The current study aimed to examine the relationship between school composition variables (socioeconomic and migration background of students) and teacher collaboration and teacher self-efficacy.
Multilevel analyses of variance were conducted using the Dutch Teaching and Learning International Survey 2018 data.
Teacher collaboration was positively related to teacher self-efficacy in classroom management, instruction and student engagement as well as to teacher self-efficacy in multicultural classrooms. Furthermore, the analyses showed a positive relationship between socioeconomic background and teacher self-efficacy in classroom management and student engagement, but a negative relationship between migration background and teacher self-efficacy in classroom management. Finally, teachers working in schools with higher proportions of students from low socioeconomic and/or migration backgrounds reported higher levels of self-efficacy to teach in multicultural classrooms.
Former research on this topic has generally been small in scale and has employed qualitative data in urban contexts only, mostly focusing on either socioeconomic or migration backgrounds. The present study used a quantitative approach and examined both types of diversity.
Introduction
The research on teacher self-efficacy (TSE) in diverse schools generally shows that teachers feel less efficacious in schools with relatively high proportions of students from migration and low socioeconomic backgrounds (e.g. Geerlings et al., 2018; Tucker et al., 2005). TSE refers to teachers' personal belief to be able to perform all the actions needed for the teaching job (Täshner et al., 2025) and TSE is strongly associated with practises related to classroom quality (Zee and Koomen, 2016) and evaluated teacher performance (Klassen and Tze, 2014). Furthermore, high levels of TSE are associated with high levels of student learning and motivation (TALIS, 2018). TSE is highly relevant to study in relation to equity and diversity in schools because it has a strong connection to beneficial outcomes (see Täschner et al., 2025). This relevance is especially important considering the growing diversity in student backgrounds, which is driven by migration patterns (Vertovec, 2023), as well as the ongoing gaps in learning outcomes and educational attainment (OECD, 2024). An important contributor to TSE is the extent to which teachers collaborate (see, e.g. Hargreaves and Fullan, 2012; Muijs et al., 2004; Vangrieken et al., 2015). The impact of teacher collaboration has been identified in some research on schools with diverse student populations but is not consistently observed. Furthermore, most previous studies have been small in scale and have employed qualitative approaches (see, e.g. Preston et al., 2017).
In the current study, we investigate the impact of teacher collaboration on TSE in schools that vary according to students' socioeconomic and migration backgrounds by conducting secondary analyses of the 2018 Dutch Teaching and Learning International Survey (TALIS) data. TALIS is one of the largest OECD education studies. The TALIS 2018 survey saw participation from 260,000 teachers and 15,000 school leaders across 48 countries and economies. The Dutch TALIS 2018 dataset that is used for the present study comprises data from 1884 secondary school teachers and 116 schools. The survey gathers teachers' school leaders' perceptions on work-related aspects such as professional development, teaching beliefs, teacher self-efficacy, practises, work assessment, job satisfaction, professional needs, workload and stress, leadership, work climate, and school population composition (TALIS 2018).
Below, we elaborate on the concepts of teacher collaboration and TSE, with a specific focus on diversity in migration and socioeconomic backgrounds.
Teacher collaboration
A recent review by Weddle (2022) has shown that research on teacher collaboration consistently shares a set of characteristics of effective teacher collaboration, mostly referring to the seminal work of Stoll et al. (2006). These characteristics include “shared values and vision, collective responsibility, reflective professional inquiry, and the promotion of group as well as individual learning” (Weddle, 2022, p. 8). Generally, collaboration can vary from incidental interactions to collaborating in strong professional learning communities (PLCs) involving the above characteristics. The TALIS theoretical framework views teacher collaboration as “a prominent and essential pillar of teacher professionalism” (TALIS, 2018, p. 149) and Stoll's work is foundational to the two TALIS scales that measure teacher collaboration on teacher exchange and professional collaboration. More concretely, the teacher exchange scale refers mostly to shared values and collective responsibility, and the professional collaboration scale refers mostly to reflective professional inquiry and group learning. The individual learning characteristic seems to more implicit in the two scales.
Teacher collaboration seems to be important in schools serving students from diverse backgrounds in terms of migration and socioeconomic status (SES). Preston et al. (2017) distinguished eight essential components of effective urban high schools (where “effectiveness” is understood as improving the achievement of disadvantaged students), one of which is a culture of learning and professional behaviour. Such a culture comprises a variety of formats that involve teacher collaboration, such as teacher learning communities, teacher inquiry groups, PLCs, and communities of teacher practice. However, most of this research has been small in scale and has employed qualitative data and case study designs in urban contexts only (Preston et al., 2017). Therefore, the question remains whether the positive role of teacher collaboration can be confirmed using a quantitative approach and whether this role is similar in schools that vary according to student backgrounds.
Teacher self-efficacy
Self-efficacy, a concept introduced in Bandura's social-cognitive theory (1977), can be defined as one's judgement of their own ability to execute a given type of task (Bandura, 1977). Efficacious teachers believe that they have at least some control over what happens in the classroom, which results in higher levels of teacher effort and, subsequently, higher levels of student performance (Malinen et al., 2013; Tschannen-Moran and Hoy, 2001; Zee and Koomen, 2016). In line with the review of TSE research of Zee and Koomen (2016), TALIS (OECD, 2014) conceptualises TSE as a multidimensional concept consisting of three core components: classroom management (teachers' beliefs regarding their ability to establish an orderly learning environment, for example to be able to communicate clear expectations with regard to student behaviour and get students to follow classroom rules), instruction (teachers' confidence in using a variety of teaching and assessment practises, for example to be able to craft good questions, and employ alternative instructional strategies), and student engagement (teachers' beliefs regarding their capacity to support and motivate their students, for example to be able to help students to value learning and get them to believe they can do well). Zee and Koomen (2016) showed that TSE was consistently related to teacher psychological well-being: self-efficacious teachers seem to suffer less from stress, emotional exhaustion, and burnout (2016).
Tucker et al.’s (2005) review of studies on TSE in diverse classrooms describes research showing that teacher efficacy in working with diverse students is often low and argue that this finding may in part explain the achievement gap. They attribute these lower scores to racial attitudes and a general sense of unpreparedness for diverse classrooms. In a similar vein, Romijn et al. (2020) argue, referring to the work of Gay (2002) on culturally responsive pedagogy, that teaching in diverse classrooms requires a specific set of skills, more than just awareness and respect for diverse viewpoints. It also requires knowledge on students' backgrounds and validating students' cultural identities in the classroom. Given the little preparation for teaching in diverse classrooms in many programs for initial teacher education (see also Severiens et al., 2014; Alhanachi et al., 2021), a lower TSE is to be expected. However, a few recent studies (Fackler et al., 2020; Romijn et al., 2020) have shown a positive relationship between school diversity and TSE. Fackler et al. (2020) analysed the TALIS, 2018 data (ISCED level 2) and showed higher levels of TSE in instruction, student engagement, and classroom management in schools with larger proportions of low-SES students. In their study on primary school teachers, Romijn et al. (2020) also observed higher levels of TSE in diverse classrooms. They attributed their finding to the fact that many urban teachers explicitly choose to teach in this particular context. These teachers' generally positive attitude towards diversity may positively impact their TSE levels.
Some of the studies that investigated TSE in diverse schools did not use general measures of self-efficacy but specific measures with reference to the diversity of the student population. Romijn et al. (2020) proposed a two-dimensional model of TSE that distinguishes between general TSE and diversity-related TSE. Similarly, Siwatu (2011; Siwatu and Starker, 2010) described a specific measure for the concept of culturally responsive teaching self-efficacy (CRTSE). Examples of CRTSE are being able to ensure that students with and without a migration background work together, and being able to reduce ethnic stereotyping amongst students. The work by Siwatu showed low efficacy in culturally responsive teaching among elementary and middle school preservice teachers. The findings of Siliunas et al. (2024), who employed the same CRTSE measure, were consistent with Siwatu's findings in the sense that self-efficacy doubts were higher with regard to addressing students' cultural and linguistic differences. Therefore, it seems relevant when investigating TSE in diverse schools, to also examine TSE specifically related to teaching in diverse classrooms.
The relationships between teacher collaboration and TSE
A possible explanation of the relationship between collaboration and TSE is that interacting positively in, for example, a PLC can provide the relevant sources of self-efficacy (Täschner et al., 2025; Weiβenrieder et al., 2015, see also Lakshmanan et al., 2011; Mintzes et al., 2013). In their meta-analysis, Täschner et al. (2025) have investigated interventions on TSE employing a framework of the possible four sources of TSE, that is mastery and vicarious experiences, social persuasion and physiological reactions. Teacher collaboration is one of the possible interventions that may provide a source of social persuasion. They conclude that interventions including mastery and vicarious experiences have a positive impact on TSE, and this is also true for social persuasion but to a lesser extent. The role of physiological reactions remains unclear as this source was hardly ever investigated. Täschner et al. (2025) conclusion with regard to the minor role of social persuasion (described as verbal support and encouragement) indicates that PLCs and teacher collaboration need to include more than “social persuasion” activities and offer mastery and vicarious experiences as well. Examples of these are trying out new teaching methods and observation in each other's classrooms and providing feedback.
Relevant to the current paper is the role of teacher collaboration in diverse classrooms. The number of studies that examine both TSE and teacher collaboration in the context of diversity is, however, small. Takahashi's study on the development of self-efficacy in communities of practice (2011), conducted among junior high school teachers in classrooms with high proportions of ethnic minority and low-income students, showed that communities of practice can help teachers escape a negative cycle of low achievement and low efficacy beliefs. Anderson and Olivier (2022) arrive at a similar conclusion in their study on the intersection of PLCs, teacher and collective efficacy, and poverty in schools. Their results indicate that PLCs in the high poverty schools they studied, strongly influence TSE (it is relevant to note here that diversity and poverty are often empirically related, but differences in achievement levels and efficacy beliefs are caused by different underlying factors). Mo et al. (2021) added to the literature by analysing the Finnish TALIS 2018 data (ISCED level 2), focusing on the relationship between PLCs and self-efficacy in multicultural classrooms. They found that the “more teachers participate in PLC activities, the more they feel self-efficacy with teaching in multicultural classrooms” (2021, p. 618). Moreover, their study showed that PLCs seem to stimulate collective enactment of multicultural practises. Mo et al. (2021) proposed that PLCs provide a suitable context for learning new practises and increasing self-efficacy in multicultural settings. Given the high levels of teacher collaboration in PLCs, by inference, we can argue that teacher collaboration may be similarly positively related to TSE in schools serving diverse student populations. Finally, He and Bagwell's recent study (2025) on relevant self-efficacy forming experiences with regard to supporting multilingual learners emphasised bridging theory and practice and collaborative learning in the form of guided exchanges, peer observations, and reflective dialogues. In summary, the studies described above suggest that teacher collaboration is generally seen as an effective feature in diverse schools as it contributes to a variety of positive outcomes, among which is TSE. Furthermore, teacher collaboration may offer a way to escape the self-perpetuating cycle in which low student achievement leads to low TSE, which in turn negatively impacts student achievement. In other words, there may be a moderating effect of teacher collaboration in the sense that high levels of collaboration counteract the negative relationship between TSE and the proportion of students from low-SES and migration backgrounds. In the present study, using the Dutch TALIS 2018 data, we aim to investigate the following research questions:
Does teacher collaboration have a positive impact on TSE?
Does the impact of teacher collaboration on TSE vary according to the composition of the student population in terms of migration and SES background?
First, it was hypothesized that teacher collaboration would be positively related to all types of teacher self-efficacy in providing classroom management, instruction and student engagement (H1). Second, it was hypothesized that student population (i.e. percentages of low-SES students and students with a migration background) would be negatively related to TSE in providing classroom management, instruction, and student engagement (H2). Third, it was hypothesized that teacher collaboration might moderate the negative relation between student population and TSE in a protective manner; for schools with low levels of teacher collaboration the negative relation between percentages of students from low SES and migration backgrounds and TSE exist, but for schools with high levels of teacher collaboration there might be no negative- or even a positive effect (H3).
The hypotheses were tested in two multilevel regression models. In the first model, we examined the relationships between student SES, student migration background, teacher collaboration, and TSE related to classroom management, instruction, and student engagement, involving all participating teachers in the TALIS 2018 sample. Because the TALIS data makes it possible to distinguish teachers who do and do not report working in a school with a high level of diversity, we ran a second model. This second model tested the same relationships as the first model, but only for teachers who reported teaching in schools with high levels of diversity. In addition, exclusively in this subsample of teachers, TALIS also measured self-efficacy in teaching in diverse classrooms. Therefore, the second model also included TSE in teaching in diverse classrooms. In other words, the hypotheses were tested first by testing a model including the TSE components of classroom management, instructional practises, and student engagement, and second by testing a model including the TSE component of teaching in diverse classrooms in a subsample of teachers that taught in diverse classrooms.
Materials and method
Respondents
The sample consisted of 1,884 teachers and 125 school leaders from 116 secondary (ISCED 2) schools that participated in the Dutch TALIS 2018 (Sapulete et al., 2018). Of the teachers, 54% were female and 46% male, with M = 42.8 years and SD = 11.85, ranging between 19 and 71 years. Of the school leaders, 38% were female and 62% male, with a = 53.9 years and SD = 7.76, ranging between 32 and 68 years. Response rates were 80% among schools and 76% among teachers.
Procedure
The data were collected in two steps: first, a random sample of 200 schools was selected, and then, within these schools, a random sample of 20 teachers was selected (see TALIS, 2018). The Dutch national TALIS partner worked closely with all relevant stakeholders (the Ministry of Education, teacher unions, and other relevant authorities) to enable sufficient participation (see Sapulete et al., 2018). The TALIS 2018 procedure implemented rigorous standards in order to ensure valid, reliable, and comparable questionnaires, data, and results, including field trials, and Statistics Canada performed all sampling and weighting procedures. All national partners were provided with extensive guidance throughout the procedure.
Within each school, a minimum response rate of 50% among the selected teachers was used as a threshold for inclusion of the school in the TALIS 2018 analysis (TALIS 2018). The principal and teacher questionnaires were administered between March and July 2018 (for more information, see TALIS, 2018).
The Teaching and learning international survey (TALIS)
TALIS had two versions, one for the school leaders consisting of 45 and one for teachers consisting of 56 (multi-item) questions on some background variables (type of school, experience etc.) and topics as, leadership, climate, induction, job satisfaction, professional self-efficacy, diversity, that were measured with multiple items using a four-point Likert scale. For the scaling of the items, scores were standardised around a M = 10 and SD = 2. The TALIS 2018 organisation considered Cronbach's alpha (α) > 0.6 to be acceptable and >0.7 as good indices for the scale. The exact number of items and the reliability indices of the scales used in this study are described below under independent and dependent variables. For a complete description of the scaling and reliability and validity tests, we refer to Chapter 11 of the technical report (TALIS, 2018).
Basic assumptions
Because of the large sample sizes and high response rates (N = 1884 for the first analysis and N = 1,074 for the second), in addition to the strict sampling procedure in order to recruit representative samples (see, TALIS technical report, 2018), normality of the data can be assumed (Zymond, 2023). As the survey was conducted under educational professionals for whom the survey outcomes are directly relevant and the survey was thoroughly tested via an extensive pilot (see TALIS technical report, 2018), we included all responses in our analysis and considered those as valid experiences and including those would provide a more reliable picture of the outcomes than excluding those. Plotting the independent variables (standardized scores) against the dependent variables classroom management (T3SECLS), instruction (T3SEINS), and establishing student engagement (T3SEENG), show that the assumption of homoscedasticity is not violated.
To check the assumption that our model will be tested more accurately using a multilevel approach instead of a unilevel approach, we will compare an independence model to a 0-model. If the model significantly improves by taking variance differences between schools into account (the independence model), the models are tested by multilevel analyses.
Variables used in this study
TSE and teacher collaboration were operationalised based on the TALIS conceptual framework and its measures. Following the TALIS terminology, we use the term “school composition” to indicate school diversity. Below are descriptions of the scales and reliability for the Dutch data, using Cronbach's alpha (α) for sets of items with tau-equivalence (i.e. where all the items have equal covariance with the true score) and omega coefficients where tau-equivalence was violated (see Deng and Chan, 2017).
Independent variables
Level 1: Teacher collaboration (TCOOPS) was composed of two subscales: exchange and coordination for teaching (four items, α = 0.69) and professional collaboration (four items, α = 0.69) (TALIS, 2018). As the present study was focused on all cooperative activities, the composite variable “teacher collaboration” (α = 0.78) was used. In the analysis, the teacher collaboration variable was centred around the grand mean (see Appendix 1 for the items).
Level 2: School diversity was measured using two school composition variables. These variables were based on the principal's perception of the percentage of students with (1) a socio-economically disadvantaged home and (2) an immigrant status or migrant background (1 = none, 2 = 0–10%, 3 = 11–30%, 4 = 31–60%, and 5 = >60%) (see Table 1). The majority of principals indicated that 0–10% of their school's students had a low-SES background, and 1–10% had a migration background. No schools were reported to be without students from a low-SES background, and 13 schools were reported to have no students with a migration background. The scores for SES and migration background were converted into dummies, with the largest category being the reference group; for SES, the reference group was 1–10% of students with a low-SES background. This means that this reference group was compared to schools with more than 10% of students with a low-SES background, as none of the included schools had no low-SES students. For migration background, the reference group was 1–10% of students with a migration background. This means that this reference group was compared to schools with no students with a migration background, as well as to schools with more than 10% of students with a migration background (see Table 1 for the frequencies of SES and migration per category). For reasons of power, category 4 = 31–60%, and 5 = >60% were combined into one dummy variable.
School principals' estimation of pupils with low socioeconomic status (SES) and migration background in their schools and the number of schools and teachers in each of the categories
| SES | Migration background | ||||
|---|---|---|---|---|---|
| Number of Schools (Teachers) | % | Number of Schools (Teachers) | % | ||
| Valid | 0% | 0 | 0 | 9 (136)1 | 7.8 (7.3) |
| 1%–10% | 69 (1,093)0 | 59.5 (57.6) | 81 (1332)0 | 69.8 (70.1) | |
| 11%–30% | 33 (561)1 | 28.4 (29.9) | 15 (263)2 | 12.9 (13.6) | |
| 31%–60% | 6 (120)2 | 5.2 (6.2) | 3 (43)3 | 2.6 (3.4) | |
| >60% | 1 (11)2 | 0.9 (1.1) | 1 (11)3 | 0.9 (0.6) | |
| Total valid | 109 (1785) | 94.0 (94.9) | 109(1785) | 94.0 (94.9) | |
| Missing | 7 (110) | 6.0 (5.1) | 7 (110) | 6.0 (5.1) | |
| Total | 116 (1895) | 100.0 | 116 (1895) | 100.0 | |
Note(s): 0reference group, 1Dummy 1, 2Dummy 2, 3Dummy 3
Dependent variables
TSE was composed of three subscales measuring the extent to which teachers feel proficient in classroom management (T3SECLS, consisting of four items, Ω = 0.81), instruction (T3SEINS, consisting of four items, Ω = 0.68), and establishing student engagement (T3SEENG, consisting of four items, Ω = 0.68) (TALIS, 2018) (see Appendix 1 for the items). A scale for efficacy in multicultural classrooms (T3SEFE, consisting of five items, Ω = 0.80) was measured in the sample of teachers who answered “Yes” to the question “Have you ever taught in a classroom of children with diverse ethnic-cultural backgrounds?” Table 2 shows the frequencies of these teachers per school composition (SES and migration background) category. As we were interested in the different types of self-efficacy, we used these subscales for the analyses.
School principals' estimation of pupils with low socioeconomic status (SES) and migration background in their schools and the number of schools and teachers in each of the categories in the sample of teachers who have taught in a classroom of children with diverse ethnic-cultural backgrounds
| SES | Migration background | ||||
|---|---|---|---|---|---|
| Number of Schools (Teachers) | % | Number of Schools (Teachers) | % | ||
| Valid | 0% | 0 | 0 | 9 (50)1 | 7.0 (4.7) |
| 1%–10% | 68 (550)0 | 59.1 (52.1) | 81 (751)0 | 70.4 (69.9) | |
| 11%–30% | 33 (384)1 | 28.7 (35.8) | 15 (196)2 | 13.0 (18.2) | |
| 31%–60% | 6 (81)2 | 5.2 (7.5) | 3 (28)3 | 2.6 (2.6) | |
| >60% | 1 (8)2 | 0.9 (0.7) | 1 (8)3 | 0.9 (0.7) | |
| Total valid | 108 (1,033) | 93.9 (96.2) | 109 (1,033) | 93.9 (96.2) | |
| Missing | 7 (41) | 6.1 (3.8) | 7(41) | 6.1 (3.8) | |
| Total | 115 (1,074) | 100.0 | 115 (1,074) | 100.0 | |
Note(s): 0reference group, 1Dummy 1, 2Dummy 2, 3Dummy 3
Analyses
Before answering the research questions, we evaluated whether multilevel modelling was more suitable than a unilevel approach. This involved comparing a null model, where variances between clusters were set to zero, with an independence model that accounted for variances within and between clusters. We used a Chi-square test of the deviance to measure the fit of our models, with a cut-off value of Δ-2LL/df > 3.84, which corresponds to the 95% critical value for a squared normal deviate for a single cell (Breslow and Day, 1987, p. 130). As indicated in Table 4, the independence model demonstrated a significantly better fit compared to the null model, where the variance between clusters was fixed.
Two analyses were conducted. First, a regression model was tested on the sample of all teachers to investigate the relationship between teacher collaboration (measured at the teacher level), school composition (SES and migration background, measured at the school level), and the three types of general TSE. Second, a similar regression model containing the added scale of TSE in multicultural classrooms was tested on the subsample of teachers that taught in diverse classrooms.
We first inspected the variance of the dependent variables by comparing the model fit of the unconditional model, in which between-school variance is fixed, to the independence model, which takes the different variances within and between schools into account. Tables 4 and 6 show that for the TSE scales, the difference between the deviance of the unconditional and independence models was significant. This indicates that the variance within and between schools differs significantly, which implies that multilevel analysis is suitable.
Correlations between teacher collaboration (TCOOPS) and teacher self-efficacy in classroom management (SECLSS), student engagement (SEENGS), and instruction (SEINSS)
| T3COOP | T3SECLS | T3SEINS | T3SEENG | |
|---|---|---|---|---|
| M (SD) | 8.95 (1.75) | 12.58 (1.80) | 12.54 (1.55) | 11.87 (1.59) |
| T3SECLS | 0.08* | |||
| T3SEINS | 0.18** | 0.34** | ||
| T3SEENG | 0.20** | 0.50** | 0.53** | – |
Note(s): *significance at the level of p < 0.05, **significance at the level of p < 0.01, *** significance at the level of p < 0.001
Regression model 1: Relationships between teacher collaboration, school composition, and TSE in classroom management, instruction, and student engagement
| All teachers | 0 model | Independence model | Level 1 | Level 2 | |
|---|---|---|---|---|---|
| Predictor | Predictors | ||||
| Estimate (SE) | Estimate (SE) | Estimate (SE) | Estimate (SE) | ||
| Level 1 | Classroom Management | ||||
| Fixed | Mean | 12.58(0.05)*** | 12.58(0.05)*** | 12.58(0.05)*** | 12.52(0.07)*** |
| Teacher collaboration | 0.08(0.03)** | 0.08(0.03)** | |||
| Random | σ2e | 3.24(0.12)*** | 3.18(0.12)*** | 3.20(0.12)*** | 3.15(0.13)*** |
| σ2u0 | 0.05(0.04) | 0.02(0.02) | 0.04(0.06) | ||
| Level 2 | 11–30% low SES | 0.25(0.12)* | |||
| >30% low SES | <0.01(0.13) | ||||
| 0% migration | 0.20(0.16) | ||||
| 11–30% migration | −0.14(0.15) | ||||
| >30% migration | −0.34(0.12)** | ||||
| Level 1 | Instruction | ||||
| Fixed | Mean | 12.54(0.04)*** | 12.54(0.04)*** | 12.54(0.04)*** | 12.53(0.06)*** |
| Teacher collaboration | 0.15(0.02)*** | 0.16(0.02)*** | |||
| Random | σ2e | 2.40(0.07)*** | 2.33(0.08)*** | 2.29(0.08)*** | 2.28(0.08)*** |
| σ2u0 | 0.06(0.03)* | 0.03(0.02) | 0.05(0.04)* | ||
| Level 2 | 11–30% low SES | 0.08(0.09) | |||
| >30% low SES | −0.03(0.25) | ||||
| 0% migration | 0.02(0.17) | ||||
| 11–30% migration | −0.17(0.13) | ||||
| >30% migration | −0.12(0.20) | ||||
| Level 1 | Student Engagement | ||||
| Fixed | Mean | 11.87(0.05)*** | 11.87(0.05)*** | 11.87(0.05)*** | 11.83(0.06)*** |
| Teacher collaboration | 0.18(0.02)*** | 0.18(0.02)*** | |||
| Random | σ2e | 2.52(0.08)*** | 2.41(0.08)*** | 2.36(0.08)*** | 2.32(0.08)*** |
| σ2u0 | 0.10(0.04)** | 0.04(0.02) | 0.10(0.05)* | ||
| Level 2 | 11–30% low SES | 0.21 (0.11)* | |||
| >30% low SES | −0.07(0.25) | ||||
| 0% migration | 0.02(0.17) | ||||
| 11–30% migration | −0.17(0.13)* | ||||
| >30% migration | −0.24(0.20) | ||||
| –2LL (Parameters) | 20191.29(6) | 20165.08(9) | 18940.24 (15) | 17890.42(33) | |
| Δ–2LL(df)a | 26.21(3)*** | 1,224.84(6) | 1,049.82(8)*** | ||
| ICC Classroom Management | 1.80% | ||||
| ICC Instruction | 2.70% | ||||
| ICC Student Engagement | 4.00% | ||||
Our multilevel models (regressed on T3SECLS, T3SEINS, and T3SEENG for analysis 1 for all teachers, and in analysis 2 with the addition of T3SEFE for the teachers who reported having taught in diverse classrooms) were built using the following steps:
The level 1 predictor “teacher collaboration” was added.
The level 2 predictors “SES” (two dummy variables) and “migration background” (three dummy variables) were added.
Random slope models for each dependent variable were tested with random slopes for teacher collaboration.
In the case of model improvement using random slope models instead of fixed slope models, analysis tests for cross-level interactions were conducted.
Results
Table 3 presents the correlations between teacher collaboration (T3COOP) and TSE in classroom management (T3SECLS), instruction (T3SEINS), and student engagement (T3SEENG) for the full sample of teachers. The TSE subscales show a modest but significant correlation with teacher collaboration. The three subscales of TSE intercorrelate moderately.
General TSE
The results showed that teacher collaboration was a significant predictor of all three TSE subscales, reflecting positive relationships between teacher collaboration and TSE in classroom management, instruction, and student engagement. This finding reflects that teachers who were engaged in exchange and coordination and who collaborated professionally reported more self-efficacy in their classroom management, instruction, and engagement with students. In other words, teachers who, for example, work with colleagues, who attend team conferences, and who engage in joint activities, also feel they are more able to calm students who are disruptive (classroom management), more able to use a variety of assessment strategies (instruction) and able to motivate students who show low interest (student engagement).
The level 2 school composition variables of SES and migration background contributed significantly to model fit. For SES, compared with the reference group (teachers in schools with 1–10% of students with a low-SES background), teachers working in schools with 11–30% of students with a low-SES background scored higher on TSE in classroom management and student engagement. For migration background, compared with the reference group (teachers in schools with 1–10% of students with a migration background), teachers working in schools with >30% of students with a migration background reported lower TSE in classroom management. In addition, compared with the reference group, teachers working in schools with 11–30% of students with a migration background had lower TSE in student engagement. (H3) Adding random slopes for the relationship between the TSE scales and teacher collaboration for all units of the level 2 variables did not further improve the model: not for TSE in classroom management (σ2T3SECLS < 0.01, p = .782), nor for instruction (σ2 T3SEINS = 0.01, p = .549) or student engagement (σ2 T3SEENG < 0.01, p = .856). Therefore, no cross-level interactions were tested. These results imply that teacher collaboration did not moderate the relationship between the level 2 composition variables (SES and migration background) and the TSE scales. This means that the relationship between teachers' collaborative practises (for example teaching jointly with colleagues) and their self-efficacy (for example being able to get students to follow classroom rules) does not depend on their classroom composition.
TSE diversity
The second model we tested, in the subsample of teachers who taught in diverse classrooms, also included the TSE measure related to teaching in diverse classrooms. Table 5 presents the correlations between teacher collaboration (T3COOP) and TSE in classroom management (SECLSS), student engagement (SEENGS), instruction (SEINSS), and diversity (SEFE) for the sample of teachers who taught in diverse classrooms. The TSE subscales show a modest but significant correlation with teacher collaboration. The four subscales of TSE intercorrelate moderately.
Correlations between teacher collaboration (TCOOPS) and teacher self-efficacy in classroom management (SECLSS), student engagement (SEENGS), instruction (SEINSS), and multicultural classrooms (T3SEFE)
| T3TCOOP | T3SECLS | T3SEINS | T3SEENG | T3SEFE | |
|---|---|---|---|---|---|
| M (SD) | 9.88 (1.81) | 12.64 (1.81) | 12.60 (1.53) | 11.95 (1.56) | 11.07 (2.12) |
| T3SECLS | 0.06* | ||||
| T3SEINS | 0.15** | 0.31** | |||
| T3SEENG | 0.19** | 0.46** | 0.51** | ||
| T3SEFE | 0.15** | 0.22** | 0.26** | 0.34** | – |
Note(s): *significance at the level of p < 0.05, **significance at the level of p < 0.01, *** significance at the level of p < 0.001
H1 in analysis 2 (see Table 6) The results on the subsample of teachers (N = 1,074, also see Table 2) who have taught in diverse classrooms show that teacher collaboration was a significant predictor of TSE in instruction and student engagement, reflecting positive relationships between teacher collaboration and TSE in instruction and student engagement. Unlike in analysis 1, teacher collaboration did not reach significance as a predictor of TSE in classroom management (p = 0.061). The results in this subset of teachers showed a significant positive relation between teacher collaboration and TSE in multicultural classrooms (p = 0.04).
Regression model 2: Relationships between teacher collaboration, school composition, and TSE in classroom management, instruction, student engagement, and multicultural classrooms
| Teachers teaching in diverse classrooms | Independence model | Level 1 | Level 2 | |
|---|---|---|---|---|
| Predictor | Predictors | |||
| Estimate (SE) | Estimate (SE) | Estimate (SE) | ||
| Level 1 | Classroom Management | |||
| Fixed | Mean | 12.58(0.05)*** | 12.58(0.05)*** | 12.61(0.09)*** |
| Teacher collaboration | 0.08(0.03)** | 0.06(0.03) | ||
| Random | σ2e | 3.18(0.12)*** | 3.20(0.12)*** | 3.15(0.16)*** |
| σ2u0 | 0.05(0.04) | 0.02(0.02) | 0.04(0.06) | |
| Level 2 | 11–30% low SES | 0.02(0.12) | ||
| >30% low SES | 0.15(0.13) | |||
| 0% migration | −0.06(0.24) | |||
| 11–30% migration | −0.17(0.16) | |||
| >30% migration | −0.34(0.22) | |||
| Level 1 | Instruction | |||
| Fixed | Mean | 12.58(0.05)*** | 12.54(0.04)*** | 12.63(0.07)*** |
| Teacher collaboration | 0.15(0.02)*** | 0.13 (0.03)*** | ||
| Random | σ2e | 2.33(0.08)*** | 2.29(0.08)*** | 2.24(0.12)*** |
| σ2u0 | 0.06(0.03)* | 0.03(0.02) | 0.05(0.04) | |
| Level 2 | 11–30% low SES | 0.02(0.12) | ||
| >30% low SES | −0.03(0.23) | |||
| 0% migration | 0.02(0.28) | |||
| 11–30% migration | −0.13(0.15) | |||
| >30% migration | −0.11(0.24) | |||
| Level 1 | Student Engagement | |||
| Fixed | Mean | 11.87(0.05)*** | 11.87(0.05)*** | 11.94(0.08)*** |
| Teacher collaboration | 0.18(0.02)*** | 0.17(0.03)*** | ||
| Random | σ2e | 2.41(0.08)*** | 2.36(0.08)*** | 2.21(0.11)*** |
| σ2u0 | 0.10(0.04)** | 0.04(0.02) | 0.10(0.05)* | |
| Level 2 | 11–30% low SES | 0.23 (0.13) | ||
| >30% low SES | 0.06(0.25) | |||
| 0% migration | −0.14(0.34) | |||
| 11–30% migration | −0.44(0.17)* | |||
| >30% migration | −0.24(0.20) | |||
| Level 1 | Multicultural Classrooms | |||
| Fixed | Mean | 11.07(0.08)*** | 11.00(0.08)*** | 10.82(0.09)*** |
| Teacher collaboration | 0.18(0.04)*** | 0.15(0.04)** | ||
| Random | σ2e | 4.48(0.18)*** | 4.22(0.20)*** | 4.21(0.20)*** |
| σ2u0 | 0.21(0.10)* | 0.20(0.09) | 0.13(0.09) | |
| Level 2 | 11–30% low SES | 0.47(0.17)** | ||
| >30% low SES | 0.56(0.38) | |||
| 0% migration | 0.10(0.24) | |||
| 11–30% migration | 0.09(0.30) | |||
| >30% migration | 0.48(0.23)* | |||
| –2LL (Parameters) | 24815.41(12) | 23415.06 (22) | 15398.55(48) | |
| Δ–2LL(df)a | 34.11(4)*** | 1,400.35(10) | 8016.53(26)*** | |
| ICC Classroom Management | 2.00% | |||
| ICC Instruction | 2.80% | |||
| ICC Student Engagement | 4.00% | |||
| ICC Multicultural Classrooms | 5.70% | |||
H2 in analysis 2 The level 2 school composition variables of SES and migration background contributed significantly to model fit. For SES, compared with the reference group (teachers in schools with 1–10% of students with a low-SES background), teachers working in schools with 11–30% of students with a low-SES background scored higher on TSE in their ability to teach in multicultural classrooms. Apparently, teachers in classrooms with relatively high percentages of students from low SES backgrounds, feel they are more able to cope with the challenges of a multicultural classroom and adapt their teaching to the cultural diversity of their students. For migration background, compared with the reference group (teachers in schools with 1–10% of students with a migration background), teachers working in schools with >30% of students with a migration background had lower TSE in student engagement but higher TSE in the ability to teach in multicultural classrooms. Unlike in classrooms with high percentages of students from low SES backgrounds, teachers in classrooms with high percentages of students from migration backgrounds, feel less efficacious in, for example, raising awareness of cultural differences among students and reducing ethnic stereotyping.
(H3 in analysis 2) Adding random slopes for the relationship between the TSE scales and teacher collaboration for all units of the level 2 variables did not further improve the model, not for TSE in classroom management (σ2T3SECLS = 0.01, p = 0.554), nor for instruction (σ2 T3SEINS < 0.01, p = 0.850), student engagement (σ2 T3SEENG < 0.01, p = 0.891), or multicultural classrooms (σ2 T3SEFE = 0.03, p = 0.132). Therefore, no cross-level interactions were tested. These results imply that teacher collaboration did not moderate the relationship between the level 2 composition variables (SES and migration background) and the four TSE scales.
Discussion
The present study examined the relationship between teacher collaboration and TSE. Teacher collaboration involved teacher exchange and professional collaboration, referring to the key characteristics as described by Stoll et al. (2006): shared values and vision, collective responsibility for pupils' learning, and collaboration focused on learning groups, as well as professional learning, reflective professional inquiry, openness, networks and partnerships, inclusive membership, and mutual trust, respect, and support. TSE included the three “classical” types of TSE (TSE in classroom management, instruction, and student engagement) (see, e.g. Zee and Koomen, 2016) as well as a fourth type: TSE related to teaching in diverse classrooms (Siwatu, 2011). We also examined the extent to which the school composition variables of student SES and migration background were related to these four types of TSE, as studies have questioned the general claim that teachers in schools with large proportions of students from diverse backgrounds have lower TSE (e.g. Fackler et al., 2020).
Analyses showed positive relationships between teacher collaboration and all teacher self-efficacies (i.e. classroom management, instruction, and student engagement), with the exception of efficacy in classroom management in model 2. This latter result indicates that the level of collaboration in a diverse school is not associated with their efficacy in managing their classrooms. For this subsample, teacher collaboration was positively related to self-efficacy with regard to the ability to teach in diverse classrooms. The results suggest that collaboration involving, for example, joint teaching activities, professional development, and discussing students' learning development also stimulates teachers' self-efficacy related to the challenges that arise in multicultural classrooms. The results of previous studies (Täschner et al., 2025; Weiβenrieder et al., 2015; Lakshmanan et al., 2011; Mintzes et al., 2013) are thus similar to the results of analyses of the Dutch TALIS 2018 data.
An interesting set of relationships was observed between school composition and TSE. The results from model 1 revealed that teachers in schools with a moderate proportion of students (11–30%) with a low-SES background feel more competent in managing their classrooms and engaging students than those working in schools with a low proportion of students (1–10%) from a low-SES background. In model 2, similar results were found, and we also observed that teachers experience a higher sense of self-efficacy when teaching and connecting with students from diverse backgrounds. Previous research has suggested a negative impact of school composition on TSE, in the sense that teachers in schools with larger proportions of pupils with a low-SES background generally have a lower sense of TSE. Yet, in two recent studies, an opposite effect emerged (Fackler et al., 2020; Romijn et al., 2020), similar to the effects observed in the present study. The two explanations offered in these previous studies may also apply to our study. The first explanation is that teachers in schools serving low-SES students have possibly chosen to teach in these schools, and they may have positive attitudes toward diversity and commitment to social justice goals, as well as higher levels of TSE to begin with (Romijn et al., 2020). The second explanation may also be valid, though: given that teachers of students from disadvantaged backgrounds generally have lower expectations (see, e.g.e.g. Rubie-Davis, 2014), a small number of experiences of success may result in relatively high levels of TSE. Both explanations are speculative, and the positive relationship between school composition in terms of SES and TSE, now identified in a third study, warrants further research.
Remarkably, we see a different pattern when we examine the proportion of students with a migration background. Compared with teachers working in schools with a low proportion of students with a migration background (1–10%), teachers working in schools with a high proportion of such students (more than 30%) reported less confidence in their classroom management skills. Teachers working in schools with a moderate proportion of students with a migration background (11–30%) considered themselves less competent in engaging students than the teachers who worked in schools with a low proportion of students with a migration background. The question is why is there a difference between diversity in terms of SES and diversity in terms of migration background. Do teachers feel more resistance to addressing diversity according to migration background which, in turn, leads to lower TSE (see, e.g. Parkhouse et al., 2019)? And what is the role of backgrounds of teachers themselves? As elsewhere, the teaching staff in the Netherlands is predominantly white, female and of middleclass backgrounds (CAOP, 2024). Bradbury et al. (2022) note “the lack of minoritised teachers within an education system through low recruitment and/or through poor retention is seen as (…) a cause of low attainment by minoritised students, affected by a lack of role models” (2023, p. 335). More research is needed to explain the different patterns of relationships that have been observed and the specific challenges in classrooms that are diverse.
Within the subsample of teachers who are experienced in teaching in diverse classrooms, teachers working in schools with a moderate (11–30%) or high proportion (>30%) of students with a low-SES background reported higher confidence in their abilities to teach in a diverse classroom. It seems that, regardless of the little attention paid to multicultural topics in teacher training programmes (see also Severiens et al., 2014, Alhanachi et al., 2021), teachers learn how to handle diverse classrooms from experience. However, the higher sense of TSE in addressing diversity is not coupled with a higher sense of TSE in student engagement. The relationship is the other way around: while teachers in multicultural classrooms have high TSE related to diversity, they have low TSE in student engagement. A qualitative study in which the TSE scales in all four areas are combined with observations and interviews could shed more light on these puzzling relationships.
High TSE is related to positive student outcomes, as shown by Zee and Koomen's (2016) review. Therefore, the observed relationship between teacher collaboration and TSE justifies the recommendation to implement PLCs as a means to stimulate teacher collaboration. A moderating effect was not observed, but as TSE seems to be relatively low in schools serving students from a migration background, it can be argued that teacher collaboration is especially important in these schools. Stimulating teachers to teach jointly, observe one another's classrooms, and engage in joint activities, among other collaborative activities, may lead to higher levels of TSE and, consequently, more positive student outcomes. Aside from teachers, principals as well as school boards play a role here, as they need to make sure conditions for effective PLCs are in place. Haiyan and Allan (2020) describe conditions as structural (engaging formal structures for collaboration), cultural (collaboratively developing shared vision and goals) and relational (stimulating positive teacher-teacher relationships). Without overemphasising the importance of PLCs and teacher collaboration, these factors may play a role in breaking the negative cycle often observed in multicultural schools, where low TSE impacts negatively on student achievement and reinforces low TSE.
Aside from recommending to stimulate teacher collaboration, another recommendation can be made with regard to TSE. The recent review of Täschner et al. (2025) clearly showed positive significant effects of interventions in the area of TSE, regardless of career stages. TSE interventions that offer opportunities for mastery and vicarious experiences, including moments of reflection seem to be most effective. Teacher collaboration seems to be an important aspect of such interventions, when they, for example, provide opportunities for exchange, and mutual classroom visits.
In the context of the present study on equity and diversity, we also recommend to include diversity as the topic to collaborate on, and the area in which it is especially important to develop TSE. Aside from a focus on orientation, knowledge and skills, Parkhouse et al. (2023) emphasise the importance of tailor PD programmes to the needs of teachers, that is to the zones with regard to equity and diversity they are situated in (referring to orientations, knowledge and skills). They note that action research seems to be a suitable PD format for doing just that, as well as a format that invites teachers to take the time and reflect, both individually as well as collectively (see also Alhanachi et al., 2021).
The recommendation with regard to stimulating teacher collaboration is based on the underlying idea that teacher collaboration stimulates TSE. However, as the TALIS data are limited in the sense that they are cross-sectional, the suggested causal relationship may in fact be a reciprocal relationship and a longitudinal study design would be needed to determine how self-efficacy and collaboration impact each other. For example, Valckx et al. (2020) theorized the other way around in the sense that TSE impacts the quality of collaboration. They observed that teachers with high levels of TSE “engage more in reflective dialogue, and are therefore more open to collaborate or take mutual responsibility” (Valckx et al., 2020, p. 295). At the same, they also noted their cross-sectional data as a limitation, and make the same suggestion of longitudinal study designs. Another limitation of the study is that TALIS relies entirely on the perceptions of teachers and principals regarding the extent to which they collaborate and their self-efficacy, and these perceptions may be affected by social desirability and bias. Despite these limitations, we feel that the results of our study make a valuable contribution to the literature, given the rigorous standards applied in the TALIS study with regard to sampling method and size and the fact that it is one of the few studies to measure both SES and migration background.
Conclusion
Schools with large proportions of low-SES students more often have teachers with high self-efficacy, whereas schools with large proportions of students with a migration background more often have teachers with low self-efficacy. The relationships observed in previous studies between teacher cooperation and TSE and between TSE and positive student outcomes warrant the recommendation to stimulate teacher collaboration, especially in schools serving students from a migration background. Longitudinal research is needed to substantiate this recommendation.
Appendix TALIS 2018 scales used
Teacher collaboration.
On average, how often do you do the following in this school (on a six-point scale ranging from “never” to “once a week or more”)?
Subscale: Teacher exchange.
Exchange or develop teaching materials with colleagues
Discuss the learning development of specific students
Work with other teachers in this school to ensure common standards in evaluations for assessing student progress
Attend team conferences
Subscale: Professional collaboration.
Teach jointly as a team in the same class
Provide feedback to other teachers about their practice
Engage in joint activities across different classes and age groups (e.g. projects)
Participate in collaborative professional learning
Teacher self-efficacy.
To what extent can you do certain activities (on a four-point scale ranging from “not at all” to “a lot”)?
Subscale: Efficacy in classroom management.
Control disruptive behaviour in the classroom
Make my expectations about student behaviour clear
Get students to follow classroom rules
Calm a student who is disruptive or noisy
Subscale: Efficacy in instruction.
Craft good questions for my students
Use a variety of assessment strategies
Provide an alternative explanation, for example when students are confused
Vary alternative instructional strategies in my classroom
Subscale: Efficacy in student engagement.
Get students to believe they can do well in schoolwork
Help my students to value learning
Motivate students who show low interest in schoolwork
Help students think critically
Subscale: Efficacy in multicultural classrooms.
Cope with the challenges of a multicultural classroom
Adapt my teaching to the cultural diversity of students
Ensure that students with and without a migrant background work together
Raise awareness of cultural differences amongst students
Reduce ethnic stereotyping amongst students
