algorithm (Raudenbush& Bryk, 2002). Variables were not centered forcomputing unstandardized coefficients. For computingstandardized coefficients, all variables were z‐standard-ized for the total sample (Hox, 2010). PASWStatistics 18(mixed models) was used to compute multilevelanalyses.In order to compare multilevel models with a different

number of variables, it is essential to have the sameamount of valid cases for all variables. To achieve this,Organization for Economic Co‐operation and Develop-ment (2009) recommended the deletion of cases withmissing values. Therefore, those 15 countries that are notrepresented in the IDV scale (see Table IV) wereexcluded from multilevel analyses. Additionally,Norway was excluded because no data were availableon the rate of economically disadvantaged students.Therefore, multilevel analyses were conducted for 32countries for Grade 4 and 40 countries for Grade 8.At the school level, list‐wise deletion of all schools

with at least one missing value in the variables includedin multilevel models would have considerably reducedthe total number of schools. One method for dealing withmissing data, currently regarded as state of the art, ismultiple imputation (e.g., Enders, 2010; Schafer &Graham, 2002). Hence, the multiple imputation proce-dure provided by PASW Statistics 18 was used to replaceschool‐level missing data by 15 imputed data sets. Thissoftware uses the sequential regression approach as thealgorithm for multiple imputation (Enders, 2010). Thefollowing 10 variables were entered into multiple

TABLE IV. (Continued)

Country IDV score GDP

SD TIMSSmathematicsGrade 4

SD TIMSSmathematicsGrade 8

Georgia m 4,662 88.43 96.46Jordan m 4,901 NP 102.21Kazakhstan m 10,863 83.81 NPMongolia m 3,236 85.45 81.49Oman m 22,816 NP 94.94Palestinian National Authority m m NP 102.44Syrian Arab Republic m 4,511 NP 82.40Tunisia m 7,520 110.81 66.52Ukraine m 6,914 84.48 89.23Yemen m 2,335 110.14 NPM 50 22,915 80.90 84.33SD 24 15,905 11.56 11.35

Note. IDV, Individualism Index; m, missing; NP, did not participate with this grade; and GDP, gross domestic product per capita on purchasing power parity.

Aggr. Behav.

192 Bergmüller

Page 12: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

imputation models for both Grade 4 and Grade 8: Schoolsize, size of school community, proportion of studentsfrom disadvantaged homes, percentage of male studentsper school, principal‐reported frequency of physicalstudent aggression, principal‐reported frequency ofverbal student aggression, principal‐reported frequencyof student classroom disturbance, principal‐reportedfrequency of student profanity, number of tested studentsper school, and school mean of mathematics achieve-ment. A detailed description of all variables was providedby Foy and Olson (2009). Data augmentation wasperformed separately for each country and each grade.PASW Statistics 18 also provides an analysis module

for multiply imputed data and computes pooled resultswith correct standard errors. All estimates reported formultilevel models were computed with this analysismodule and are therefore pooled coefficients.

RESULTS

The Relationship Between CulturalIndividualism–Collectivism and AggressiveStudent Behavior Reported by Students andPrincipals: Country‐Level Analyses

Table V summarizes the results for models in whichprincipal‐ and student‐reported measures of aggressivestudent behavior were regressed on the level ofindividualism. To perform the following analyses,student and principal data were aggregated at the countrylevel. This means that country‐level measures ofaggressive student behavior as depicted in Table IIwere regressed on IDV scores. For better resultreadability (i.e., avoiding decimal numbers with many

zeros after the dot), original IDV scores (as depicted inTable IV) were divided by 1,000 for regression analyses.This transformation only affected unstandardized regres-sion coefficients and their standard errors, which need tobe divided by 1,000 to fit the original scale. Standardizedestimates are invariant to scale transformation.OLS and robust regression analyses yielded the same

significance of findings and provided mainly the sameestimates, which indicates evidence of the sameconclusions. In this case, OLS estimates should bepreferred (see Method Section). Therefore, I only refer toestimates of OLS regression in the following. Resultsprovided mixed support for the hypothesis that aggres-sive student behavior occurs more frequent in moreindividualist countries. The results of students’ self‐reports of being aggressively victimized in school did notsupport the hypothesis. There was no significant relationfound between the IDV and the proportion of self‐identified victims at the country level for physical, verbal,or relational aggression in either grade; Grade 4:b* ¼ .04, t(31) ¼ .25, P ¼ .808 (physical aggression),b* ¼ �.26, t(31) ¼ �1.52, P ¼ .138 (verbal aggres-sion), b* ¼ .32, t(31) ¼ 1.85, P ¼ .073 (relationalaggression); Grade 8: b* ¼ �.30, t(39) ¼ �1.95,P ¼ .059 (physical aggression), b* ¼ �.25, t(39) ¼ �1.64, P ¼ .110 (verbal aggression),b* ¼ �.08, t(39) ¼ �.49, P ¼ .627 (relational aggres-sion). In models with only one predictor variable, R2shave the same significance as regression coefficients.Therefore, R2s for student‐reported victimization werenot significant. Furthermore, they were very small in size(see Table V), which indicates that IDV scores are a verypoor predictor of student‐reported measures of

TABLE V. Country‐Level Robust and OLS Regression Analysis Predicting Aggressive Student Behavior Based on IndividualismIndex for Grade 4 (Top) and Grade 8 (Bottom)

Outcomes: Form of aggression by reporter

Robust regression OLS Regression

dfB SE B SE Adjusted R2

Grade 4Physical (principals) 5.39** 1.83 5.32** 1.88 .18 32Verbal (principals) 8.86** 2.24 9.20** 2.20 .33 32Physical (students) .35** .66 .16** .63 .00 31Verbal (students) �.78** .58 �.85** .56 .04 31Relational (students) .96** .53 .93** .50 .07 31

Grade 8Physical (principals) 5.87** 1.03 5.01** 1.21 .28 40Verbal (principals) 12.49** 2.18 12.30** 2.04 .46 40Physical (students) �.82** .52 �.96** .49 .07 39Verbal (students) �.48** .50 �.85** .52 .04 39Relational (students) .08** .33 �.20** .41 .00 39

Note. Individualism index was the only predictor for all models. For better readability of results, the individualism index scores were divided by 1,000.*P < .05.**P < .01.

Aggr. Behav.

Student Aggression and Culture 193

Page 13: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

aggressive victimization. It should be noted howeverthat, even if they are not significant, some of the effects ofstudents’ self‐reports of aggressive victimization tendindicate slightly higher verbal victimization of fourth andeighth graders in more collectivist countries and alsoslightly higher physical victimization of eighth graders inmore collectivist countries.Principals’ reports of aggressive student behavior

however clearly supported the hypothesis: The higher acountry’s IDV score was themore frequent physically andverbally aggressive behavior was perceived by principals.This was consistently shown for Grade 4: b* ¼ .45,t(32) ¼ 2.83, P ¼ .008 (physical aggression) andb* ¼ .59, t(32) ¼ 4.17, P < .001 (verbal aggression) aswell as for Grade 8: b* ¼ .55, t(40) ¼ 4.14, P < .001(physical aggression) and b* ¼ .69, t(40) ¼ 6.02,P < .001 (verbal aggression). Moreover, significantamounts of variance in principal‐reported measures ofaggressive student behaviorwere explainedby the IDV.Asdepicted in Table V, R2s ranged from .18 for physicalaggression(Grade4) to .46 forverbalaggression(Grade8).This indicates that the IDV does a good job in predictingprincipal‐reported measures of aggression. Notably, theIDV does an especially good job in predicting verbalaggressionwith33%and46%of thevarianceexplainedforGrade 4 and Grade 8, respectively.2

The Relationship Between School and CountryContexts and Principal‐Reported AggressiveStudent Behavior: Multilevel Analyses

Multilevel models were computed to examine whetherthe IDV remains a significant predictor of principal‐perceived aggressive student behavior after controllingfor potential predictors at both school and country level.Two‐level models with schools at Level 1 and countriesat Level 2 were computed. The outcome variables werethe four measures of aggressive student behaviorreported by principals (physical and verbal aggressioneach for Grade 4 andGrade 8). For each outcome variabletwo models were estimated: Model 1 was a null model,and Model 2 included potential predictors at the school

and the country level. The scores of most of thesepredictors were divided by 100 (see Table VI) to enhancethe readability of results by avoiding decimal values withtoo many zeros. Only unstandardized regression coef-ficients and their standard errors were affected by thistransformation (these need to be divided by 100 to fit theoriginal scale). Standardized estimates are not affectedand also intercepts and variance components remain thesame.There are no predictor variables in the null models, and

the intercept terms are allowed to vary randomly acrosscountries. The null models show the proportion ofvariance in principal‐perceived aggressive studentbehavior between countries. If a null model revealsthat the variance estimates at the country level aresignificant, it makes sense to explain this variance usingpredictor variables. If a null model does not reveal asignificant amount of variance at the country level, it isuseless trying to explain this nonexistent variation. Foreach form of aggression, null models revealed asignificant intercept for country variance (Table VI).This indicates that there were country‐level differences inaverage levels of principal‐reported aggressive studentbehavior. Comparing the country variance to the residualvariance (i.e., variance attributable to schools withincountries) reveals the proportion of variance betweencountries. This proportion is also referred to as intra-class‐correlation (Raudenbush & Bryk, 2002). Theintraclass correlation coefficients are .13 for physicalaggression (Grade 4), .14 for verbal aggression (Grade 4),.11 for physical aggression (Grade 8), and .18 for verbalaggression (Grade 8). That is, 11–18% of the variance inprincipal‐reported measures of student aggression wasdue to variation between countries. These results showthat principals in different countries differ considerablyin their perception of aggressive student behavior. Thismeans for further analyses, that it is reasonable to includepredictors to explain country‐level variance, as there is asignificant variance at the country level.This consideration was followed by each Model 2, in

which predictor variables at the school and at the countrylevel were introduced as fixed effects. Tomake the effectsof different variables (with different scaling) comparable,standardized coefficients are also reported. As depicted inTable VI, IDV scores were still a significant predictor foreach form of aggression after controlling for school size,size of school community, proportion of students fromdisadvantaged homes per school, and percentage of malestudents per school at the school level as well as for GDPand variation in mathematics achievement at the countrylevel. This means, the more individualist and lesscollectivist a country’s culture is, the more oftenprincipals perceive aggressive student behavior evenafter controlling for the just mentioned variables. This

2 I performed additional country‐level correlation analysis in order toinvestigate the relationship between different forms of student aggressionand the other Hofstede scales (power distance, uncertainty avoidance, andmasculinity). The Power Distance scale is very similar to the IDV scale(r ¼ �.70, P < .001, for the present sample of countries). Consequentlyalso correlational results were similar: There were significant correlationsbetween Power Distance and principal‐reported aggressive studentbehavior (except for physical aggression, Grade 4). However, thesecorrelations were always smaller than the ones between principal‐reportedaggressive student behavior and the IDV scale. Furthermore, I found asignificant correlation between masculinity and student‐reported verbalaggression for Grade 4. All of the other correlations were not significant(P > .01).

Aggr. Behav.

194 Bergmüller

Page 14: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

applies to both forms of aggression and both gradecohorts. The IDVwas not only a significant predictor, butwas also the most predictive variable compared to theincluded control variables: As depicted in Table VI, b*sfor individualism were—compared to all other variables—the highest and ranged from .17 (physical aggression,Grade 8) to .28 (verbal aggression, Grade 8). School‐level variables were mostly significant but on the wholeonly explain a small proportion of variance at the schoollevel (i.e., intercepts for school variance in Models 2 areonly slightly smaller than in the null models). Comparingb*s for all forms of aggression (see Table VI), thegenerally best predictors at the school level were theproportion of students from disadvantaged homes and thepercentage of male students in a school: The more

economically disadvantaged students and the more boysthere were in a school, the higher the frequency ofaggressive student behavior reported by principals. Theincluded control variables at the country level (GDP andSD of mathematics achievement) were very weak andmostly nonsignificant predictors of aggressive studentbehavior. GDP was only a significant predictor for verbalaggression at Grade 4 and for physical aggression atGrade 8. In these cases, principals reported aggressivestudent behavior more frequently the richer a countrywas. But effect sizes were small: b*s ¼ .16 (verbalaggression, Grade 4) and .10 (physical aggression, Grade8; see Table VI). Variation of mathematics achievementin a country was a nonsignificant predictor for each formof aggression.

TABLE VI. Results of Multilevel Models Predicting Principal‐Perceived Student Aggression From Culture, School Context, andCountry Context for Grade 4 (Top) and Grade 8 (Bottom)

Physical aggression Verbal aggression

Model 1 Model 2 Model 1 Model 2

B SE B SE b t B SE B SE b t

Grade 4Fixed effectsIntercept 1.84�� 0.05 0.78�� 0.49 — 2.19�� 0.07 1.36�� 0.50 — —

Level 1 (school)School size/100 0.02�� 0.01 .09 2.43 0.01�� 0.01 .06 1.42Comm./100 2.93�� 1.10 .07 2.68 3.11�� 1.22 .06 2.54Disadv./100 9.09�� 1.51 .15 6.02 12.74�� 1.32 .16 9.66Male % 0.40�� 0.06 .11 6.82 0.41�� 0.07 .09 5.81

Level 2 (country)IDV/100 0.74�� 0.21 .23 3.49 0.95�� 0.22 .24 4.28GDP/10,000 0.05�� 0.03 .09 1.43 0.11�� 0.04 .16 2.94SD maths/100 0.13�� 0.52 .02 0.26 �0.63�� 0.55 �.06 �1.15

Variance componentsIntercept for country variance 0.09�� 0.02 0.07�� 0.02 — — 0.14�� 0.04 0.07�� 0.02 — —

Residual variance 0.60�� 0.02 0.57�� 0.02 — — 0.85�� 0.03 0.81�� 0.03 — —

Grade 8Fixed effectsIntercept 1.74�� 0.04 1.33�� 0.30 — — 2.23�� 0.07 1.73�� .46 — —

Level 1 (school)School size/100 0.02�� 0.00 .14 5.12 0.02�� 0.00 .10 4.78Comm./100 2.02�� 0.93 .05 2.17 0.71�� 0.96 .01 0.74Disadv./100 6.36�� 1.34 .10 4.75 7.66�� 1.46 .09 5.25Male (%) 0.42�� 0.04 .14 10.37 0.36�� 0.05 .09 7.46

Level 2 (country)IDV/100 0.49�� 0.16 .17 3.16 1.11�� 0.24 .28 4.68GDP/10,000 0.04�� 0.02 .10 2.03 0.05�� 0.03 .09 1.73SD maths/100 �0.42�� 0.32 �.06 �1.30 �0.72�� 0.48 �.08 �1.50

Variance componentsIntercept for country variance 0.06�� 0.01 0.03�� 0.01 — — 0.18�� 0.04 0.08�� 0.02 — —

Residual variance 0.50�� 0.01 0.47�� 0.01 — — 0.81�� 0.02 0.79�� 0.02 — —

Note Pooled results are reported. Residuals are weighted by normalized school weight. Comm., size of school community; Disadv., proportion of studentsfrom disadvantaged homes; Male (%), percentage of male students; IDV, individualism index; GDP, gross domestic product on purchasing power parity; andMaths, mathematics achievement.�P < .05.��P < .01.

Aggr. Behav.

Student Aggression and Culture 195

Page 15: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

Discussion

This study moved the focus away from the individuallevel towards the country level in order to explain studentaggression. Associations between cultural individual-ism–collectivism and different forms of student aggres-sion were examined using the huge TIMSS 2007database which includes 62 countries and more than400,000 fourth‐ and eighth‐grade students along withtheir principals’ reports. Some questions of the back-ground questionnaires for students and principalsfocused on aggressive student behavior. Students wereasked, whether they were victims of physical, verbal, orrelational aggression; principals were asked, how oftenphysical and verbal student aggression were a problem intheir school. Using both traditional OLS regressionanalyses and robust regression analyses, I found that thelevel of individualism does not predict the percentage ofself‐identified student victims but that it does predictaggressive student behavior as reported by principals:Principals perceived aggressive student behavior morefrequently, the more individualist, and hence lesscollectivist, their country’s culture. These results werehighly consistent with regard to different grade cohorts(Grade 4 and Grade 8) and to different forms ofaggression. Moreover, multilevel analysis showed thatIDV scores are still a significant predictor of principal‐reported aggressive student behavior after controlling forschool‐ and country‐level variables that have previouslybeen found to predict aggressive student behavior. Theseresults were consistent across different forms of aggres-sion and across different grade cohorts.

Why do Results Differ by Reporter’sPerspective?

The present results only provide partial support for thehypothesis that aggressive student behavior occurs morefrequently the more individualist, and hence less collec-tivist, a country’s culture is. Data on aggressive victimiza-tion obtained from students did not support the hypothesis(with somemeasures even leaning slightly in the oppositedirection), whereas data on aggressive student behaviorobtained from principals consistently supported thehypothesis, even after other potential predictors werecontrolled for. This raises the question of why individual-ism–collectivism was related to the principals’ reports ofaggressive student behavior but not the students’ self‐reports of aggressive victimization. Before elaboratingmoreonthereasons for this, it shouldbenoted thatdifferentreporters (teachers/peers/observers) often report differentfrequencies of the same negative student behavior (e.g.,Tomada & Schneider, 1997; Weisz, Chaiyasit, Weiss,Eastman, & Jackson, 1995). Furthermore, student self‐reported victimization by other students and principals’

reports on aggression amongst the students in their schoolis actually different constructs. However, both constructsare indicators of aggressive student behavior that relate tothe same forms of aggression (physical and verbal).Therefore, the question of why the relationship betweencultural individualism–collectivism and aggressive stu-dent behavior differs dependingon the informant hasyet tobe answered.One reason for this discrepancy may lie in the

formulation of context‐free items in TIMSS.3 As outlinedin the method section, the TIMSS aggression items didnot include a description of the context in which thebehavior occurred, and in particular, students andprincipals were not asked whether or not they perceivedthe negative student behavior as intended by the offender.Ignoring the offender’s intention is a limitation toaggression research because aggressive behavior isdefined by the intention to harm another person (DeWallet al., 2011). Therefore, it is only the offender’s intent toharm that makes harmful behavior aggressive and onlythe victim’s perception of the offender’s intent to harmprovides the experience of victimization. Moreover,ignoring the offender’s intent may create a culture‐specific reporting bias. As Cheng et al. (2013) describedin their meta‐analysis, members of individualist societiestend to consider the outcomes of events to be contingentupon their own actions, under personal control, andrelatively independent of the context. In contrast,members of collectivist societies are likely to view eventoutcomes as strongly influenced by outside forces andlargely determined by the context, particularly theinterpersonal context, in which they occur.Because members of individualist societies tend to

perceive any demonstrated behavior as intended by theactor, they are more likely to describe any harmfulbehavior as aggressive (i.e., intended by the offender).Members of collectivist societies however tend to see anydemonstrated behavior as largely influenced by thesituation and therefore do not tend to consider harmfulbehavior as intended by the offender and henceaggressive (cf. Benderlioglu, 2003). If information aboutthe context had been included in the present items, Iwould expect a) members of collectivist societies toreport a lower frequency of aggression (than actuallymentioned) because they tend to see behavior as largelydetermined by the context and b) members of individu-alist societies to report approximately the same frequencyof aggression as mentioned because they tend to seebehavior as largely as intended by the actor.In the present study, items from both students and

principals were formulated context free. The lack of

3 I thank an anonymous reviewer for this comment.

Aggr. Behav.

196 Bergmüller

Page 16: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

context however may have had less effect on principalresponses than student responses.Principals usuallydonotwitness aggressive student behavior themselves, butaggressive acts are usually observed by other schoolpersonnel (mostly teachers)who then inform the principal.It also seems unlikely that the incidents reported to theprincipals are a random sample of all aggressive behaviorthat occurs in the school. Instead, these acts of aggressionare probably more salient (more serious and/or morefrequent) thanusual.That is, principalsonlybecomeawareof behavior that is clearly outside of the norm. Moreimportantly, the aggressive acts reported by principleswere identified by at least one reasonably neutral observerand may therefore have been identified more accuratelythan the acts remembered and reported by victims.Consequently, I regard principal reports as more valid

indicators of aggressive student behavior than studentself‐reports of victimization. There is however somesupport for the notion that students providemore accuratereports than principals. I describe these reasons in thefollowing along with some critical statements as to why Iconsider them largely implausible.First, the frequencies of aggressive student behavior

that teachers report to principals might be culturallybiased. It could be that teachers in collectivist societiesreport aggressive student behavior less often becausethey are more likely to take into account the context inwhich the behavior occurred and view the behavior asless severe. A study by Weisz, Suwanlert, Chaiyasit,Weiss, and Walter et al. (1988) indeed showed thatteachers and parents in collectivist Thailand rated achild’s problems like disobedience and fighting as lessserious, less worrisome, less likely to reflect personalitytraits, and more likely to improve with time than theirindividualist US counterparts. However, these ratingswere not reflected by the frequency with which Thaiteachers report aggressive behavior: Thai teachersactually reported student problem behavior more oftenthan American teachers (Weisz, Suwanlert, Chaiyasit,Weiss, Achenbach et al., 1988; Weisz et al., 1995) eventhough structured observation showed lower levels ofproblem behavior for Thai students than US students(Weisz et al., 1995). Instead, these findings are in linewith cultural normativeness theory (e.g., Dodge, Coie, &Lynam, 2006), which maintains that when viewed in thecontext of various cultures or social contexts, the sameact might be classified as aggressive or nonaggressivedepending on how it contrasts with base rates for similarbehavior in each of these contexts. In this respect,teachers in low aggressive (collectivist) cultures mayeven consider lower levels of aggression out of norm andreport them to their principals. Therefore, the fact thatprincipals in collectivist countries report aggressivestudent behavior less often cannot be explained by a

reporting bias as evidence indicates that teachers incollectivist countries tend to report aggressive studentbehavior more often.Second, it is possible that the students’ reports are more

accurate than the principals’ if students in collectivistsocieties perform more covert and less overt forms ofaggression. Here, teachers in collectivist societies couldmake fewer reports to their principals because theywitnessless overt forms of aggressive student behavior. Indeed,aggressive behavior in the presence of teachers is strictlyprohibited in collectivist societies (e.g., China; Chen andFrench, 2008). Moreover, victims in collectivist culturesmay be less likely to report their experiences to teachers(e.g., Kanetsuna, Smith, & Morita, 2006). Because theopen expression of aggressive behavior is less accepted incollectivist societies than individualist societies, Forbeset al. (2009) formulated the hypothesis that more covert(relational) forms of aggression occur more frequently incollectivist societies whereas direct forms (physical andverbal aggression) occur less frequently. Althoughsupporting evidence can be found in a study by Murray‐Harvey, Slee, and Taki (2010), which described forms ofrelational aggression to be more frequent in collectivistJapan than in individualist Australia, Forbes et al. (2009)were unable to find any empirical evidence for theirhypothesis. In a three‐country study including the UnitedStates, Poland, and China, Forbes et al. (2009) found ratesof indirect aggression to be lowest in China and highest inthe United States, thereby contradicting the initialhypothesis. Moreover, a study by Lansford et al. (2012)showed that direct and relational forms of aggression arehighly correlated at the country level but unrelated tocollectivism–individualism. In this nine‐country study,physical and relational aggression were over‐representedin collectivist Kenya and Jordan and underrepresented incountries asdifferent asChina,Columbia, Sweden, and theUnited States. Also, the present results showed thatrelational victimization reported by students is not morefrequent in collectivist societies than individualist socie-ties. There is even a small (nonsignificant) effect in theopposite direction which indicates that relational victimi-zation is more frequent in individualist societies, under-pinning the results of Forbes et al. (2009). As such, thepresent study did not find any support for the notion thatstudents in collectivist countries enact more relationalforms of aggression that are not perceivable by theirteachers and thus not reported to and perceived byprincipals.All in all, there is no convincing evidence for the notion

that student reports are more accurate than principalreports; on the other hand however, it seems highlyplausible that principal reports are more valid becauseprincipals are only made aware of aggressive studentbehavior that has been observed and reported to them by

Aggr. Behav.

Student Aggression and Culture 197

Page 17: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

relatively neutral observers (teachers or other schoolpersonnel). These observers probably only report morefrequent and/or more severe acts of aggression whichconvey a clear intent to harm rather than all of the harmfulstudent behavior they are aware of. Therefore, it is likelythat principal reports are less biased by the use of contextfree items than student responses because students mayalso report unintended harmful behavior. Overall Iconsider the results based on principal reports to bemore valid than the results based on student reports.Moreover, the results based on principals’ reportsshowed that there is a significant relationship betweenrates of aggressive student behavior and culture:Aggression rates were found to be higher in moreindividualist and hence less collectivist countries andthis relationship persisted even after controlling for anumber of context characteristics at the school andcountry level.

Strengths and Limitations

The present study makes a substantial step forward ininvestigating the relationship between cultural values andaggressive behavior. Preceding studies on the relationbetween cultural individualism and aggression (Bergeron& Schneider, 2005; Forbes et al., 2009) have notsystematically distinguished between different reportingperspectives, were rather limited in the range of countries(e.g., no Arab or African countries were included), anddid not include representative samples of individuals forthe countries under investigation.The present study was based on a broad empirical basis

of 62 countries across the globe, and more than 400,000students and their principals. The comparability of thesamples between countries and grades was ensured as“all participants [i.e., countries] followed the uniformsampling approach specified by the TIMSS 2007 sampledesign with minimum deviations” (Joncas, 2008, p. 77).The specified sampling procedure also ensured represen-tative samples of fourth‐ and eighth‐grade students foreach country.A potential limitation with regard to the included

countries is that not all of the 62 countries participated inthe testing at both Grade 4 and Grade 8. Therefore, thesample of countries was not the same when comparingfourth‐ and eighth‐grade results. However, as the resultsof Grade 4 and Grade 8 were highly consistent, theanalysis of different countries for Grade 4 and Grade 8may not be problematic. Another limitation related tocountries is that the IDV score was not available for 15nations. Therefore, these nations were excluded from allcomputations using the IDV.In the present study, culture was operationalized using

IDV scores derived from Hofstede et al.’s (2010) culturalvalues framework. These scores are assigned at the

country level and thus represent a country‐levelaggregation of individuals’ characteristics. Within‐society variance in the endorsement of cultural valuesmay be larger for some countries, especially countrieswith a great diversity of ethnic groups (e.g., United Statesand Australia) and those collectivist countries that haveundergone rapid modernization (e.g., Mainland China,and South Korea; see Cheng et al., 2013). In futurestudies, culture should therefore be additionally assessedat an individual level next to the society level.Another potential limitation is that each form of

aggressive student behavior was measured using onlyone survey item per informant and grade. It must be notedthat a small number of items is a potential source ofconstruct bias (Van de Vijver & Tanzer, 1997).Moreover,the TIMSS aggression items did not specify theinterpersonal context in which the harmful studentbehavior occurred. Particularly, the items did not describewhether or not the harmful behavior was intended by theoffender. This is especially important to cross‐culturalaggression research because members of collectivistcountries were shown to rely more on context whereasmembers of individualist societies judge behavior moreas intentional and under control of the performer. Futurecross‐cultural studies on aggression should try toovercome these limitations by (a) using multiple itemsto assess aggressive student behavior, by (b) includingthe intention/purpose of the offender, (c) by specifyingthe interpersonal context in which the aggressivebehavior occurred, and (d) by including structuredbehavior observation alongside students’ and principals’or teachers’ reports. Such a multi‐method approachwould allow for a more detailed understanding of howthe rates of aggressive student behavior reported bystudents and principals relate to actual rates.The present study brought new, systematic insight into

student aggression across different cultures. Relying onprincipals’ reports of aggressive student behavior, it wasshown that countries’ dominant cultural values (individ-ualism vs. collectivism) were definitely related tofrequencies of aggressive student behavior. Futurestudies should test whether this relationship generalizesacross different age groups, across gender and acrossdifferent subpopulations in multicultural societies.

ACKNOWLEDGEMENTS

I am grateful to Alexander Robitzsch and UlrikeKipman for their methodological advice and to RebekkaWanka and Angela Whale for language editing.

REFERENCES

AFPPD/UNFPA. 2003. Violence against women in South Asia—A regionalanalysis. Retrieved from http://www.unfpa.org.np/pub/vaw/VAW_RE-G_Analysis.pdf

Aggr. Behav.

198 Bergmüller

Page 18: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

Akiba, M., LeTendre, G., Baker, D., & Goesling, B. (2002). Studentvictimization: National and school system effects on school violence in37 nations.American Educational Research Journal, 39, 829–853. doi:10.3102/00028312039004829

Benderlioglu ZA. 2003. Perception of hostility and blameworthiness, anger,and aggression in the US, Turkey, and China (Doctoral dissertation, TheOhio State University). Retrieved from http://etd.ohiolink.edu/send‐pdf.cgi/Benderlioglu%20Zeynep%20A.pdf?osu1054591695

Bandura, A. 1973. Aggression: A social learning analysis. EnglewoodCliffs, NJ: Prentice‐Hall.

Benbenishty, R., & Astor, R. A. (2008). School violence in an internationalcontext: A call for global collaboration in research and prevention.International Journal of Violence and School, 7, 59–80.

Bergeron, N., & Schneider, B. H. (2005). Explaining cross‐nationaldifferences in peer‐directed aggression: A quantitative synthesis.Aggressive Behavior, 31, 116–137. doi: 10.1002/ab.20049

Bergmüller, S., & Wiesner, C. (2009). Schulische Gewalt‐ und Aggres-sionserfahrungen 15‐/16‐Jähriger [Experiences of school violence andaggression of 15‐/16‐year‐old]. In: C. Schreiner & U. Schwantner(Eds.), PISA 2006: Österreichischer Expertenbericht [PISA 2006:Austrian expert report] (pp. 314–320). Graz, Austria: Leykam.

Bergmüller, S., & Wiesner, C. (2012). Aggressives Schülerverhalten:Auftretenshäufigkeit und Zusammenhang mit schulischer Leistungsse-lektion [Aggressive student behavior: Prevalence and its relation toschool tracking]. In: F. Eder (Hrsg) (Ed.) PISA 2009. NationaleZusatzanalysen für Österreich [PISA 2009. Additional national analysesfor Austria] (pp. 131–165). Münster, Germany: Waxmann.

Bronfenbrenner, U. 1979. Basic concepts. In: U. Bronfenbrenner (Ed.) Theecology of human development (pp. 16–44). Cambridge, MA: HarvardUniversity Press.

Brown, D. E. 2004. Human universals, human nature, human culture.Daedalus, 133, 47–54. doi: 10.1162/0011526042365645

Central Intelligence Agency. (n.d.). The world factbook: GDP—per capita(PPP). Retrieved from http://hdrstats.undp.org/en/indicators/91.html

Chen, X., & French, D. (2008). Children’s social competence in culturalcontext. Annual Review of Psychology, 59, 591–616.

Chen, X., Hastings, P., Rubin, K. H., Chen, H., Cen, G., & Stewart, S. L.(1998). Childrearing attitudes and behavioral inhibition in Chinese andCanadian toddlers: A cross‐cultural study. Developmental Psychology,34, 677–686.

Cheng, C., Cheung, S‐F., Chio, J. H‐M., & Chan, M‐P. S. (2013). Culturalmeaning of perceived control: A meta‐analysis of locus of control andpsychological symptoms across 18 cultural regions. PsychologicalBulletin, 139, 152–188. doi: 10.1037/a0028596

Cole, P.M., Tamang, B. L., & Shrestha, S. (2006). Cultural variations in thesocialization of young children’s anger and shame.Child Development,77, 1237–1251. doi: 10.1111/j.1467‐8624.2006.00931.x

Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multipleregression/correlation analysis for the behavioral sciences (3rd ed.)Mahwah, NJ: Erlbaum.

Crystal, D. S., Chen, C., Fuligni, A. J., Stevenson, H.W., Hsu, C. C., Ko, H.J. … Kimura, S., (1994). Psychological maladjustment and academicachievement: A cross cultural study of Japanese, Chinese, andAmerican high school students. Child Development, 65, 738–753.doi: 10.1111/j.1467‐8624.1994.tb00780.x

DeWall, C. N., Anderson, C. A., & Bushman, B. J. (2011). The generalaggression model: Theoretical extensions to violence. Psychology ofViolence, 1, 245–258. doi: 10.1037/a0023842

Dodge, K. A., Coie, J. D., & Lynam, D. (2006). Aggression and antisocialbehavior. In: N. Eisenberg (Ed.)Handbook of child psychology, social,emotional, and personality development (6th ed., Vol. 3, pp. 719–788).New York, NY: Wiley.

Enders, C. K. 2010. Applied missing data analysis. New York: TheGuilford Press.

Erberber,E.,Arora,A.,&Preuschoff,C. (2008).Developing theTIMSS2007backgroundquestionnaires. In: J. F.Olson,M.O.Martin,& I.V.S.Mullis(Eds.), TIMSS 2007 technical report (pp. 45–62). Chestnut Hill, MA:Boston College.

Forbes, G. B., Zhang, X., Doroszewicz, K., & Haas, K. B. (2009).Relationships between individualism–collectivism, gender, and director indirect aggression: A study in China, Poland, and the USAggressiveBehavior, 35, 24–30. doi: 10.1002/ab.20292

Foy, P., & Olson, J. F. (Eds.), (2009). TIMSS 2007 user guide for theinternational database. Chestnut Hill, MA: Boston College.

Fuchs, M. 2008. Impact of school context on violence at schools: A multi‐level analysis. International Journal of Violence and School, 7, 20–42.

Hawker, D., & Boulton, M. (2000). Twenty years’ research on peervictimization and psychosocial maladjustment: A meta‐analytic reviewof cross‐sectional studies. Journal of Child Psychology and Psychiatry,41, 441–455. doi: 10.1111/1469‐7610.00629

Hofstede, G. H. 2001. Culture’s consequences: Comparing values,behaviors, institutions, and organizations across nations (2nd ed.).Thousand Oaks, CA: Sage.

Hofstede, G. H. 1984. Culture’s consequences: International differences inwork‐related values. Newbury Park, CA: Sage.

Hofstede, G. H., Hofstede, G. J., & Minkov, M. (2010). Cultures andorganizations: Software of the mind (3rd ed.) NewYork, NY:McGraw‐Hill.

Hox, J. J. 2010.Multilevel analysis: Techniques and applications (2nd ed.).New York, NY: Routledge.

Johansone, I. 2008. Quality assurance in the TIMSS 2007 data collection. In:J. F. Olson, M. O.Martin, & I. V. S.Mullis (Eds.), TIMSS 2007 technicalreport (pp. 113–140). Chestnut Hill, MA: Boston College.

Johansone, I., & Malak, B. (2008). Translation and national adaptations ofthe TIMSS 2007 assessment and questionnaires. In: J. F. Olson, M. O.Martin, & I. V. S. Mullis (Eds.), TIMSS 2007 technical report. (pp. 63–75). Chestnut Hill, MA: Boston College.

Johansone, I., & Neuschmidt, O. (2008). TIMSS 2007 survey operationsprocedures. In: J. F. Olson,M. O.Martin, & I. V. S.Mullis (Eds.), TIMSS2007 technical report. (pp. 93–112). ChestnutHill,MA:BostonCollege.

Joncas,M. 2008. TIMSS 2007 samplingweights and participation rates. In:J. F. Olson, M. O. Martin, & I. V. S. Mullis (Eds.), TIMSS 2007technical report (pp. 153–192). Chestnut Hill, MA: Boston College.

Kanetsuna, T., Smith, P. K., & Morita, Y. (2006). Coping with bullying atschool: Children´s recommended strategies and attitudes to school‐based interventions in England and Japan. Aggressive Behavior, 32,570–580. doi: 10.1002/ab.20156

Kim, U., Triandis, H. C., Kagitçibasi, C., Choi, S.‐C., & Yoon, G. (1994).Individualism and collectivism: Theory, method, and applications.Thousand Oaks, CA: Sage.

Lansford, J. E., Skinner, A. T., Sorbring, E., Di Giunta, L., Deater‐Deckard,K., Dodge, K. A.…, Chang, L., (2012). Boys’ and girls’ relational andphysical aggression in nine countries. Aggressive Behavior, 38, 298–308. doi: 10.1002/ab.21433

Mullis, I. V. S., & Martin, M. O. (2008). Overview over TIMSS 2007. In:J. F. Olson, M. O. Martin, & I. V. S. Mullis (Eds.), TIMSS 2007technical report. (pp. 1–12). Chestnut Hill, MA: Boston College.

Murray‐Harvey, R., Slee, P. T., & Taki, M. (2010). Comparative and cross‐cultural research on school bullying. In: S. R. Jimerson, S. M. Swearer,& D. L. Espelage (Eds.), Handbook of bullying in schools (pp. 35–47).New York: Routledge.

Nisbett, R. E., & Cohen, D. (1996). Culture of honor: The psychology ofviolence in the South. Denver, CO: Westview Press.

O’Leary, D. 1990. Robust regression computation using iterativelyreweighted least squares. SIAM Journal on Matrix Analysis andApplications, 11, 466–480.

Organization for Economic Co‐operation and Development. (2009). PISAdata analysis manual—SPSS (2nd ed.). Paris, France: Author.

Aggr. Behav.

Student Aggression and Culture 199

Page 19: The Relationship Between Cultural Individualism-Collectivism and Student Aggression Across 62 Countries

Oyserman, D., Coon, H., & Kemmelmeier, M. (2002). Rethinkingindividualism and collectivism: Evaluation of theoretical assumptionsand meta‐analyses. Psychological Bulletin, 128, 3–73. doi: 10.1037//0033‐2909.128.1.3

R Development Core Team. (2008). R: A language and environment forstatistical computing. Vienna, Austria: R Foundation for StatisticalComputing.

Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models:Applications and data analysis methods (2nd ed.). ThousandOaks, CA:Sage.

Rubin, K. H., Burgess, K. B., & Coplan, R. J. (2002). Social withdrawaland shyness. In: P. K. Smith&C. H. Hart (Eds.),Blackwell handbook ofchildhood social development (pp. 330–352). Mladen, MA: Blackwell.

Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the stateof the art. Psychological Methods, 7, 147–177. doi: 10.1037/1082‐989X.7.2.147

Tomada, G., & Schneider, B. H. (1997). Relational aggression, gender, andpeer acceptance: Invariance across culture, stability over time, andconcordance among informants. Developmental Psychology, 33, 601–609. doi: 10.1037/0012‐1649.33.4.601

Triandis, H. C. 2004. Culture and social behavior. New York: McGraw‐Hill.

Triandis, H. C., & Suh, E. M. (2002). Cultural influences on personality.Annual Review of Psychology, 53, 133–160. doi: 10.1146/annurev.psych.53.100901.135200

United Nations Development Programme. 2009. Human DevelopmentReport: GDP per capita (PPP US$). Retrieved from http://hdrstats.undp.org/en/indicators/91.html

Van de Vijver, F. J. R., & Tanzer, N. K. (1997). Bias and equivalence incross‐cultural assessment: An overview. European Review of AppliedPsychology, 47, 263–280.

Weisz, J. R., Chaiyasit, W., Weiss, B., Eastman, K. L., & Jackson, E. W.(1995). A multimethod study of problem behavior among Thai andAmerican children in school: Teacher reports versus direct observa-tions. Child Development, 66, 402–415. doi: 10.1111/j.1467‐8624.1995.tb00879.x

Weisz, J. R., Suwanlert, S., Chaiyasit, W., & Walter, B. R. (1987). Over‐and undercontrolled problems among Thai and American children andadolescents: The wat and wai of cultural differences. Journal ofConsulting and Clinical Psychology, 55, 719–726. doi: 10.1037/0022‐006X.55.5.719

Weisz, J. R., Suwanlert, S., Chaiyasit, W., Weiss, B., Achenbach, T. M., &Trevathan, D. (1988). Epidemiology of behavioral and emotionalproblems among Thai and American children: Teacher reports for ages6–11. Journal of Child Psychology and Psychiatry, 30, 471–484. doi:10.1111/j.1469‐7610.1989.tb00260.x

Weisz, J. R., Suwanlert, S., Chaiyasit, W., Weiss, B., Walter, B. R., &Anderson, W.W. (1988). Thai and American perspectives on over‐ andunder‐controlled child behavior problems: Exploring the thresholdmodel among parents, teachers, and psychologists. Journal ofConsulting and Clinical Psychology, 56, 601–609. doi: 10.1037/0022‐006X.56.4.601

Zahn‐Waxler, C., Fiedman, R. J., Cole, P. M., Mizuta, I., & Hiruma, N.(1996). Japanese and United States preschool children’s responses toconflict and distress.Child Development, 67, 2462–2467. doi: 10.1111/j.1467‐8624.1996.tb01868.x

Aggr. Behav.

200 Bergmüller