er

Vikfur

is geneactices are cross-nationally equivalent.evel changes its meaning.ith valu

Teacher self-efcacyCross-national

Thus, mean score differences between countries mainly reect cultural value orientations and responsestyles.

surveys, such as the Teaching and Learning International Survey(TALIS), initiated by the OECD (Organisation for Economic Co-operation and Development; OECD, 2009), and the Teacher Edu-cation and Development Study in Mathematics (TEDS-M; Tattoet al., 2008). This raises the question to what extent the positiveassociations of teacher self-efcacy with, among other things,

st, the question isat is largely rootedl contexts as well.the ndings fromd at country level.nd cross-nationalvement tests? Ademonstration of

the universal relevance and comparability of the concept of teacherself-efcacy. We set out to examine these, as yet, untested pre-mises. We report a secondary analysis of data from OECDs TALIS(OECD, 2009, 2010b), complemented by several other cross-national data sets. First, we compare the structure and psycholog-ical function of the teacher self-efcacy construct across 23 coun-tries. Second, we examine whether variations at the teacher- andcountry-levels have the same meanings, and we analyze thenomological network of country-level differences.

* Corresponding author. Tel.: 49 (0)69 24708 222.

Contents lists available at

ea

.e

Teaching and Teacher Education 35 (2013) 92e103E-mail address: vieluf@dipf.de (S. Vieluf).Being convinced of ones own ability is a fundament of humanagency, also for teachers. A large body of research shows teacherself-efcacy to be closely related not only with teachers well-being(e.g., Brouwers & Tomic, 2000), but also with their professionalpractices (e.g., Gibson & Dembo, 1984) and the educational out-comes of their students (e.g., Anderson, Greene, & Loewen, 1988).On account of its high relevance for teaching and learning, teacherself-efcacy has recently been included in international large scale

policy. Two questions need to be addressed. Firwhether the construct of teacher self-efcacy thin US research can be applied in other culturaSecond, it needs to be questioned whetherstudies on the individual level can also be founCan teacher self-efcacy help us to understaperformance differences on educational achieprerequisite for answering these questions is a1. Introduction student performance and job satisfaction that were found in earlierstudies really inform us on possible implications for educationalStructural equivalenceCultural valuesResponse styles

2013 Elsevier Ltd. All rights reserved. Associations with other beliefs and pr Aggregating the scale to the country-l Country-level variation is explained w

a r t i c l e i n f o

Article history:Received 4 November 2012Received in revised form22 May 2013Accepted 27 May 2013

Keywords:0742-051X/$ e see front matter 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.tate.2013.05.006e orientations and response styles.

a b s t r a c t

In the present study, teacher self-efcacy was examined in a cross-national setting. The cross-nationalgeneralizability of the scale and the meaning of cross-national variation in mean scores were investi-gated. Using data from TALIS involving 73,100 teachers in 23 countries, teacher self-efcacy was shownto have a similar unifactorial structure and equivalent positive correlations with teaching practices andjob satisfaction across countries. At the country level, signicant correlations were only found for jobsatisfaction; in addition, teacher self-efcacy was related to collectivism, modesty, and extremity scoring.We examined teacher self-efcacy in a The unifactorial structure of the scale ralizable across countries.cross-national setting.Teacher self-efcacy in cross-national p

Svenja Vieluf a,*, Mareike Kunter b, Fons J.R. van deaGerman Institute for International Educational Research, Schlostrae 29, 60486 FranbGoethe University Frankfurt am Main, Germanyc Tilburg University, The Netherlands

h i g h l i g h t s

Teaching and T

journal homepage: wwwAll rights reserved.spective

jver c

t am Main, Germany

SciVerse ScienceDirect

cher Education

lsevier .com/locate/ tate

ache1.1. Teacher self-efcacy

The construct of teacher self-efcacy is grounded largely in twoinuential psychological theories of the 20th century: Locus ofControl (Rotter, 1966) and Social Cognitive Theory (Bandura, 1977,1986). In Locus of Control Theory teacher self-efcacy isdescribed as the belief in the ability to affect student performanceover and above the inuences of students home environments(Armor et al., 1976). This belief may depend on teachers evalua-tions of their ability to perform behaviours effective towardsreaching this goal, but it may also depend on the conviction thatstudent performance is malleable by teachers. In the present articlewe refer to Banduras (1977) Social Cognitive Theory that distin-guishes between efcacy and outcome expectations. Based on thistheory, self-efcacy of teachers can be dened as individual beliefsin their capabilities to perform specic teaching tasks at a speciedlevel of quality in a specied situation (Dellinger, Bobbett, Olivier,& Ellett, 2008, p. 4).

Teacher self-efcacy reects previous experiences, beliefs, andbehaviour (e.g., Raudenbush, Rowan, & Fai Cheong, 1992) and it is apowerful predictor of future behaviour, especially of classroomteaching practices (e.g., Gibson & Dembo, 1984; Smylie, 1988).Moreover, it is associated with outcomes, such as teacher burnoutand job satisfaction (e.g., Brouwers & Tomic, 2000; Caprara,Barbaranelli, Borgogni, & Steca, 2003; Klassen & Chiu, 2010; Mo,Pazzaglia, & Ronconi, 2010; Skaalvik & Skaalvik, 2007) as well asstudent achievement (e.g., Anderson et al., 1988; Ashton & Webb,1986; Caprara, Barbaranelli, Steca, & Malone, 2006; Ross, 1992;for a review, see Tschannen-Moran, Woolfolk Hoy, & Hoy, 1998).

1.2. The cross-national generalizability of teacher self-efcacy

Teacher self-efcacy could be inuenced by national cultures inseveral ways. First, the basic structure of the construct may beculturally contingent, which would imply that behaviours and be-liefs associated with teacher self-efcacy would vary across cul-tures and that there is no basis for comparing the construct acrossnations. Second, the strength of associations with educationalprocesses and outcomes may vary, which would suggest that thepsychological and practical relevance of the construct varies acrosscountries. Third, we may observe cross-national differences inaverage teacher self-efcacy, which could reect genuine cross-national differences as well as differences in self-presentationalnorms.

In discussing cross-national ndings, we distinguish betweenstructure- and level-oriented studies (van de Vijver & Leung, 1997).The former refers to studies involving a comparison of the factorstructure and of relationships with other variables (e.g., the asso-ciation between teacher self-efcacy and job satisfaction), whereasthe latter refers to comparisons of means (e.g., cross-nationalcomparisons of teacher self-efcacy scores).

1.2.1. Structure-oriented comparisons of teacher self-efcacyBanduras (1997) view that self-efcacy is universal has been

challenged (e.g., Pajares, 2007; Schooler, 1990). It has been arguedthat the evaluation of ones capabilities is more important to oneswell-being and satisfaction in individualistic cultures where tiesbetween individuals are rather loose and an I consciousness,autonomy, and individual needs and rights are valued. In contrast,collectivistic cultures stress a We consciousness, collective iden-tity, interpersonal connectedness, harmony, solidarity, duty, andconformity. This emphasis on the in-group may result in individualself-evaluations having a weaker impact on a persons well-beingas compared to the evaluation of the own group (Kim, Triandis,

S. Vieluf et al. / Teaching and TeKagtbas, Choi, & Yoon, 1994; Markus & Kitayama, 1991).Bandura (2002) opposed this view, arguing personal efcacy isvalued not because of reverence for individualism but because astrong sense of efcacy is vital for successful adaptation and changeregardless of whether it is achieved individually or by groupmembers working together (p. 272). In Banduras view, no cross-national differences in teacher self-efcacy would be expectedalong the individualism-collectivism dimension. This debate hasfar-reaching consequences for the study of teacher self-efcacy:Applying the construct in non-Western cultures is only appro-priate if it has a basically similar structure, psychological function,and effect on educational outcomes across countries.

The few empirical studies that have examined the universalityof the psychological structure of teacher self-efcacy yieldedinconsistent results. Klassen et al. (2009) employed questionnaireitems designed to measure teacher self-efcacy in a sample of 1212elementary/middle school and secondary school teachers in vecountries: Canada, Cyprus, Korea, Singapore, and the United States.Across these countries a common three factor structure was iden-tied (teacher self-efcacy for instructional strategies, teacher self-efcacy for student engagement, and teacher self-efcacy forclassroom management). The study also showed cross-nationallysimilar patterns of correlations between teacher self-efcacy andjob satisfaction. A cross-nationally similar factorial structure ofteacher self-efcacy was also reported by Ho and Hau (2004), whocompared the factorial structure of a teacher self-efcacy instru-ment comprising four subscales (personal instruction efcacy,personal discipline efcacy, personal guidance efcacy, and beliefsabout external inuences) in a sample of 246 secondary schoolteachers from Australia and Hong Kong. However, contrary toKlassen et al. (2009), they found cross-national variation in thecorrelations between different sub-dimensions. Considerablecross-national differences in the factor structure of teacher self-efcacy were reported by Lin, Gorrell, and Taylor (2002). Theyapplied the Gibson and Dembo (1984) teacher self-efcacy instru-ment to 481 preservice early childhood or elementary schoolteachers in Taiwan and the United States. The original two scales(personal teaching efcacy and general teaching efcacy) could notbe replicated in Taiwan, suggesting that teachers in both countrieshave different conceptions of their own efcacy.

1.2.2. Level-oriented comparisons of teacher self-efcacyResults of previous comparisons of mean scores for teacher self-

efcacy across countries are more consistent than those ofstructure-oriented studies. East Asian teachers are frequentlyshown to have lower average self-efcacy scores than their westerncounterparts (e.g., Ho & Hau, 2004; Klassen et al., 2009; Lin et al.,2002). However, the interpretation of these differences is still alargely unresolved issue. When questionnaire scales are aggregatedto the country level, the often implicit assumption is made that theaggregate score has ameaning similar to that at the individual level.This would mean that countries with higher scores of teacher self-efcacy have teachers who actually feel more efcacious. Uponcloser scrutiny, this implication is far from evident. At the indi-vidual teacher level, teacher self-efcacy is a judgement of onesown teaching ability which, as argued before, is related to theteachers level of job satisfaction, classroom teaching practices, andstudent outcomes. If the aggregate score had an equivalent mean-ing, similar relationships would be expected at the country level.The few studies that have compared teacher self-efcacy interna-tionally, however, found comparatively low average scores forteachers in East Asian sites, such as Taiwan (Lin et al., 2002),Singapore (Klassen et al., 2009), and Hong Kong (Ho & Hau, 2004).As repeatedly found in studies conducted by the OECD, these ed-ucation systems traditionally have high student performance levels

r Education 35 (2013) 92e103 93in international comparisons (e.g., OECD, 2010a) and teachers with

achecomparatively high levels of professional knowledge (Blmeke,Kaiser, & Lehmann, 2010). This seeming paradox (lower teacherself-efcacy in countries with higher teacher performance) sug-gests that country and individual differences in teacher self-efcacydo not have the same meaning.

An alternative explanation comes from cross-cultural psychol-ogy. As described above, values related to individualism andcollectivism are likely to affect the way the self is dened indifferent countries. These values may not only moderate the asso-ciation of teacher self-efcacy with their psychological well-being,but may also affect the average level of teachers evaluations oftheir own capabilities. Accounting for cross-national differences inteacher self-efcacy beliefs by referring to individualism-collectivism may be problematic. The individualism-collectivismdimension is very broad (Triandis, 1995) and individualism-collectivism strongly correlates with afuence (e.g., Hofstede,2001). As a consequence, correlations with individualism-collectivism may refer to the numerous differences between rich(individualistic) and poor (collectivistic) countries. Similarly,individualism-collectivism can be viewed as comprising variousinterrelated lower-level constructs, each of which could account for(part of) the cross-national differences observed.

We explore three possible mechanisms in the present study. Therst is internal locus of control, or the belief that outcomes dependon ones own behaviour rather than being random or under thecontrol of powerful others or external conditions (Rotter, 1966).People in collectivistic countries have been found to show aweakerinternal locus of control (Heine & Lehman, 1995), but a highercondence in collective control when compared to people inindividualistic countries (Yamaguchi, Gelfand, Ohashi, & Zemba,2005). As teachers locus of control and self-efcacy are closelyrelated constructs (Tschannen-Moran & Woolfolk Hoy, 2001), cul-tural inuences on perceived internal locus of control are likely toaffect a teachers perception of his or her own ability to teach.

Second, cross-national differences in self-beliefs are oftenexplained by referring to differences in self-enhancement needs.Self-enhancement refers to the motivation to seek activities andsocial situations conducive to maintaining a positive self-image,which causes a positive bias in self-evaluations. It has beenargued that this motive is weaker in collectivistic cultures wherethe subjective well-being depends more on belongingness to anesteemed in-group than on self-success (Heine & Hamamura,2007).

Third, differences in teacher self-efcacy in individualistic andcollectivistic countries also may reside in norms about its publicexpression (e.g., Kobayashi & Greenwald, 2003; Kurman, 2003).These may not only be related to the individualism-collectivismdimension. Rather, ethnographic research suggests that they maybe different in societies with Confucian-based versus Mediterra-nean honour collectivism (see also Uskul, Oyserman, & Schwarz,2010). Both types of cultures attach great importance to socialinterdependence, but they differ in strategies to maintain in-groupcohesion. Confucian-based collectivism encourages modesty,tting in, and not offending others to assure harmony within thegroup (e.g., Heine, 2007; Markus & Kitayama, 1991). This can giverise to a public understatement of ones own positive traits andabilities, but also to a general tendency to be moderate in theexpression of opinions (Chen, Lee, & Stevenson, 1995). In contrast,Mediterranean honour collectivism attaches great importance toprotecting group honour and reputation. Because honour does notdirectly result from behaviour but rather from its social evaluation,honour collectivism is likely to elicit positive self-presentations(e.g., Nisbett & Cohen, 1996; Peristiany, 1965). In individualisticcultures social interactions are rather contractual and negotiable.

S. Vieluf et al. / Teaching and Te94To ensure a fair basis for these negotiations, norms in individualisticsocieties encourage honesty and authenticity in self-presentationsand attitude statements (Triandis, 1995).

The meaning of cross-national differences in teacher self-efcacy has consequences for educational research and policy.The use of country means for teacher self-efcacy as indicators ofeducation systems is only valid if the meaning of these countrymeans is similar to the meaning of individual-level scores. In otherwords, a low average score for a country would be a reason forconcern, if teacher self-efcacy was linked with teachers psycho-logical well-being and student achievement not only at the indi-vidual level, but also at country level. However, if country-levelscores for teacher self-efcacy were affected by cross-nationaldifferences in self-presentational norms, they could not be inter-preted analogously to the individual level. This would render themeans for teacher self-efcacy incomparable. An inuence ofculturally shaped self-enhancement motives would additionallycall into question the relevance of examining individual-level dif-ferences in teacher self-efcacy in collectivistic societies. It can beconcluded that the meaning of country-level scores of teacher self-efcacy has implications for policy and that we cannot take thesedifferences at face value as such differences could be engendered byvarious unintended factors, such as self-presentational norms.

1.3. Aims of the present study

The main aim of the present study is to gain a better under-standing of cross-national similarities and differences related toteacher self-efcacy. We examine, rst, whether teacher self-efcacy has the same meaning cross-nationally, and second,whether there are cross-national differences in average levels ofteacher self-efcacy.

To answer the rst question, structural equation modelling wasused to investigate cross-national equivalence of teacher self-efcacy and its nomological network. We used data from 23 edu-cation systems, thereby providing a rather rigorous test of thecross-national generalizability of teacher self-efcacy. The rsthypotheses compare differences across countries at the individuallevel:

I.1. Teacher self-efcacy has a similar basic factor structureacross countries. This was examined by analyzing whether thesame items load on one teacher self-efcacy factor across countries,and by testing for equivalent factor loadings and intercepts in alatent modelling approach.

I.2. Teacher self-efcacy has the same nomological networkacross countries. This was tested by comparing the teacher-levelcorrelations between teacher self-efcacy and presumably relatedconstructs across countries. Drawing on prior research, we hy-pothesized that teacher self-efcacy would relate positively to jobsatisfaction and effective classroom teaching practices within allcountries.

The next hypothesis refers to the comparison of teacher self-efcacy at the individual and at the country level:

II.1. The nomological networks of teacher self-efcacy at indi-vidual and country-level are different. This multilevel equivalenceis tested by comparing the correlational pattern at the individuallevel with the pattern that results after variables are aggregated tothe country level.

Additional hypotheses refer to an alternative explanation forcountry-level differences.

II.2. Country-level differences in teacher self-efcacy are relatedto cultural value orientations. We expect teacher self-efcacy tocorrelate negatively with collectivism. Three mechanisms for thiseffect were discussed: we expect teacher self-efcacy to be posi-tively associated with an internal locus of control and self-

r Education 35 (2013) 92e103enhancement, and we expect positive respective negative

lesson I present a short summary of the previous lesson, andStudents evaluate and reect upon their own work. The StudentOrientation Scale consisted of four items: Students work in smallgroups to come upwith a joint solution to a problem or task, I givedifferent work to the students that have difculties learning and/orto thosewho can advance faster, I askmy students to suggest or tohelp plan classroom activities or topics, and Students work ingroups based upon their abilities. The Enhanced Activities Scalealso included four items: Students work on projects that require atleast one week to complete, Students make a product that will beused by someone else, I ask my students to write an essay inwhich they are expected to explain their thinking or reasoning atsome length, and Students hold a debate and argue for a partic-ular point of view which may not be their own. Response cate-gories were: never or hardly ever; in about one-quarter of lessons; inabout one-half of lessons; in about three-quarters of lessons; and inalmost every lesson. Higher scores indicated more frequent use ofeffective practices (Vieluf et al., 2010).

Student reading achievement scores as country-level variableswere derived from the Programme for International StudentAssessment (PISA) 2009 cycle (OECD, 2012). The PISA tests are

achecorrelations with two indicators of self-presentational norms:modesty values and extremity scoring.

2. Method

The present study was a secondary analysis of data from severalcross-national studies. Teacher self-efcacy and other teacher mea-sureswere taken from the OECDTeaching and Learning InternationalSurvey (TALIS; OECD, 2009, 2010b). TALIS examined 23 educationsystems where the ministries of education had agreed to fund andparticipate in the study: Australia, Austria, Belgium (Flemish), Brazil,Bulgaria, Denmark, Estonia, Hungary, Iceland, Ireland, Italy, Republicof Korea, Lithuania, Malaysia, Malta, Mexico, Norway, Poland,Portugal, Slovak Republic, Slovenia, Spain, and Turkey. Followingquestionnaire development, translation, a pilot study, and aeld trial,the questionnaires were administered to principals and teachers inthe second half of 2007 and the rst half of 2008. The mode ofadministration was paper-and-pencil and online. A rigorous qualitycontrol programmewas implementedduring all phases of the survey.

At the country level, the data were merged with indicators ofeducation systems from the OECD (2012) as well as with culturalvalue data (House, Hanges, Javidan, Dorfman, & Gupta, 2004).

2.1. Participants

The TALIS sample consisted of 73,100 teachers in4362 schools andwas representative of the population of teachers who provided in-struction in programmes at the lower secondary level in each of theparticipating countries. A two-stage stratied probability samplingdesignwas implemented: First, 200 (ormore) schoolswere randomlyselected from the population of schools within an education system,where ISCED Level 2 education (lower secondary education; for amore detailed explanation of ISCED levels, see UNESCO, 2006) wasprovided. Second, at least 20 teachers who teach regular classes andwho do not also act as principals were randomly selected withinthese schools. In the nal sample, two thirds of the teachers werefemale. As to age,17%were under 30 years old, 61%were between 30and 50, and 22% were more than 50 years old. When reporting ontheir experience, 31% of the teachers in the international sample re-ported having taught for more than 20 years, 29% reported between10 and 20 years, 37% had between 1 and 10 years of experience, and3% were in their rst year of teaching. In the breakdown by subject,30% of the teachers reported teaching reading, 22%mathematics, 23%science, 27%social studies,19% foreign languages,15% technology,19%arts, 15% physical education, 14% religion or ethics, 9% practicalvocational skills, and 5%, another subject.

For most analyses at the individual level the full sample of73,100 teachers was used. For Multiple Group Conrmatory FactorAnalysis (MGCFA) and regression analysis only, a subsample of1000 cases per country was drawn to assure that all countries had asimilar inuence on t indexes. The 23,000 teachers were selectedproportionally to their nal estimation weight, giving those caseswith relatively larger weights a higher probability of being selected.For analyses at the country level, country means for the fullweighted TALIS sample of 73,100 teachers were related to in-dicators from other large scale studies and public databases.

2.2. Measures

Table 1 shows summary statistics for all variables, includingreliabilities.

2.2.1. Teacher self-efcacyThe construct was measured with four items: I feel that I am

S. Vieluf et al. / Teaching and Temaking a signicant educational difference in the lives of mystudents, If I try really hard, I can make progress with even themost difcult and unmotivated students, I am successful with thestudents in my class, I usually know how to get through to stu-dents. These formed a single factor. Hence, while most previouscross-national studies examined multiple dimensions of teacherself-efcacy, TALIS used a relatively short measure of generalteacher self-efcacy. Response categories were: strongly agree;agree; disagree; and strongly disagree. Less than 1% of the teachersused the response category strongly disagree; therefore, disagreeand strongly disagree were combined (Vieluf, Leon, & Carstens,2010).

2.2.2. Educational processes and outcomesJob satisfaction was measured with a single item on the teacher

questionnaire. Teachers were asked to indicate their endorsementof the statement All in all, I am satised with my job on a four-point Likert scale, ranging from strongly disagree to strongly agree.

Classroom teaching practices were measured with three scalesin the teacher questionnaire. The Structuring Scale consisted of veitems: I explicitly state learning goals, I review with the studentsthe homework they have prepared, I ask my students toremember every step in a procedure, At the beginning of the

Table 1Summary statistics of variables.

Variable M SD Min Max Reliability

Teacher-level indicatorsT: Teacher self-efcacy 0.05 1.04 3.14 2.45 0.76T: Job satisfaction 3.11 0.62 1.00 4.00 ea

T: Teaching practices: structuring 0.11 1.03 3.52 2.45 0.73T: Teaching practices:

student orientation0.39 1.10 3.52 2.45 0.70

T: Teaching practices:enhanced activities

0.67 1.25 1.66 4.35 0.72

Country-level indicatorsP: Reading literacy achievement 482.52 30.48 412.00 539.00 0.86c

GL: In-group collectivism 5.17 0.61 3.53 5.88 0.77STD: Locus of control 9.45 1.37 7.36 12.09 eb

E: Modesty 0.00 0.35 0.41 0.73 eaT: Extremity scoring 0.28 0.07 0.17 0.45 0.97

a Single item or index.b No reliabilities reported in original study.c Reliability coefcient exists only for the student level. T TALIS; P PISA;

GL GLOBE; STD Smith, Trompenaars, & Dugan; E European Social Survey.

r Education 35 (2013) 92e103 95paper-and-pencil tests with multiple-choice, short-answer, and

shared variance. The remaining 36 items have an average absoluteintercorrelation of r 0.06, p > 0.05. Thus, the assumption ofheterogeneity of content was met.

2.3. Statistical modelling

In the present study, we compared the factor structure ofteacher self-efcacy, its correlations with other variables, and itsmean scores across countries. To this end, we rst employed a se-ries of structural equation models. These models (Multiple GroupConrmatory Factor Analysis, MGCFA) allowed us to test a) whether

acher Education 35 (2013) 92e103extended-response items. The items were developed by an inter-national consortium in cooperation with groups of experts in thesubject areas. In the PISA studies, scaling was done using the Raschmodel. PISA studies have been carried out every three years acrossall OECD countries with a number of economic partners. The pre-sent study used PISA 2009 population estimates for the country-level average reading achievement.

2.2.3. Cultural value orientationsIndividualism-collectivism as a country variable was oper-

ationalized using the in-group collectivism scale from the GlobalLeadership and Organizational Behaviour Effectiveness ResearchProgramme (GLOBE; House et al., 2004). GLOBE used a second scaleto measure institutional collectivism. We chose to use only theindex for in-group collectivism, because it better reects how theconstruct individualism/collectivism is typically understood in theliterature and it is more closely associated with other collectivismindicators (House et al., 2004). It is dened as the degree to whichindividuals express pride, loyalty, and cohesiveness in their orga-nizations or families (House et al., 2004, p. 30). The scale wasoperationalized with four items referring to descriptive norms(how things are done in the respondents country). An exampleitem is: In this society, children take pride in the individual ac-complishments of their parents. These questions were answeredby 17,370 middle managers from 951 organizations in 62 countrieson a seven-point Likert scale ranging from strongly agree to stronglydisagree, and they were reverse scored. Cultural value orientationshave been shown to be relatively stable over time (e.g., Hofstede,2001), legitimating their use as a stable culture-level constructand combining with scores from other studies and differentsamples.

Country-level scores for an internal/external locus of controlwere taken from Smith, Trompenaars, and Dugan (1995). They gavethe Rotter (1966) scale with 23 forced-choice items to 8841 man-agers and organization employees in 43 countries. One exemplaryitem is: Many of the unhappy things in peoples lives are partly dueto bad luck versus Peoples misfortunes result from the mistakesthey make.

Self-enhancement was assessed as the strength of the associa-tion between teacher self-efcacy and job satisfaction within eachof the countries participating in TALIS. This operationalization isbased on the argument that the self-enhancement motive is re-ected in the dependence of personal well-being on positive self-evaluations. The TALIS questions used to assess both constructsare described above.

Modesty was operationalized with one item from the thirdround of the European Social Survey (http://www.europeansocialsurvey.org/) carried out in 2006 that asked explicitly aboutthe personal importance of modesty. The wording of this itemwas:It is important to him/her to be humble and modest. He/she triesnot to draw attention to him/herself. A total of 40,278 participantsin 23 countries indicated howmuch a person to whom the value ofmodesty is important is like them on a six-point Likert scaleranging from very much like me to not like me at all. The sample wasrepresentative of all persons aged 15 and over who are residentwithin private households in each of the participating countries.(For a more detailed description see Schwartz, 2003.)

An index for extremity scoring was computed based on theTALIS questionnaire data. It was calculated as the proportion ofresponses in the two extreme response categories (strongly disagreeand strongly agree) divided by the total number of items. In linewith the literature, a heterogeneous set of 32 items was chosen(e.g., Baumgartner & Steenkamp, 2001). All items from TALIS withthe response options disagree and agreewere used; only the teacher

S. Vieluf et al. / Teaching and Te96self-efcacy items were excluded to avoid an overestimation ofthe measurement qualities of our teacher self-efcacy measurewere similar across countries and b) whether similar correlationsoccurred between teacher self-efcacy and other constructs in all ofthe countries. (Appendix B includes a more detailed discussion ofthese analyses.) For the cross-national comparisons of mean scoreswe used the latent means derived from these MGCFA models. Next,intraclass correlations, that is, the variance distribution betweenthe country-, school-, and teacher-levels were analyzed with theprogramme HLM (Raudenbush & Bryk, 2002; Snijders & Bosker,1999). Finally, we computed country-level correlations betweenthe latent country means and other country-level indicators.

3. Results

In the presentation of results, we rst describe analyses of cross-national similarities and differences in the factorial structure andthe nomological network of teacher self-efcacy; and second, weaddress country means and their country-level associations witheducational processes and outcomes and cultural valueorientations.

3.1. Cross-national comparison of the structure of teacher self-efcacy

Hypothesis I.1 supposed equivalence of the basic factor structureof teacher self-efcacy across countries. We used three nestedMGCFA models to test this hypothesis. The results are specied inTable 2 (see also Vieluf et al., 2010). In the MGCFA models onegradually tests different levels of equivalence; a comparison of thet indices of the models allows us then to determine which level ofequivalence can be assumed.

The rst line of Table 2 shows t indices for a model that testedthe most basic level of equivalence, the so-called conguralequivalence. This model was supported, indicating the same fouritems form a single teacher self-efcacy factor within each of the 23countries participating in TALIS e a precondition for all furthercross-national comparisons of teacher self-efcacy (van de Vijver &Poortinga, 1982). We then tested the second level of equivalence,the so-called metric equivalence, indicating that all factor loadingsof the teacher self-efcacy items on the latent factor are similar

Table 2Model t for a multiple group conrmatory factor analysis with different parameterrestrictions.

Model Model t

Freeparameters

LL Dc2 BIC CFI RMSEA

Congural equivalence 276 55039 e 112844 0.98 0.09Metric equivalence 210 55182 222** 112468 0.97 0.07Scalar equivalence 122 57066 3318** 115354 0.75 0.14

Note. LL Log-likelihood; BIC sample-size adjusted Bayesian information crite-

rion; CFI Comparative Fit Index; RMSEA Root Mean Square Error of Approxi-mation. **p < 0.01.

across countries. This level is needed for making valid cross-national comparisons of intra-cultural score differences and ofcorrelations with third constructs (van de Vijver & Poortinga, 1982).The t of this model (shown in the second line of Table 2) was onlyslightly poorer as compared to the congural equivalence model.Hence, metric equivalence could also be established; our teacherself-efcacy scale had a largely similar metric across countries.Cross-national comparisons of mean scores additionally require so-called scalar equivalence (van de Vijver & Poortinga, 1982). Thislevel was not supported by our results. The t indices shown in thethird line of Table 2 were noticeably poorer than those of the metricequivalence model. Hence, the TALIS teacher self-efcacy scale didnot have the same origin cross-nationally.

In addition to the factorial structure, we also compared corre-lations with other variables across countries. Hypothesis I.2postulated that teacher self-efcacy had the same nomologicalnetwork within each of the countries. To test this hypothesis, weexamined latent correlations of teacher self-efcacy with jobsatisfaction and the three classroom teaching practices (i.e., struc-turing, student orientation, and enhanced activities). The resultsshowed that teacher self-efcacy is associated positively with allfour variables. A strong correlation was observed with job satis-

S. Vieluf et al. / Teaching and Teachefaction, r(73,098) 0.53, p < 0.01. Correlations with the three di-mensions of classroom teaching practices were weaker, yet stillsignicant, r(73,098) 0.20, p < 0.05 for structuring, r(73,098) 0.24, p< 0.05 for student orientation, and r(73,098) 0.23, p< 0.05for enhanced activities. To examine the similarity of this patternacross countries, for each of the indicators we compared the t of amodel in which the correlation coefcients were allowed to varyacross countries with a model in which they were identical. Table 3shows that each of the four restricted models tted equally well toits corresponding free model. From this we can conclude that allfour correlations were cross-nationally equivalent. Across coun-tries, teachers with higher self-efcacy report to be more satisedwith their job and to use all three types of classroom teachingpractices more often than their colleagues.

In sum, our ndings show that teacher self-efcacy has the samepsychological meaning in all countries involved, because it has asimilar factor structure and it shows the same pattern of positivecorrelations with job satisfaction and classroom teaching practices.The scale further has a basic level of cross-national comparability in

Table 3Model t for a multiple group model relating teacher self-efcacy to job satisfactionand classroom teaching practices with and without restrictions on correlationcoefcients.

Model Model t

Free parameters Log-likelihood Dc2 BIC CFI RMSEA

Teacher self-efcacy with job satisfactionr free 279 242942 e 489009 0.93 0.08r restricted 257 243119 207** 489116 0.93 0.08

Teacher self-efcacy with classroom teaching practices e structuringr free 279 279276 e 561679 0.96 0.06r restricted 257 279344 122** 561568 0.96 0.06

Teacher self-efcacy with classroom teaching practices e student orientationr free 279 274107 e 551340 0.96 0.06r restricted 257 274186 129** 551251 0.96 0.06

Teacher self-efcacy with classroom teaching practices e enhanced activitiesr free 279 263683 e 530491 0.96 0.06r restricted 257 263768 132** 530415 0.96 0.06

Note. LL Log-likelihood; BIC sample-size adjusted Bayesian information crite-

rion; CFI Comparative Fit Index; RMSEA Root Mean Square Error of Approxi-mation. **p < 0.01.a qualitative and quantitative sense. More specically, it can bevalidly used for comparisons of intra-cultural score differences andcorrelations with other variables. Note however, that comparisonsof mean scores might be biased by the lack of scalar equivalence.

3.2. Cross-national comparison of the level of teacher self-efcacy

The second aim of this study was to gain a deeper under-standing of whether and why the level of teacher self-efcacyvaries across cultures. To do this, rst we compared the meanscores for teacher self-efcacy across countries and second, wecorrelated them with educational processes and outcomes andcultural value orientations.

Our previous analyses suggested that teacher self-efcacy scalesshowed metric but not scalar equivalence. Thus, teacher-level cor-relations can be compared validly across countries, but mean scorecomparisonsmay be biased. However, the country-level correlationbetween latent means assuming scalar versus metric equivalencewas high, r(21) 0.91, p < 0.01. Thus, we concluded that the lack ofscalar equivalence is fairly inconsequential for analyzing country-level correlations. Nevertheless, comparisons between singlecountries should be interpreted cautiously.

Fig. 1 shows the latent means for scores based on a factor modelwith varying intercepts. Teachers in Norway had by far the highestaverage self-efcacy scores. Relatively high scores were also foundin the other Northern European countries (Iceland and Denmark),the English-speaking countries (Ireland and Australia) and inAustria. The Mediterranean and South American countries hadrather low means (especially Spain and Brazil, but also Portugal,Malta, Mexico, and Turkey, the only exception was Italy). Compar-atively low scores were also found in the Eastern European coun-tries (especially in Estonia, Hungary, the Slovak Republic, andPoland, while they were a bit higher in Lithuania, Slovenia, andBulgaria). Finally, by far the lowest average teacher self-efcacyscore was observed in the Republic of Korea. As commonly foundin cross-national studies (Poortinga, Van de Vijver, & Van Hemert,2002), the variation was noticeably smaller between countriesthan within countries. At 87%, the share of variance at the indi-vidual level was the largest, 4% of the variance was found at theschool level, and 8% was found at the country level.

We further examined country-level correlations of teacher self-efcacy to test hypotheses II.1 and II.2 (the results are presented inTable 4). Hypothesis II.1 stated that we expected to nd a differentnomological network of teacher self-efcacy at the country level ascompared to the individual teacher level. To test this hypothesis, wecomputed country-level correlations of teacher self-efcacy witheducational processes and outcomes (i.e., the three dimensions ofclassroom teaching practices, job satisfaction as well as studentreading literacy achievement). It should be noted that country-levelindicators were not available for all countries; so, the actual num-ber of countries used in computing these correlations differs acrossvariables (actual numbers are presented in Table 4). Positive andsignicant correlations with all seven indicators would point tosimilarity inmeaning at the individual and country levels; however,as we had expected, the correlations did not provide strong supportfor this similarity. Although correlations of job satisfaction andstructuring classroom teaching practices with teacher self-efcacyat the country level showed positive signs, they were small, andonly the correlation with job satisfaction reached signicance. Nocountry-level correlations were observed with student orientationor enhanced activities. Most notably, correlations with studentachievement were not signicant. The small sample sizes at thecountry level could have an impact on the signicance of the cor-relations; however, the global patterning of the correlations does

r Education 35 (2013) 92e103 97not suggest a close correspondence of correlations at the individual

terva

S. Vieluf et al. / Teaching and Teache98and country levels and is more in line with the interpretation thatteacher self-efcacy has a different meaning at the teacher levelthan it does at the country level.

An alternative interpretation of country-level variation inteacher self-efcacywas formulated in hypothesis II.2, which statedthat teacher self-efcacy would be negatively linked to collec-tivism. This hypothesis was conrmed: In-group collectivism wasstrongly and positively associated with teacher self-efcacy at thecountry level (see Table 4).1 Three alternative interpretations wereproposed: cross-national differences in internal locus of control,the need for self-enhancement, or self-presentational norms. Eachof these could explain (alone or in combination) the associationbetween collectivism and teacher self-efcacy. Our results

Fig. 1. Latent means and condence inconrmed the latter more than the two former mechanisms.Modesty and extremity scoring were both signicantly correlatedwith teacher self-efcacy at the country-level (see Table 4). Themore modesty and humbleness were valued in a country and theweaker the tendency to choose the extreme points of Likertresponse scales was, the lower the average teacher self-efcacyscore. In contrast, no signicant country-level associations wereobserved with locus of control or self-enhancement.

To gain a better understanding of these complex results, partialcorrelations controlling for the four scales that showed signicantcountry-level correlations with teacher self-efcacy (job satisfac-tion, in-group collectivism, modesty, and extremity scoring) arereported in Table 4. The pattern of correlations remained largelysimilar after excluding the effects of these third variables: Thecountry-level correlations of teacher self-efcacy with job satis-faction and extremity scoring remained positive and strong andthose with in-group collectivism and modesty remained negativeand strong, even after correction for the respective other variables.Likewise, the partial country-level correlations of teacher self-efcacy with all other variables remained non-signicant, nomatter whether any of the control variables was included in themodel. However, there were some exceptions: The partial

1 We also analysed correlations with individualism-collectivism indicators fromHofstede (2001) and Schwartz (1992, 1994). The results were largely similar, so theywere not discussed due to space limitations.correlation of teacher self-efcacy with in-group collectivism wasnon-signicant when modesty was controlled for e in contrast tothe negative zero-order correlation between both constructs. Alsothe partial correlation with extremity scoring declined and becamenon-signicant with modesty as covariate. Finally, the inclusion ofmodesty also revealed a strongly positive partial country-levelcorrelation of teacher self-efcacy with structuring classroomteaching practices. These results may suggest mediation, con-founding or suppression effects. However, all three partial corre-lations were based on a country sample that was considerablysmaller than that for the zero-order correlation. Thus, the observeddifferences between corrected and uncorrected correlations mayalso be due to sample uctuations.

ls for teacher-self-efcacy by country.

r Education 35 (2013) 92e103Altogether these results show that, at the country level, teacherself-efcacy was closely associated with job satisfaction, in-groupcollectivism, modesty, and extremity scoring, whereas no correla-tions were observed with any of the other variables. In-groupcollectivism, modesty, and extremity scoring further alsoappeared to be closely linked with each other.

To understand whether the distinction between Confucian-based and Mediterranean honour collectivism is important forunderstanding the association between collectivism, modesty, andteacher self-efcacy, we additionally carried out some moredescriptive analyses. We compared mean scores for teacher self-efcacy, modesty, and in-group collectivism across the groups ofConfucian-based collectivistic (South Korea), Mediterranean hon-our collectivistic (Brazil, Italy, Malta, Mexico, Portugal, Slovenia,Spain, and Turkey) and individualistic countries (Australia, Austria,Belgium (Flemish), Denmark, Iceland, Ireland, and Norway). Weadditionally distinguished a fourth group of Eastern Europeancountries (Bulgaria, Estonia, Hungary, Lithuania, Poland, and theSlovak Republic). These countries also have high collectivismvalues, but they are neither characterized by Confucian traditionnor do they belong to the group of Mediterranean countries. Table 5shows that all groups of collectivistic countries had higher in-groupcollectivism scores than the individualistic countries, which vali-dates their assignment to the groups. With regard to teacher self-efcacy and modesty our hypotheses were partly conrmed: Asexpected, the lowest modesty and the highest teacher self-efcacymeans were observed in individualistic countries, whereas the

Table 4Zero-order and partial country-level correlations with teacher self-efcacy controlling fo

Scale Correlation Partial corr

Job satisfac

n r df r

T: Job satisfaction 23 0.40* e eT: Teaching practices: structuring 23 0.19 20 0.T: Teaching practices: student orientation 23 0.09 20 T: Teaching practices: enhanced activities 23 0.00 20 P: Reading achievement 21 0.17 18

0.L

0.

, Tro

S. Vieluf et al. / Teaching and Teacheaverage teacher self-efcacy score was especially low in theConfucian-based collectivistic country (South Korea). However, theMediterranean honour collectivistic and the Eastern Europeancollectivistic countries also had high values for modesty and lowteacher self-efcacy means.

4. Discussion

The overarching goal of the present study was to enhance ourunderstanding of cross-national similarities and differences inteacher self-efcacy. We rst examined whether the structure ofthe construct was similar across countries e a basic preconditionfor all other cross-national comparisons. Second, we compared thestrength of associations of teacher self-efcacy with job satisfactionand classroom teaching practices across countries to test whetherthe psychological and practical relevance of teacher self-efcacywas linked with national cultures. Third, we analyzed cross-national variation in the average teacher self-efcacy. This latteranalysis provided information about the interpretability of countrymeans as indicators of education systems. The structure-orientedanalyses of factor structures and teacher-level correlations arediscussed separately from the level-oriented analyses of countrymeans.

4.1. Cross-national generalizability and comparability of teacherself-efcacy

GL: In-group collectivism 15 L0.46* 12STD: Locus of control 17 0.07 14T: Self-enhancement 23 0.21 20E: Modesty 13 L0.61** 10T: Extremity scoring 23 0.59** 20

Note. Signicant correlations in bold. T TALIS; P PISA; GL GLOBE; STD Smith(one-tailed).Structure-oriented analyses were conducted to determine thecross-national generalizability of the factor structure and nomo-logical network of teacher self-efcacy. We contrasted two hy-potheses: Banduras (1997) view which holds that self-efcacy is auniversal construct with a similar function across cultures andSchoolers (1990) assumption that positive feelings about ones

Table 5Mean scores for teacher self-efcacy, modesty, and in-group collectivism in fourgroups of countries.

GL: In-groupcollectivism

E: Modesty T: Teacherself-efcacy

Confucian-based collectivism 5.54 ea 0.23Mediterranean honor collectivism 5.44 0.37 0.01Eastern European collectivism 5.39 0.09 0.06Individualism 4.42 0.28 0.08a No modesty score was available for South Korea, the only country in the TALIS

sample representing a Confucian-based collectivism. GL GLOBE; E EuropeanSocial Survey; T TALIS.own abilities are more important to well-being and behaviour incountries with cultural values that foster individuality and self-reliance. The latter hypothesis would imply that individualismand collectivism (or closely related constructs such as countryafuence) moderate the association of teacher self-efcacy witheducational processes and with outcomes. The ndings of thepresent study support Banduras view of the universality of self-efcacy. We found that the same construct can be used to ac-count for responses to teacher self-efcacy items in different cul-tural groups and that the construct is manifested by items in thesame way across all countries participating in TALIS. Further, wecould replicate results of previous monocultural studies that foundteachers with higher self-efcacy to be more satised with theirjobs (e.g., Klassen & Chiu, 2010) and to report more effectiveteaching practices (e.g., Gibson & Dembo, 1984) and we could showthat these relationships were similar in a variety of cultural set-tings. Thus, it seems that regardless of their cultural backgrounds,teachers tend to develop a coherent sense of their professionalcapability and that this sense is linked to their professionalbehaviour and well-being.

From these ndings it can be concluded that examining teacherself-efcacy is relevant not only in the individualistic countrieswhere Locus of Control (Rotter, 1966) and Social Cognitive(Bandura, 1977, 1986) theories have been developed, but also incountries characterized by more collectivistic value orientations,because teacher self-efcacy seems to full a similar psychological

r teacher job satisfaction, collectivism, modesty, and extremity scoring.

elation controlling for:

tion In-groupcollectivism

Modesty ERS

df r df r df r

12 0.43 10 0.49 20 0.48*04 12 0.21 10 0.77** 20 0.400.23 12 0.08 10 0.08 20 0.050.22 12 0.07 10 0.36 20 0.090.07 11 0.33 10 0.26 18 0.090.49 - - 5 0.13 12 0.5006 11 0.04 8 0.19 14 0.130.08 12 0.06 10 0.51 20 0.200.67** 5 0.48 - - 10 0.4264** 12 0.62** 10 0.38 - -

mpenaars, & Dugan; E European Social Survey. *p < 0.05 (one-tailed). **p < 0.01

r Education 35 (2013) 92e103 99function independent of the national context of an education sys-tem. Moreover the ndings attest a basic level of cross-nationalcomparability of the teacher self-efcacy scale that has been usedin TALIS: The strength of its teacher-level associations with otherconstructs can be validly compared across different countries.

4.2. Cross-national variation in teacher self-efcacy and itsmeaning

In addition to the structure of the construct, we examined cross-national differences in the mean level of teacher self-efcacy. Thendings suggest that the mean was high in the Northern Europeanparticipating countries, rather low in Eastern European countries,and particularly low in the Republic of Korea, the only SoutheastAsian country that participated in TALIS. Southern and CentralEuropean countries as well as South American countries were inbetween. Thus, we could replicate previous ndings on cross-national differences between Southeast Asian and English-speaking countries in teacher self-efcacy (e.g., Ho & Hau, 2004;Klassen et al., 2009) and expand this previous research by

acheincluding a larger sample of diverse countries. Next, we tried to ndout how to interpret these cross-national differences.

At the individual level, teacher self-efcacy is the belief in theown capability to perform specic teaching tasks at a specied levelof quality in a specied situation (Dellinger et al., 2008, p. 4). It isassociated with teachers job satisfaction, classroom teachingpractices and students outcomes (e.g., Anderson et al., 1988). Ourresults conrmed these previous ndings at the individual level,but we found a more complicated pattern of correlations at thecountry level. We found average job satisfaction to correlate posi-tively with teacher self-efcacy as has been found at the individuallevel, although all other country level correlations with indicatorsof educational processes and outcomes were insignicant. Thiscould be a consequence of the small sample size at country-level;however, the country-level correlations of teacher self-efcacywith student orientation and enhanced activities were close tozero, and correlations with student achievement were absent.Therefore, it is quite likely that the meaning of teacher self-efcacyat individual level is different from the meaning of teacher self-efcacy at country level (called non-isomorphism in the multi-level literature; Van de Vijver, van Hemert, & Poortinga, 2008).

This conclusion receives additional support from the nding ofcountry-level correlations between teacher self-efcacy and cul-tural value orientations. Theory and empirical studies in the eld ofcross-cultural psychology (e.g., Heine & Hamamura, 2007; Kurman,2003; Uskul et al., 2010) had suggested a negative association be-tween teacher self-efcacy and collectivism. This hypothesis wasconrmed. Thus, we can conclude that teachers report a higherself-efcacy in countries where interdependence is emphasizedmore.

Three mechanisms for this effect were derived from theory. Weexpected teacher self-efcacy to be associated with internal locusof control, self-enhancement, and self-presentational norms. Thepattern of correlations supported the latter, but not the two formermechanisms. Strong correlations were observed with modesty andextremity scoring, whereas neither an internal locus of control norself-enhancement was associated with teacher self-efcacy at thecountry level. The reservation must be made, however, that theTALIS data may not have provided us with the best operationali-zation of self-enhancement: Often the association between self-evaluations and well-being is computed controlling for achieve-ment (Kurman, 2003), but TALIS did not provide any measure ofteacher competence. Moreover, the limitations that arose from thesmall sample size at the country level have to be taken into account.Under these qualiers, our ndings suggest that the countryaverage for teacher self-efcacy is linked particularly to cross-national differences in self-presentational norms.

Moreover, results of partial correlations can be interpreted asindicating that modesty mediates the effect of in-group collec-tivism; a larger set of countries would be needed to conduct a full-edged mediation analysis. In any case, collectivism, modesty, andextremity scoring were not only linked with teacher self-efcacy,but also closely related to each other. Expressing opinions andespecially beliefs about the self in a modest waymay help maintainclose interpersonal afliations by leading to greater in-group har-mony, which is of strong concern in collectivistic cultures (e.g.,Mau, 2000). In countries with this pattern of norms and values,teachers tend to make less favourable judgements of their capa-bility when they respond to self-efcacy items.

This may hold true not only for teachers and self-efcacy, butalso for other population groups and other types of self-evaluations.Kurman (2003), for example, foundmodesty tomediate the effect ofthe country on the academic self-evaluations of high-school stu-dents in Singapore and Israel. She also found a correlation between

S. Vieluf et al. / Teaching and Te100self-evaluations and individualism-collectivism but, similar to ourresults, this was weaker than the associationwith modesty. Van derGaer, Grisay, Schulz, and Gebhardt (2012) examined a larger countrysample and showed that country differences in students scienceself-concept can be explained by modesty bias. These authors didnot examine cultural value orientations and they operationalizedmodesty bias as Confucian heritage. Results of the present studysuggest that moderation in self-presentation not only is valued incultures with a Confucian tradition, but also can help explain dif-ferences in self-evaluations between more and less collectivisticEuropean countries. It is likely that we would have found evenstronger effects of modesty and collectivism if more Confucian so-cieties had been included in our sample, but evenwith the Republicof Korea as the only Confucian country, modesty values have beenshown to be highly relevant for explaining cross-national differ-ences in teacher self-efcacy.

In summary, our study suggests that teacher self-efcacy is notthe same at the individual and country level. Cultural norms andvalues that inuence self-presentation, such as norms regardingthe expression of modesty and extremity that are particularlyimportant in Confucian societies but also in other countries wherecollectivistic values are highly endorsed, have a bearing on cross-national differences in teacher self-efcacy. The inuence of thesefactors will be small in comparisons of teacher scores within asingle country but more prominent in cross-national comparisons.Our ndings should not be interpreted as a plea to enlargethe concept of teacher self-efcacy so as to include these culturalvalues; rather, our ndings suggest that in cross-national assess-ments of teacher self-efcacy we have to consider that cross-national differences are inuenced by differential normsregarding self-presentation. Therefore, variation between countriesin the responses to the teacher self-efcacy scale used in TALIScannot be interpreted analogously to variation between teachers. Alow country mean is more indicative of cultural modesty normsthan of low teacher self-efcacy. This speaks against the use ofaggregated teacher self-efcacy scores as indicators of educationsystems for policy purposes.

4.3. Limitations and future research

The present study used data from the OECD large-scale TALISsurvey on teacher self-efcacy and related it to country indicatorsfrom other studies and public databases. Experts from variouscultures have been involved in the development of the TALISquestionnaire, and the instrument has been tested thoroughly in apilot study and a eld trial. The teacher self-efcacy scale had aninternal consistency that was comparable to established measures(Henson, Kogan, & Vacha-Haase, 2001). Associations with teacherbeliefs about the nature of teaching and learning, classroomteaching practices, and several school policies and practices sup-ported its validity (OECD, 2009). However, due to time constraints,TALIS used only four items to operationalize teacher self-efcacy.Previous research had suggested that teacher self-efcacy is amultidimensional construct, and that measures should be situa-tion specic (e.g., Tschannen-Moran & Woolfolk Hoy, 2001;Woolfolk Hoy, 2004). However, the TALIS scale is a rather gen-eral measure of teacher-self-efcacy. For future studies it wouldbe preferable to use a longer scale, for instance, with questionsconcerning self-efcacy for instructional strategies, classroommanagement, and student engagement. This would allow morein-depth analysis of whether all different dimensions and theirintercorrelations are equivalent across countries. Moreover,asking more specic questions about the capability to performteaching tasks in certain situations could reduce the effect ofstandards and reference groups on country-level self-evaluations

r Education 35 (2013) 92e103(Van de Gaer et al., 2012).

acheMany previous studies have compared only two or threecountries due to the high cost of collecting data in multiple coun-tries. In the present study we analyze data from a comparativelylarge sample of 23 countries. However, an even larger samplewould be desirable in future studies to allow for an examination ofthe factor structure at the country level (e.g., Hox &Maas, 2001) andto facilitate testing mediation-hypotheses, also at the country level.In addition, these samples should include several countries char-acterized by Confucian-based as well as Mediterranean honourcollectivism to systematically compare these two variants ofcollectivism.

Finally, the cross-sectional design of the present study providesnew insights into the cross-national generalizability of associatingteacher self-efcacy with other educational processes and out-comes. Future longitudinal and experimental studies could helpdisentangle the causes and effects of these associations both at theteacher and at the country level.

4.4. Conclusion

In spite of these limitations, our results bear on several impor-tant issues. Comparisons of the structure and nomological networkat individual level suggest that teacher self-efcacy is not just aphenomenon of individualistic cultures; rather, a largely equivalentconstruct can be found across 23 quite diverse countries eincluding one Confucian-based collectivistic country, severalMediterranean honour collectivistic countries, and several EasternEuropean collectivistic countries. Moreover, teacher self-efcacyrelates positively to teachers job satisfaction and classroomteaching practices, independent of the cultural background of theteachers. Thus, it seems to full a similar psychological functioncross-nationally. These ndings corroborate the validity andgeneralizability of theory on teacher self-efcacy and support therelevance of collecting data and addressing teacher self-efcacy inresearch cross-nationally. This result is especially important in lightof the growing internationalization of research and educationalpolicy. Recently, teacher self-efcacy measures have been includedin large scale surveys such as TALIS (OECD, 2009) and TEDS-M(Tatto et al., 2008). This transport of theory to other cultural set-tings can be problematic, if its universality is merely assumed, butnot tested. Yet, our ndings support that it is meaningful toexamine teacher self-efcacy within the diverse countries thathave participated in TALIS.

Our ndings also inform about the degree of cross-nationalcomparability of the TALIS scale. The high degree of cross-national equivalence of the teacher self-efcacy scale indicatesthat for the purpose of comparing teacher-level correlations withthird variables across countries, it can produce valid results. At thesame time, cross-national differences in teacher self-efcacy scoresshould be interpreted with caution in that these comparisons ofmean scores may be biased and do not reect differences in teacherself-efcacy. While for individual teachers low self-efcacy beliefscan indicate a need for support, countries with low teacher self-efcacy scores are often those where students are high achievers.Rather, the variation at country level reects differences in culturalvalue orientations and self-presentational values, and this ndingmay not be limited to teachers or self-efcacy measures (e.g., vande Gaer et al., 2012; Kurman, 2003). Non-isomorphism does notjeopardize the validity of self-evaluation scales at the individuallevel, but it has important consequences for the design and inter-pretation of future cross-national studies. To measure the self-evaluation of a teacher population that predicts country-levelvariation in achievement, it may be necessary to take normsregarding self-presentations into account (e.g., by developing

S. Vieluf et al. / Teaching and Temeasures of these norms). Alternatively, it would be possible todevelop new instruments aimed specically at operationalizing theconstruct at this level of analysis. Inuences of response stylescould be reduced by using other item formats such as forced choiceand anchoring vignettes, or by drawing on indirect measurementtechniques such as the Implicit Association Test (IAT; Greenwald,McGhee, & Schwartz, 1998) or word-fragment completion tasks(see, e.g., Hetts, Sakuma, & Pelham,1999).Whatever option is viableand preferred, our study clearly points to the need to go beyondmerely administering self-reporting questionnaires on teacher self-efcacy in cross-national studies to advance our understanding.

Finally, our results also have consequences for teacher educa-tion. Supporting teachers with developing a strong sense of efcacyis sometimes considered an important goal of teacher education,professional development, and school improvement, especially inthe USA (e.g., Ashton, 1984; Fritz, Miller-Heyl, Kreutzer, & MacPhee,1995; Ross, 1995). With increasing international exchange and arenewed interest in policy borrowing, it has also become an issue inother countries. Such transfer could have counter-productive ef-fects if the psychological function of teacher self-efcacy dependedon the cultural contexts. Yet, our ndings support the relevance ofaddressing teacher self-efcacy in teacher education programmesat least in the 23 countries that participated in TALIS.

Acknowledgements

We would like to thank Michael Davidson and Ben Jensen(project leaders, OECD), Ralph Carstens and Steffen Knoll (projectmanagers at IEA-DPC, the international study contractor), as well asDavid P. Baker, Aletta Grisay and Jaap Scheerens (members of theTALIS Questionnaire Expert Group) for excellent collaborationduring instrument development and data preparation.

Appendix. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.tate.2013.05.006.

References

Anderson, R. N., Greene, M. L., & Loewen, P. S. (1988). Relationships among teachersand students thinking skills, sense of efcacy, and student achievement. AlbertaJournal of Educational Research, 34, 148e165.

Armor, D. J., Conry-Oseguera, P., Cox, M., King, N., McDonnell, L., Pascal, A., et al.(1976). Analysis of the school preferred reading program in selected Los Angelesminority schools. Santa Monica, CA: RAND Corporation. Retrieved from http://www.rand.org/pubs/reports/2005/R2007.pdf.

Ashton, P. (1984). Teacher efcacy: a motivational paradigm for effective teachereducation. Journal of Teacher Education, 35(5), 28e32. http://dx.doi.org/10.1177/002248718403500507.

Ashton, P. T., & Webb, R. B. (1986). Making a difference: Teachers sense of efcacy andstudent achievement. New York, NY: Longman. Bandura, A. (1977). Social learningtheory. New York, NY: General Learning Press.

Bandura, A. (1977). Self-efcacy: toward a unifying theory of behavioral change.Psychological Review, 84, 191e215. http://dx.doi.org/10.1037/0033-295X.84.2.191.

Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory.Englewood Cliffs, NJ: Prentice-Hall.

Bandura, A. (1997). Self-efcacy: The exercise of control. New York, NY: Freeman.Bandura, A. (2002). Social cognitive theory in cultural context. Applied Psychology:

An International Review, 51, 269e290. http://dx.doi.org/10.1111/1464-0597.00092.

Baumgartner, H., & Steenkamp, J.-B. E. M. (2001). Response styles in marketingresearch: a cross-national investigation. Journal of Marketing Research, 38, 143e156. http://dx.doi.org/10.1509/jmkr.38.2.143.18840.

Blmeke, S., Kaiser, G., & Lehmann, R. (2010). TEDS-M 2008: Professionelle Kompetenzund Lerngelegenheiten angehender Primarstufenlehrkrfte im internationalenVergleich [TEDS-M 2008: Professional competence and learning opportunities ofprimary school teachers in international perspective]. Mnster, Germany:Waxmann.

Brouwers, A., & Tomic, W. (2000). A longitudinal study of teacher burnout and

r Education 35 (2013) 92e103 101perceived self-efcacy in classroom management. Teaching and Teacher Edu-cation, 16, 239e253. http://dx.doi.org/10.1016/S0742-051X(99)00057-8.

acheCaprara, G. V., Barbaranelli, C., Borgogni, L., & Steca, P. (2003). Efcacy beliefs asdeterminants of teachers job satisfaction. Journal of Educational Psychology, 95,821e832. http://dx.doi.org/10.1037/0022-0663.95.4.821.

Caprara, G. V., Barbaranelli, C., Steca, P., & Malone, P. S. (2006). Teachers self-efcacybeliefs as determinants of job satisfaction and students academic achievement:a study at the school level. Journal of School Psychology, 44, 473e490. http://dx.doi.org/10.1016/j.jsp.2006.09.001.

Chen, C., Lee, S., & Stevenson, H. W. (1995). Response style and cross-culturalcomparisons of rating scales among East Asian and North American students.Psychological Science, 6, 170e175. http://dx.doi.org/10.1111/j.1467-9280.1995.tb00327.x.

Dellinger, A. B., Bobbett, J. J., Olivier, D. F., & Ellett, C. D. (2008). Measuring teachersself-efcacy beliefs: development and use of the TEBS-Self. Teaching andTeacher Education, 24, 751e766. http://dx.doi.org/10.1016/j.tate.2007.02.010.

Fritz, J. J., Miller-Heyl, J., Kreutzer, J. C., & MacPhee, D. (1995). Fostering personalteaching efcacy through staff development and classroom activities. Journal ofEducational Research, 88, 200e208. http://dx.doi.org/10.1080/00220671.1995.9941301.

van de Gaer, E., Grisay, A., Schulz, W., & Gebhardt, E. (2012). The reference groupeffect: an explanation of the paradoxical relationship between academicachievement and self-condence across countries. Journal of Cross-CulturalPsychology. Advance Online Publication, http://dx.doi.org/10.1177/0022022111428083.

Gibson, S., & Dembo, M. H. (1984). Teacher efcacy: a construct validation. Journal ofEducational Psychology, 76, 669e682. http://dx.doi.org/10.1037/0022-0663.76.4.569.

Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individualdifferences in implicit cognition: the implicit association test. Journal of Per-sonality and Social Psychology, 74, 1464e1480. http://dx.doi.org/10.1037/0022-3514.74.6.1464.

Heine, S. J. (2007). Culture and motivation. What motivates people to act in theways that they do? In S. Kitayama, & D. Cohen (Eds.), Handbook of culturalpsychology (pp. 714e733) New York, NY: Guilford Press.

Heine, S. J., & Hamamura, T. (2007). In search of East Asian self-enhancement.Personality and Social Psychology Review, 11, 1e24. http://dx.doi.org/10.1177/1088868306294587.

Heine, S. J., & Lehman, D. R. (1995). Cultural variation in unrealistic optimism:does the West feel more invulnerable than the East? Journal of Personalityand Social Psychology, 68, 595e607. http://dx.doi.org/10.1037/0022-3514.68.4.595.

Henson, R. K., Kogan, L. R., & Vacha-Haase, T. (2001). A reliability generalizationstudy of the teacher efcacy scale and related instruments. Educational andPsychological Measurement, 61, 404e420. http://dx.doi.org/10.1177/00131640121971284.

Hetts, J. J., Sakuma, M., & Pelham, B. W. (1999). Two roads to positive regard: im-plicit and explicit self-evaluation and culture. Journal of Experimental SocialPsychology, 35, 512e559. http://dx.doi.org/10.1006/jesp.1999.1391.

Ho, I. T., & Hau, K.-T. (2004). Australian and Chinese teacher efcacy: similaritiesand differences in personal instruction, discipline, guidance efcacy and beliefsin external determinants. Teaching and Teacher Education, 20, 313e323. http://dx.doi.org/10.1016/j.tate.2003.09.009.

Hofstede, G. (2001). Cultures consequences. Comparing values, behaviors, institutions,and organizations across nations. Thousand Oaks, CA: Sage.

House, R. J., Hanges, P. J., Javidan, M., Dorfman, P. W., & Gupta, V. (Eds.). (2004).Leadership, culture, and organizations: The GLOBE study of 62 societies. ThousandOaks, CA: Sage.

Hox, J. J., & Maas, C. J. M. (2001). The accuracy of multilevel structural equationmodeling with pseudobalanced groups and small samples. Structural EquationModeling: A Multidisciplinary Journal, 8, 157e174. http://dx.doi.org/10.1207/S15328007SEM0802_1.

Kim, U., Triandis, H. C., Kagtbas, C., Choi, S.-C., & Yoon, G. (Vol. Eds.). (1994). Cross-cultural research and methodology series: Vol. 18. Individualism and collectivism:Theory, method, and applications. Thousand Oaks, CA: Sage.

Klassen, R. M., Bong, M., Usher, E. L., Chong, W. H., Huan, V. S., Wong, I., et al. (2009).Exploring the validity of a teachers self-efcacy scale in ve countries.Contemporary Educational Psychology, 34, 67e76. http://dx.doi.org/10.1016/j.cedpsych.2008.08.001.

Klassen, R. M., & Chiu, M. M. (2010). Effects on teachers self-efcacy and jobsatisfaction: teacher gender, years of experience, and job stress. Journal ofEducational Psychology, 102, 741e756. http://dx.doi.org/10.1037/a0019237.

Kobayashi, C., & Greenwald, A. G. (2003). Implicit-explicit differences in self-enhancement for Americans and Japanese. Journal of Cross-Cultural Psychol-ogy, 34, 522e541. http://dx.doi.org/10.1177/0022022103257855.

Kurman, J. (2003). Why is self-enhancement low in certain collectivist cultures?Journal of Cross-Cultural Psychology, 34, 496e510. http://dx.doi.org/10.1177/0022022103256474.

Lin, H.-L., Gorrell, J., & Taylor, J. (2002). Inuence of culture and education on U.S.and Taiwan preservice teachers efcacy beliefs. Journal of Educational Research,96, 37e46. http://dx.doi.org/10.1080/00220670209598789.

Markus, H. R., & Kitayama, S. (1991). Culture and the self: implications for cognition,emotion, and motivation. Psychological Review, 98, 224e253. http://dx.doi.org/10.1037/0033-295X.98.2.224.

Mau, W. C. (2000). Cultural differences in career decision-making styles and self-

S. Vieluf et al. / Teaching and Te102efcacy. Journal of Vocational Behavior, 57, 365e378. http://dx.doi.org/10.1006/jvbe.1999.1745.Mo, A., Pazzaglia, F., & Ronconi, L. (2010). When being able is not enough. Thecombined value of positive affect and self-efcacy for job satisfaction inteaching. Teaching and Teacher Education, 26, 1145e1153. http://dx.doi.org/10.1016/j.tate.2010.02.010.

Nisbett, R. E., & Cohen, D. (1996). Culture of honour: The psychology of violence in theSouth. Boulder, CO: Westview.

OECD (Ed.). (2009). Creating effective teaching and learning environments: First resultsfrom TALIS. Paris, France: OECD Publishinghttp://dx.doi.org/10.1787/9789264068780-en.

OECD (Ed.). (2010a). PISA 2009 results: What students know and can do studentperformance in reading, mathematics and science. Paris, France: OECD Publish-inghttp://dx.doi.org/10.1787/9789264091450-en.

OECD (Ed.). (2010b). TALIS technical report. Paris, France: OECD Publishinghttp://dx.doi.org/10.1787/9789264079861-en.

OECD (Ed.). (2012). PISA 2009 technical teport. Paris, France: OECD Publishinghttp://dx.doi.org/10.1787/9789264167872-en.

Pajares, F. (2007). Culturalizing educational psychology. In F. Salili, & R. Hoosain(Eds.), Research in multicultural education and international perspective. Culture,motivation, and learning. A multicultural perspective (pp. 19e42). Charlotte, NC:Information Age Publishing.

Peristiany, J. G. (Ed.). (1965). Honour and shame: The values of the Mediterraneansociety. London, United Kingdom: Weinfeld and Nicolson.

Poortinga, Y. H., Van de Vijver, F. J. R., & Van Hemert, D. A. (2002). Cross-culturalequivalence of the big ve. A tentative interpretation of the evidence. InR. R. McCrae, & J. Allik (Eds.), The ve-factor model of personality across cultures(pp. 271e292). New York, NY: Kluwer Academic/Plenum Publishers.

Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications anddata analysis methods (2nd ed.). Newbury Park, CA: Sage.

Raudenbush, S. W., Rowan, B., & Fai Cheong, Y. (1992). Contextual effects on the self-perceived efcacy of high school teachers. Sociology of Education, 65, 150e167.http://dx.doi.org/10.2307/2112680.

Ross, J. A. (1992). Teacher efcacy and the effect of coaching on student achieve-ment. Canadian Journal of Education, 17, 51e65. http://dx.doi.org/10.2307/1495395.

Ross, J. A. (1995). Strategies for enhancing teachers beliefs in their effectiveness:research on a school improvement hypothesis. Teachers College Record, 97,227e251.

Rotter, J. (1966). Generalized expectancies for internal versus external control ofreinforcements. Psychological Monographs, 80. whole issue.

Schooler, C. (1990). Psychological effects of complex environments during the lifespan: a review and theory. In C. Schooler, & K. W. Schaie (Eds.), Cognitivefunctioning and social structure over the life course (pp. 24e49). Norwood, NJ:Ablex.

Schwartz, S. H. (1992). Universals in the content and structure of values: theo-retical advances and empirical tests in 20 countries. In M. Zanna (Ed.), Ad-vances in experimental social psychology, Vol. 25 (pp. 1e65). Orlando, FL:Academic Press.

Schwartz, S. H. (1994). Beyond individualism/collectivism: new dimensions ofvalues. In U. Kim, H. C. Triandis, C. Kagtbas, S. C. Choi, & G. Yoon (Eds.),Individualism and collectivism: Theory application and methods (pp. 85e119).Newbury Park, CA: Sage.

Schwartz, S. H. (2003). A proposal for measuring value orientations across nations.Chapter 7 in the questionnaire development package of the European SocialSurvey. Retrieved from http://www.europeansocialsurvey.org/weuropean/images/downloads/Dec2012/chapter_07.pdf.

Skaalvik, E. M., & Skaalvik, S. (2007). Dimensions of teacher self-efcacy and re-lations with strain factors, perceived collective teacher efcacy, and teacherburnout. Journal of Educational Psychology, 99, 611e625. http://dx.doi.org/10.1037/0022-0663.99.3.611.

Smith, P. B., Trompenaars, F., & Dugan, S. (1995). The Rotter locus of control scale in43 countries: a test of cultural relativity. International Journal of Psychology, 30,377e400. http://dx.doi.org/10.1080/00207599508246576.

Smylie, M. A. (1988). The enhancement function of staff development: organiza-tional and psychological antecedents to individual teacher change. AmericanEducational Research Journal, 25, 1e30. http://dx.doi.org/10.3102/00028312025001001.

Snijders, T. A. B., & Bosker, R. J. (1999). Multilevel analysis: An introduction to basicand advanced multilevel modeling. London, United Kingdom: Sage.

Tatto, M. T., Schwille, J., Senk, S., Ingvarson, L., Peck, R., & Rowley, G. (2008). Teachereducation and development study in mathematics (TEDS-M): Conceptual frame-work. East Lansing, MI: Teacher Education and Development InternationalStudy Center, College of Education, Michigan State University.

Triandis, H. C. (1995). Individualism and collectivism. Boulder, CO: Westview Press.Tschannen-Moran, M., & Woolfolk Hoy, A. (2001). Teacher efcacy: capturing an

elusive construct. Teaching and Teacher Education, 17, 783e805. http://dx.doi.org/10.1016/S0742-051X(01)00036-1.

Tschannen-Moran, M., Woolfolk Hoy, A., & Hoy, W. K. (1998). Teacher efcacy: itsmeaning and measure. Review of Educational Research, 68, 202e248. http://dx.doi.org/10.3102/00346543068002202.

UNESCO. (2006). International standard classication of education I S C E D 1997.Retrieved from http://www.uis.unesco.org/Library/Documents/isced97-en.pdf.

Uskul, A. K., Oyserman, D., & Schwarz, N. (2010). Cultural emphasis on honor,modesty, or self-enhancement: implications for the survey-response process:

r Education 35 (2013) 92e103survey methods in multinational, multiregional and multicultural context. InJ. A. Harkness, M. Braun, B. Edwards, T. Johnson, L. Lyberg, P. Ph. Mohler, B.-