2. Do students have equitable access to effective teachers and learning environments?

Teacher quality is by far the most important benefit schools can provide to students. A large body of research literature shows that teachers have a powerful impact on students’ outcomes (Aaronson, Barrow and Sander, 2007[1]; Rivkin, Hanushek and Kain, 2005[2]). It is long-lasting (Chetty et al., 2011[3]), and it is not limited to academic achievement or other cognitive outcomes. There is now robust evidence that teachers can also raise students’ social and emotional skills (Blazar and Kraft, 2017[4]; Jackson, 2018[5]).

There is less consensus about what exactly makes a “good” teacher, however. Many researchers have struggled to identify teachers’ characteristics and teaching practices that are robustly correlated with students’ outcomes (Rivkin, Hanushek and Kain, 2005[2]). This is partly due to the fact that teaching is a complex and multidimensional activity. Different teachers are often good at different things, and measured teaching effectiveness can be influenced by contextual factors that are outside of teachers’ influence: the quality of the “match” between the teacher and the school (OECD, 2012[6]) can be important as well as the match between teachers’ and students’ socio-demographic characteristics (Dee, 2005[7]; Fairlie, Hoffmann and Oreopoulos, 2014[8]; Gershenson et al., 2018[9]; Gershenson, Holt and Papageorge, 2016[10]; Lim and Meer, 2017[11]). This implies that teachers can be more or less effective depending on the environment they operate in and the students they are assigned to.

Similarly, different teaching styles and practices can be especially beneficial for some students (more than for others), and this makes it difficult to identify teaching practices that can be considered “superior” to others. Le Donné, Fraser and Bousquet (2016[12]) find that cognitive activation strategies are less strongly related to mathematics achievement in schools where a larger share of students comes from a socio-economically disadvantaged background. Caro, Lenkeit and Kyriakides (2016[13]) similarly conclude that the association between cognitive activation and performance is stronger for socio-economically advantaged students and in schools with a positive disciplinary climate. Direct, teacher-centred instruction is also believed to be superior for disadvantaged, at-risk students (Butler, 2020[14]), and student-centred instruction can, therefore, amplify socio-economic gaps in achievement (Clark, Kirschner and Sweller, 2012[15]; Kirschner, Sweller and Clark, 2006[16])

A consensus is, however, slowly growing on what constitutes “quality” or “effective” teaching (OECD, 2020[17]) thanks to a large set of studies that have tried to unpack the relationship between teaching and learning in order to identify teachers’ attributes and teaching practices that are more likely to facilitate the cognitive and the socio-emotional development of students This large and growing body of literature informs the conceptual framework of surveys like the Teaching and Learning International Survey (TALIS), and as a consequence the questionnaires administered as part of the survey. Different frameworks give more or less emphasis to different aspects, and sometimes use different terminologies to refer to similar concepts. The TALIS framework emphasises a number of practices related to instructional quality that have received much attention in the literature (Ainley and Carstens, 2018[18]). Good teaching requires a well-managed classroom in which disruptions are minimised and learning time is maximised. Good teachers must be able to communicate in a clear and comprehensive way; they should help students gain a deep understanding of the subject by requiring them to evaluate, integrate and apply knowledge to solve problems; they should be able to provide effective support to students, listening to their needs, respecting their ideas, and encouraging them; they should provide constructive feedback through both formative and summative assessments. Effective teachers should also, of course, be competent professionals: they should possess and continue to develop appropriate content and pedagogical knowledge as well as affective and motivational competencies, and this knowledge should inform their teaching practices (Guerriero, 2017[19]).

This chapter draws on data from TALIS 2018. It examines how teachers’ characteristics and practices that the research literature has shown to be robustly correlated with students’ achievement are distributed across schools.

Given that the analysis aims at informing policies about the allocation and relocation of teachers in order to achieve more equitable outcomes for students, the distinction between teachers’ characteristics and practices is particularly relevant. The former are in a sense fixed and portable assets that teachers always possess irrespective of the schools where they are employed. Practices, on the contrary, are the result of an explicit choice made by teachers who are teaching in a given context (and as such they are elicited by means of the TALIS questionnaire). Nothing ensures that they would adopt the same practices in a different school or even with different students in the same school. Indeed, it can be argued that the ability of a teacher to adapt their methods of instructions to the specific learning needs of their students is an important element of quality teaching, especially for educational systems that value inclusiveness (Brussino, 2021[20]; OECD, 2012[6]; Peterson et al., 2018[21]).

The analysis will look at inequalities in students’ access to effective teachers from two different angles and using different tools. A first approach aims at assessing whether teachers with certain traits are clustered in a restricted number of schools. Clustering arises whenever similar individuals (in this case, teachers with similar characteristics) end up together (in this case, working in the same school). Clustering can be the result of teachers’ behaviours (similar teachers are more likely to apply to the same school) as well as of schools’ behaviour (when a school tends to only hire teachers sharing a narrow set of characteristics). This report will not be able to disentangle which of these mechanisms is at play.

Clustering will be measured by the dissimilarity index, which capture to what extent the distribution of teachers belonging to different groups depart from what would be observed if teachers were allocated across schools in a perfectly random way. The index (commonly used as a measure of segregation) is related to the proportions of teachers of either one of two groups that have to be displaced in order to achieve a perfectly even distribution, i.e. a situation where the shares of teachers of different types in each school equals the shares observed in the overall population. It ranges from 0 to 1, where 0 represents the situation of perfect evenness and 1 the situation of maximum unevenness, in which all teachers of one type are concentrated in a single school.1 More details on the dissimilarity index are contained in Box 2.1.

Analysis based on the dissimilarity index addresses issues related to equality in a broad sense: by being only focused on characteristics of teachers, it disregards characteristics of the students as well as the fact that students themselves sort across schools on the basis of their personal characteristics (OECD, 2019[22]). The dissimilarity index tells us the likelihood that teachers with certain characteristics are clustered together in the same school. This information can be useful for policy makers to better understand the process of teachers’ sorting in their country.

Randomly assigning teachers to schools, however, might not help in addressing concerns related to equity. In equitable education systems, the achievement of educational potential does not depend on personal and social circumstances, including factors such as gender, ethnic origin, immigrant status, special education needs and giftedness (OECD, 2017[23]; OECD, 2012[6]). To achieve this goal, it may be necessary to devote more resources (including more effective teachers) to disadvantaged students who, for different reasons, do not compete on an equal footing with their more advantaged peers. This is the distinction between horizontal and vertical equity (OECD, 2017[24]), where the former considers the overall fair provision of resources to each part of the school system (providing similar resources to the alike) and the latter involves providing disadvantaged groups of students or schools with additional resources based on their needs.

Issues related to equity are more directly assessed by looking at whether “better” teachers (a shortcut here for teachers’ characteristics and teaching practices robustly associated with higher student proficiency) are more or less likely to teach disadvantaged students. Unfortunately, TALIS contains little information about the characteristics of all the students that each surveyed teacher teaches to. The only information that can be exploited is at the school level. In particular, the chapter will focus on three different variables that capture important features of the population of students served by the teachers.2

The first variable is the socio-economic composition of the student body. Schools where more than 30% of students come from socio-economically disadvantaged homes3 are classified as “disadvantaged schools”, and schools where less than 10% of the students are socio-economically disadvantaged are classified as “advantaged schools”.

The second variable refers to the location of the school, distinguishing between rural and urban areas. Schools located in rural areas often cater to students with particular socio-economic profiles and may face a distinctive set of challenges (Echazarra and Radinger, 2019[25]); urban and rural schools can also differ in their ability to attract and retain teachers.

The third variable relates to school governance, distinguishing between public and private schools.4 The public or private management of schools is an important factor in many countries in explaining the segregation of students according to their socio-economic background (OECD, 2019[22]). It is important, however, to acknowledge that the relevance of urban-rural or public-private divides can vary across countries and national contexts. (The share of teachers and schools by each school type are presented in Tables A.B.2 and A.B.3 in Annex C.)

The implicit assumption underlying this analysis is that all students in a given school are equally “exposed” to all the teachers in the school (or, equivalently, that students are randomly sorted into classes). The validity of this assumption varies across countries depending on the particular institutional arrangements governing class formation, the assignment of teachers to classes, and whether such arrangements change from grade to grade.

It is also important to acknowledge that the analysis is purely descriptive and focuses on a single characteristic of the school (be it student composition, location, or type of governance). Schools can differ in many potentially important ways that explain the differences in the characteristics and practices of teachers working in a certain school. For instance, rural schools tend to be smaller than urban schools. School size can be, in itself, a factor driving teachers’ application decisions. The results of the analysis should not then be interpreted in a causal sense (as school characteristics determining the prevalence of certain teachers), and should be complemented with country-specific information about the specific structure of the education system.

After having investigated the allocation of teachers’ characteristics and practices, the chapter will look at the prevalence of effective learning environments in different types of schools. Schools are effective learning environments when all stakeholders – students, parents, teachers, principals – co-operate and complement each other, and interact in a way that delivers learning outcomes superior to what would be expected by simply looking at the sum of each individuals’ contributions. The analysis in this chapter will not cover all elements that contribute to creating effective learning environments. Rather, it will narrowly focus on two indicators that capture the quality of the leadership exerted by principals: this extends the analysis conducted on the allocation of teachers across different schools to principals. Two indicators will be examined: the degree to which principals engage in instructional leadership and whether all teachers have access to mentoring. These are arguably two indicators over which principals have significant margin of manoeuvre, and they can, therefore, be considered as reasonable proxies of principals’ quality (intended as principals’ ability to create effective learning environments).

Finally, the focus of the analysis will shift to how teachers’ characteristics and practices are related and how this relationship varies with school characteristics. This will be particularly informative from a policy perspective. While the first part of the chapter analyses different dimensions of teaching quality in isolation, in practice, teaching is performed by people who possess a bundle of these characteristics and choose to adopt certain practices. Policies trying to achieve better outcomes for students by modifying the allocation of teachers must necessarily take these interrelationships into account.

The finding that more experienced teachers are on average more effective in raising the performance of their students is probably one of the most robust and most widely accepted in the literature (Papay and Kraft, 2015[29]). Studies generally find that effectiveness increases steeply in the first few years of teachers’ careers, and remain constant afterwards. Evidence is now emerging, however, that teachers can keep improving much later in their career (Wiswall, 2013[30]), and that whether or not they do so might well depend on whether they have the opportunity to work in a supportive professional environment (Kraft and Papay, 2014[31]).

In the following analysis, teachers with more than ten years of professional experience are labelled as “experienced teachers” as opposed to “less experienced teachers” who have ten years or less of teaching experience. For most countries, the share of experienced teachers in the overall teacher population ranges between 50 and 70% (Figure 2.2).

Portugal, Saudi Arabia and Turkey are the three TALIS countries where experienced teachers are more likely to be concentrated in a small number of schools (Figure 2.1) with values of the dissimilarity index larger than 0.5 (as compared to an OECD average of about 0.3). In the case of Portugal, though, this result must be taken with care in light of the fact that more than 90% of teachers in Portugal are classified as “experienced”. In such cases, looking at allocation across schools may not be particularly meaningful as, inevitably, most students will have access to some experienced teachers (see the footnote of Figure 2.1).

The dissimilarity index for experienced teachers shows much lower values than the OECD average in England (United Kingdom), Finland and Singapore, signalling that the share of experienced teachers in a given school is likely to be similar to the share in the overall population in these countries (Figure 2.1).

An alternative way to look at this issue, which overcomes difficulties related to the interpretation and cross-country comparability of the dissimilarity index, relies on examining the proportion of the overall variance in teacher experience that lies between schools (as opposed to the variance within schools). This method works well in the case of continuous variables such as teachers’ experience but is not without drawbacks. In particular, this indicator is more sensitive, for instance, to differences between schools where teachers have on average 20 years of experience and schools where teachers have on average 30 years of experience. But if teaching effectiveness increases with experience only up to 10 years, the difference between the two sample schools would, in fact, not matter.

Nevertheless, it is reassuring that the two indicators deliver similar results. While the proportion of variance in teachers’ experience that lies between schools is 8.2% on average across OECD countries participating in TALIS, this indicator is as high as 37% in Saudi Arabia, 33% in Turkey and 20% or more in Colombia and Portugal, which are the same countries with high values of the dissimilarity index (Table 2.4).

In many of the countries participating in TALIS, experienced teachers are more likely to work in schools with a low concentration of socio-economically disadvantaged students (less than 10% of the student body) than in schools where disadvantaged students constitutes more than 30% of the student body (Figure 2.2). The differences between these two types of schools in the share of experienced teachers are particularly large in Australia, Estonia, the Flemish Community of Belgium and Saudi Arabia, where they range between 13 and 18 percentage points. In a few other countries, however, experienced teachers are more likely to work in schools with a large share of disadvantaged students. This is, in particular, the case of Colombia (with a difference of 19 percentage points), Shanghai (China) (14 percentage points) and Israel (12 percentage points).

In many countries, the differences between private and public schools are also particularly large. More often than not, experienced teachers are over-represented in public schools (Figure 2.2). Colombia, the United Arab Emirates and Viet Nam are the countries where this phenomenon is more evident. Australia, Korea, New Zealand and Singapore are among the few countries, on the other hand, where experienced teachers are more likely to work in private schools.

In most countries there are no significant differences between schools located in rural settings (towns with, at most, 3 000 inhabitants) and schools located in urban areas (cities with more than 100 000 inhabitants) (Figure 2.2). Where differences along these dimensions exist, they tend to be quite large in both directions. In Turkey, for instance, the differences in the share of experienced teachers between urban and rural schools is as high as 34 percentage points, meaning that experienced teachers are largely concentrated in urban schools. Divides of similar magnitudes are observed in Romania and Saudi Arabia. Austria, Norway, the United Arab Emirates and the United States are examples of countries in which experienced teachers are over-represented in rural schools.

A unique feature of TALIS is the amount of detailed information it contains about the content of initial teacher education programmes (OECD, 2019[32]). This allows this report to go beyond studies that have traditionally focused on the simple effect of teachers’ licensing. As the content and requirements of teachers’ education vary greatly both across countries and within countries over time, it is particularly important to have more precise information on the kind of initial education received by teachers.

Arguably, the type and quality of teacher education are important determinants of teacher knowledge, which in turn is significantly related to student achievement (Baumert et al., 2010[33]; Hill, Rowan and Ball, 2005[34]). For the purpose of the analysis presented in this report, it is useful to combine the rich information on initial teacher education collected in TALIS in a simple binary indicator of “comprehensive education”, which identifies teachers whose initial formal education and training covered a broad set of topics, including content, pedagogy, classroom practice, cross-curricular skills, teaching in mixed ability setting and classroom management.

On average across OECD countries participating in TALIS, about 40% of teachers received a comprehensive initial education (Figure 2.4). For most countries, this share ranges between 30 and 60%. There is a much smaller proportion of teachers with comprehensive initial education in the Czech Republic (17%) and Spain (17%) whereas they constitute a large majority in the United Arab Emirates (73%) and Viet Nam (86%).

Variations across countries in the dissimilarity index (which captures the degree to which teachers with a comprehensive initial education are unevenly distributed across schools) are limited (Figure 2.3). The dissimilarity index varies in the vast majority of countries between 0.2 and 0.3. It is relatively small in Malta (0.16) and relatively large in Iceland and Viet Nam (above 0.35). The result for Viet Nam should, however, be interpreted with care, given the very large share of teachers in the country with comprehensive initial education.

Similarly, in most countries there are no large differences between different types of schools in the share of teachers who received a comprehensive education (Figure 2.4). Differences between private and public schools are slightly more pronounced in a number of countries. Public schools generally are more likely to employ comprehensively educated teachers, especially in countries such as Japan, Kazakhstan, Sweden and Viet Nam. In Belgium (including its Flemish Community) and Denmark, on the other hand, these teachers are more likely to work in private schools.

In only a handful of countries are there differences in the share of comprehensively educated teachers between rural and urban schools (Figure 2.4). Where these differences exist, they are in favour of rural schools: this is the case for Brazil, Croatia and Romania, in particular.

In seven countries, comprehensively educated teachers are more likely to work in schools with a high concentration of socio-economically disadvantaged students (Figure 2.4). This is the case in particular of Ciudad Autónoma de Buenos Aires (hereafter CABA [Argentina]), Israel and Italy. These teachers are overly represented in advantaged schools in Belgium (including its Flemish Community) and Spain.

Self-efficacy captures individuals’ perceptions of their capabilities of performing a task. Such perceptions can influence actual behaviours and, thus, performance. A vast literature in education has showed robust positive association between self-efficacy and performance. This is true for students, where self-efficacy correlates with academic performance (Honicke and Broadbent, 2016[35]), as well as for teachers, with higher self-efficacy being associated with higher quality instructional practices (Holzberger, Philipp and Kunter, 2013[36]) and better student outcomes (Caprara et al., 2006[37]; Woolfolk Hoy and Davis, 2006[38]).

Contrary to self-confidence, self-efficacy is conceptually related to specific tasks. This is why TALIS elicits teachers’ self-efficacy in a number of dimensions: classroom management, instruction, and student engagement. An overall self-efficacy scale can be computed as an average of these different dimensions (OECD, 2019[39]).

Within this report, overall self-efficacy is considered as a “fixed” characteristic of teachers such as years of experience or the content of initial education. It must be acknowledged, though, that self-efficacy is definitely more prone to change over time. It is often conceptualised as being context-specific (Tschannen-Moran and Hoy, 2001[40]), in which case relocating teachers to different schools is no guarantee that the teacher would keep the same level of self-efficacy in the new school.

Teachers with high self-efficacy are identified as those in the top quarter of the overall self-efficacy scale in their own country. Doing so has two great advantages for analysis. First, the analysis does not suffer from possible lack of measurement invariance,5 making the results more comparable across countries. Second, the analysis based on the dissimilarity index is more robust because by construction the share of high-self-efficacy teachers is not negligible (being equal to 25%) and is the same in all countries.

Figure 2.5 shows that there is a fair degree of heterogeneity across countries in the extent to which teachers with high self-efficacy are allocated across schools. The largest degree of clustering is observed in Alberta (Canada), Belgium, New Zealand and South Africa, where the dissimilarity index is between 0.37 and 0.40.

In spite of the imbalances signalled by the dissimilarity index, there are actually few differences between schools of different types in terms of the share of teachers with high self-efficacy (Figure 2.5). In Finland and Singapore, private schools are more likely to employ teachers with high self-efficacy while the opposite is true in CABA (Argentina) and Norway. A more consistent divide is between urban and rural schools: in Australia, Estonia, Finland, France, Italy, Lithuania and Sweden, high self-efficacy teachers are significantly more likely to work in urban schools rather than in rural schools. The only country where the opposite is true is Chile.

In only three countries are differences between advantaged and disadvantaged schools observed (Figure 2.5). Teachers with high self-efficacy are more likely to work in disadvantaged schools in South Africa, and more likely to work in advantaged schools in Belgium and Spain.

TALIS 2018 collected information on a number of instructional practices that the research literature has identified as being important elements of teaching quality. This chapter will focus in particular on three practices for which the empirical literature has been able to establish the most robust relationship with students’ achievement: cognitive activation, clarity of instruction, and time spent on actual teaching.

In order to mitigate concerns about social desirability bias, rather than using an agreement scale (i.e. asking teachers how much they agree with a given activity being important or desirable), TALIS relies on frequency scales, asking teachers how often they perform certain activities that characterise a given practice. Combining answers to different items allows the construction of a scale that measures the degree to which teachers adopt a given practice. This is how the scales of cognitive activation and clarity of instruction were constructed (OECD, 2019[39]).

Time spent on actual teaching is computed by directly asking teachers about the share of class time they typically spend on actual teaching as opposed to administrative tasks and keeping order in the classroom.

For all three practices, the analysis focuses on teachers in the top quarter of the national distribution of the measure. As was the case for self-efficacy, this maximises the extent to which the dissimilarity index can be compared across countries.

Cognitive activation consists of instructional activities that require students to evaluate, integrate and apply knowledge within the context of problem solving (Ainley and Carstens, 2018[18]; Lipowsky et al., 2009[41]). The use of cognitive activation has been shown to be related to higher student achievement. This is the case for mathematics in the Programme for International Student Assessment (PISA) (Le Donné, Fraser and Bousquet, 2016[12]) although the relationship seems to be less strong in schools with a high share of disadvantaged students. On the other hand, using TIMSS data, Bellens et al. (2019[42]) conclude that instructional quality (of which cognitive activation is an element) has similar positive effects on the achievement of all students. Cognitive activation strategies have also been found to moderate the positive effect of pedagogical content knowledge on students’ achievement (Baumert et al., 2010[33]).

Among TALIS participants, South Africa is the country where teachers relying more on cognitive activation strategies are most unevenly distributed across schools (dissimilarity index: 0.41), followed by Brazil, Chile, New Zealand, Saudi Arabia and Viet Nam (dissimilarity index: 0.37). The allocation is much more even in France, Malta, Portugal and Singapore (i.e. dissimilarity index between 0.19 and 0.24) (Figure 2.6).

However, such imbalances do not necessarily mean there is uneven student access to teachers who effectively use cognitive activation strategies. In most countries, the differences between advantaged and disadvantaged schools in this regard are not statistically significant. Yet, there are four countries – Austria, Israel, Lithuania and Portugal – where the share of teachers who frequently rely on cognitive activation is higher in socio-economically advantaged schools than in disadvantaged schools (Figure 2.6).

More frequently, divides emerge between public and private schools. Cognitive activation practices are more common in private schools in six TALIS participating countries and territories, and the differences are largest in Finland (21 percentage points), Singapore (14 percentage points) and the Czech Republic (11 percentage points). Chile and Kazakhstan are the only two countries where the divide is in favour of public schools.

In only six countries are some differences between urban and rural schools observed. In Australia, Estonia, Lithuania, Norway and the United Arab Emirates, cognitive activation practices are more likely to be used in urban schools while in Turkey the opposite is true.

Clarity of instruction is conceptualised in TALIS as the ability to set clear and comprehensive instruction and learning goals, to connect new and old topics, and to provide students with a summary of the lesson at its end (Ainley and Carstens, 2018[18]). Various studies have shown how this practice is related to positive student outcomes, including learning motivation, achievement and satisfaction (Hines, Cruickshank and Kennedy, 1985[43]; Seidel, Rimmele and Prenzel, 2005[44]). In TIMSS, students that reported higher scores for their teachers on this dimension tended to have better performance in mathematics and science (Mullis et al., 2020[45]). In Blazar (2015[46]), (lack of) clarity is part of a broader construct of mathematical error and imprecisions, which is found to be negatively correlated to students’ achievement.

As was the case for cognitive activation practices, there is evidence of an uneven allocation of teachers who frequently adopt clarity of instruction techniques across schools but there is less evidence of teachers’ sorting in schools with particular characteristics (Figure 2.7). The largest departure from a random allocation of teachers occurs in Alberta (Canada) and Chile, with dissimilarity index values above 0.4. Smaller values (still above 0.2) are recorded in France, the French Community of Belgium and Portugal.

In 12 countries and territories, teachers who rely most on clarity of instructions tend to be concentrated in public schools (Figure 2.7). The difference with respect to private schools is largest in Italy (15 percentage points), Australia (13 percentage points) and the United States (13 percentage points). Finland and Singapore are the only countries where clarity of instruction is more frequently adopted in private schools.

Fewer differences emerge according to school location. Teachers tend to adopt clarity of instruction more frequently in urban schools in seven countries, in particular in Alberta (Canada), Finland and the United States. Yet, in Hungary, Romania and South Africa, this practice is more common in rural than urban schools.

Differences according to the socio-economic composition of the student body are present in only two countries (Australia and Chile); in both cases, they are to the benefit of disadvantaged schools, whose teachers are more likely to often use clarity of instruction techniques.

More instruction time during class translates into higher student achievement (Carroll, 1963[47]; Muijs et al., 2014[48]; Schmidt, Zoido and Cogan, 2014[49]). This has been shown to hold across different settings, using different data and different empirical strategies. In PISA, differences in weekly instructional time can account for cross-country differences in students’ achievement (Lavy, 2015[50]). The strength of this relationship varies across countries and depends on the classroom environment (Rivkin and Schiman, 2015[51]). In Germany, shortening the length of the school year had a negative effect on educational achievement, increasing grade repetition and resulting in fewer students attending higher secondary school tracks (Pischke, 2007[52]). In Denmark, a large-scale randomised trial showed that increases in instruction time lead to higher student learning (Andersen, Humlum and Nandrup, 2016[53]).

The literature on teaching quality has stressed the ability of teachers to maximise instruction time as one important component of classroom management (Ainley and Carstens, 2018[18]; Kane et al., 2010[54]; Stronge et al., 2007[55]).

In TALIS, classroom management is captured by a question on the disciplinary climate observed in the classroom. A different question, though, gets closer to identifying the instruction time to which students are exposed: it asks teachers how class time is allocated between different tasks such as administrative tasks, keeping order and actual teaching. Data from the TALIS-PISA link study show that students of teachers who spend a higher share of class time on actual teaching perform better in the PISA assessment (OECD, 2021[56]).

Data from TALIS 2018 show that teachers who are in the top quarter of the national distribution in terms of the share of class time they spend on actual teaching are far from being equally represented across schools. On average across OECD countries that participated in TALIS, the dissimilarity index equals 0.33 (Figure 2.8). In Georgia, South Africa and Viet Nam, the distribution is particularly uneven, with the index exceeding 0.45. In Lithuania, Portugal and Shanghai (China) the value of the index is smallest, although still above 0.25.

Large and systematic differences are observed between different types of schools in the majority of countries. Teachers that spend more class time on actual teaching are much more likely to work in advantaged schools as well as private schools (Figure 2.8). As for the former, the divides are particularly large (above 20 percentage points) in Alberta (Canada), Denmark, and New Zealand. Shanghai (China) is the only territory in which disadvantaged schools are more likely to employ teachers in the top quarter of the distribution of the share of class time spent teaching. Differences between private and public schools are largest in Singapore (32 percentage points), Kazakhstan (17 percentage points), Australia (16 percentage points), Denmark (15 percentage points) and New Zealand (15 percentage points). The only countries in which public schools are more likely to employ teachers who spend a high share of class time in actual teaching than private schools are Italy and Japan.

Differences according to school location are less common. In nine countries, rural schools are more likely to employ teachers who spend a large share of class time in actual teaching, with differences particularly large (20 percentage points or above) in Colombia and Spain. Differences are in favour of urban schools in Australia, Hungary, Kazakhstan and Lithuania.

While teachers are undoubtedly the most important school-related factor that contributes to students’ achievement, many other things can help improve learning, often by facilitating, improving and complementing the work of teachers. Much research has emphasised the importance of creating “effective learning environments” (Ainley and Carstens, 2018[18]). In such environments, the overall results are often larger than what the sum of individual components would deliver elsewhere. Many different actors can play a role in helping create such environments, including students themselves, parents, and teachers, but school principals play a crucial role in this respect. One could argue that a school principal’s most important job is to create an effective learning environment for teachers and students to work together. Principals are, therefore, ultimately important for student outcomes, although the relationship is likely to be indirect, operating through teachers or an overall school climate that is more conducive to learning (Hallinger, 2011[57]). Recent literature in economics has also highlighted the importance of managerial skills for principals (Bloom et al., 2015[58]; Di Liberto, Schivardi and Sulis, 2015[59]).

TALIS 2018 surveyed both teachers and principals, collecting a wide range of indicators describing the institutional environment and conditions under which teachers operate. This includes, among others, practices related to teachers’ professional development and information about the characteristics of the principals and the type of tasks they are more engaged with. As the broad goal of this chapter is to investigate to what extent students have equal and fair access to effective learning environments in their specific school, the following analysis will focus on two indicators that can vary across schools and that can be plausibly shaped and influenced by school principals: principals engaging in instructional leadership and the availability of mentoring programmes for teachers in the school.

The survey was designed in such a way that these indicators are only available at the school level. As there is only one principal in any given school, it is not possible to examine their distribution across schools by means of indicators such as the dissimilarity index. In the following sections, only differences between different types of schools will be reported.

Instructional leadership refers to the actions that a principal takes to promote growth in students’ learning (Flath, 1989[60]). Concrete actions that principals can take to improve the quality of instructions and, therefore, students’ learning include managing the curriculum, attending to teachers’ professional development needs and creating a culture of collaboration. In TALIS, instructional leadership is measured by asking principals how much they support teachers co-operating to develop new practices; and how much they ensure that teachers feel responsible for their students’ outcomes and take responsibility for improving their teaching skills. Indeed, previous findings from TALIS show a positive association between instructional leadership and collaboration among teachers (OECD, 2016[61]), which is in turn beneficial for students’ learning (Goddard et al., 2015[62]).

The degree to which principals are actually able to implement these actions can vary significantly from country to country as the boundaries of school autonomy are often regulated at the national level. For this reason, the analysis will focus on principals that are in the top quarter of the national distribution of the index of instructional leadership, i.e. principals who show a higher tendency to adopt instructional leadership actions when facing a similar set of institutional rules.

Very few statistically significant differences between different types of school emerge when looking at the likelihood of employing a principal who scores in the top quarter of the national distribution in terms of instructional leadership (Figure 2.9). This is in part due to the large margin of error associated with estimated differences: the number of schools surveyed in TALIS is relatively limited, rarely exceeding 200. In Colombia and Lithuania this calibre of principals are more likely to work in socio-economically advantaged schools but the opposite is true in France and in Mexico. The difference between private and public schools is positive in Colombia but negative in France and Norway (meaning that principals who score high on the instructional leadership scale are more likely to work in public schools in these two countries). Slightly more consistent results emerge when looking at geographical factors, with urban schools more likely to have principals that adopt an instructional leadership style in Denmark, Estonia, England (United Kingdom) and Lithuania.

Mentoring and induction programmes are an important element of teachers’ continuing professional development. They can be especially important for novice teachers, who are generally less effective than more experienced ones.

The literature on the effects of mentoring programmes is relatively new (Ingersoll and Strong, 2011[63]; Jackson, Rockoff and Staiger, 2014[64]). More recent and rigorous studies generally find positive effects on students’ achievement (Glazerman et al., 2010[65]; Rockoff, 2008[66]). Interestingly, practices that can be considered similar in spirit to mentoring programmes and that are not explicitly targeted to novice teachers such as protocols for teachers’ peer observations have been found to be beneficial not only for students of the observed teacher but also for students of the observee (Burgess, Rawal and Taylor, 2021[67]).

Initiatives that favour teachers’ collaboration and mutual help in professional development can be relatively simple, and principals can put them in place. Despite this, they are still relatively uncommon in countries that participated in TALIS 2018 (OECD, 2019[32]).

Except for a few countries, teachers’ access to mentoring programmes is similar across advantaged and disadvantaged schools. However, in Austria, CABA (Argentina), Lithuania and Romania, the share of principals who reported that all teachers in their school had access to a mentoring programme is higher in socio-economically advantaged schools (Figure 2.10). The United States is a notable exception as mentoring programmes are more prevalent in schools with more disadvantaged students.

In most countries, mentoring is likely to be equally present in urban and rural schools. Exceptions are Alberta (Canada) and Latvia, where access to mentoring programmes is higher in urban schools (Figure 2.10).

Evidence is mixed on the difference in teachers’ access to mentoring in public and private schools (Figure 2.10). In Brazil, Colombia and Mexico, there is more likely to be mentoring for all teachers in private schools but the opposite holds true in Estonia, France and the United Arab Emirates.

So far the analysis has focused on a single teacher characteristic or teaching practice at a time. However, teachers have a wide range of characteristics and practices they can employ. In order to inform teacher allocation policies, it is important to understand how these different characteristics and practices are related to each other.

As it is impossible to analyse all the dimensions involved, the focus will be on the relationship between two important teaching practices (cognitive activation and class time spent on actual teaching) and teachers’ characteristics. This will be informative about possible changes in teaching practices in a school when teachers with certain characteristics have been relocated to that school.

The analysis will also investigate whether the relationship between teachers’ characteristics and teaching practices varies depending on schools’ characteristics. This tells us about the extent to which different teachers will adapt or change their practices depending on the environment they operate in.

The analysis relies on multilevel regression models. These models can estimate the relationship between a given practice and teachers’ characteristics, and how that varies with school characteristics, while at the same time taking into account the nested structure of the data (i.e. the fact that teachers are clustered within particular schools). Schools and teachers, therefore, form the two levels of the model. Adding an interaction term between teachers’ characteristics (such as teachers’ years of experience) and school characteristics (such as the share of disadvantaged students) is informative on how the strength of the relationship between a given practice and a given characteristic (captured by the coefficient associated to teachers’ characteristics) changes when looking only at schools with those given characteristics (schools with a high share of disadvantaged students in all the analyses presented in this section). A negative sign for the coefficient associated to the interaction terms signals, for instance, that the strength of the relationship between teaching practice and teachers’ characteristics is weaker in disadvantaged schools (and the other way around for a positive coefficient).

Multilevel models not only deliver estimates of regression coefficient (how much the odds of teachers adopting a certain practices varies with teachers’ characteristics) they also provide estimates of how much the overall variance in practices varies between and within schools. Looking at how the between-school variance changes after adding teachers’ characteristics to the model is informative about the possible outcomes of teachers’ reallocation as it tells how much difference between schools one can expect after accounting for the fact that different schools employ teachers with different characteristics.

Table 2.1 summarises the relationship between the use of cognitive activation and the three teacher characteristics analysed previously. It also shows how the relationship varies depending on the socio-economic background of the students they teach. In essentially all countries that participated in TALIS, teachers with a higher sense of self-efficacy are more likely to adopt cognitive activation practices. On average across OECD countries, however, this relationship is weaker in disadvantaged schools. Chile and Korea are two exceptions in the sense that the likelihood that teachers with high self-efficacy adopt cognitive activation strategies is actually higher in disadvantaged schools.

Almost equally robust is the positive relationship between cognitive activation practices and having received a comprehensive initial teacher education. The relationship is positive in most countries, and negative only in the French Community of Belgium (Table 2.1). In a few countries (Czech Republic, Israel, Malta, Slovenia, Spain) the strength of this relationship is even stronger in disadvantaged schools.

More mixed results emerge when looking at the relationship between teachers’ years of professional experience and the use of cognitive activation. On average across the OECD countries that participated in TALIS 2018 the relationship is negative, meaning that more experienced teachers are less likely to rely on cognitive activation (Table 2.1). However, the relationship is actually positive in many countries like Bulgaria, Hungary, Kazakhstan, Lithuania, the Netherlands, the Russian Federation and the Slovak Republic. In the majority of countries, the strength of this relationship does not seem to depend on the share of disadvantaged students in the school.

Do differences in teachers’ characteristics account for the variation between schools in terms of the use of cognitive activation practices? An answer to this question is contained in Table 2.19. The table tells us first that the share of variation in the use of cognitive activation that lies between schools is small on average across OECD countries (2.7%), although there are some differences across countries. The share of variance between schools is not identical to the dissimilarity index but it does measure similar concepts. It is, therefore, not surprising that the country with the highest dissimilarity index value (South Africa) is also the country with the highest share of between-school variance at 20%. Similarly, the share of between-school variance is among the lowest in France (0.2%) and Slovenia (0.6%), which are also countries with a low level of the dissimilarity index (Tables 2.8 and 2.19).

Balancing the composition of teachers across schools would greatly reduce between-school variation in the use of cognitive activation. In models that control for teacher characteristics, variation between schools (as captured by the between-schools standard deviation) drops on average across OECD countries by 21% compared to an empty model that does not include those variables (Table 2.19). With very few exceptions, the decline is consistent in most of the countries that participated in TALIS.

The relationships between the share of class time spent on actual teaching and teachers’ characteristics are summarised in Table 2.2. A strong and consistently positive relationship between time spent on teaching and teachers’ experience and sense of self-efficacy emerges in almost all the countries participating in TALIS.

The strength of the relationship with self-efficacy appears stronger in more disadvantaged schools on average across OECD countries and especially in some countries like Estonia, Hungary, Latvia, Malta and the Slovak Republic (Table 2.2). This is less the case for teachers’ experience: if anything, there is evidence that the relationship between experience and time spent teaching is actually weaker in more disadvantaged schools in a few countries.

Having received a comprehensive initial education does not appear, on the other hand, to be robustly related to the amount of time spent on actual teaching.

Differences between schools in the share of class time spent on actual teaching are larger than differences in the use of cognitive activation: the share of variance between schools is on average 8% across OECD countries and exceeds 15% in Australia, Brazil, Bulgaria, Georgia, and the United States – these are all countries that also have a dissimilarity index level that is higher than the average (Tables 2.12 and 2.27).

Controlling for teachers’ characteristics, variation between schools would decrease, but not by a large amount: on average 7% across OECD countries (Table 2.27). The reduction would be highest (above 14%) in Australia, Belgium, Croatia, England (United Kingdom), Latvia and Mexico. However, there are a few countries (Alberta (Canada), Italy and Portugal) that would observe an increase in between-school variance following a relocation of teachers according to those characteristics. This means that, under the current allocation in these countries, teacher characteristics act as a counterbalance to school characteristics. This reduces the difference between schools in the average share of class time teachers spend on actual teaching.


[1] Aaronson, D., L. Barrow and W. Sander (2007), “Teachers and student achievement in the Chicago public high schools”, Journal of Labor Economics, Vol. 25/1, pp. 95-135, https://doi.org/10.1086/508733.

[18] Ainley, J. and R. Carstens (2018), “Teaching and Learning International Survey (TALIS) 2018 Conceptual Framework”, OECD Education Working Papers, No. 187, OECD Publishing, Paris, https://dx.doi.org/10.1787/799337c2-en.

[53] Andersen, S., M. Humlum and A. Nandrup (2016), “Increasing instruction time in school does increase learning”, Proceedings of the National Academy of Sciences of the United States of America (PNAS), Vol. 113/27, pp. 7481-7484, https://doi.org/10.1073/PNAS.1516686113.

[33] Baumert, J. et al. (2010), “Teachers’ mathematical knowledge, cognitive activation in the classroom, and student progress”, American Educational Research Journal, Vol. 47/1, pp. 133-180, https://doi.org/10.3102/0002831209345157.

[42] Bellens, K. et al. (2019), “Instructional quality: Catalyst or pitfall in educational systems’ aim for high achievement and equity? An answer based on multilevel SEM analyses of TIMSS 2015 data in Flanders (Belgium), Germany, and Norway”, Large-scale Assessments in Education, Vol. 7/1, pp. 1-27, https://doi.org/10.1186/s40536-019-0069-2.

[46] Blazar, D. (2015), “Effective teaching in elementary mathematics: Identifying classroom practices that support student achievement”, Economics of Education Review, Vol. 48, pp. 16-29, https://doi.org/10.1016/j.econedurev.2015.05.005.

[4] Blazar, D. and M. Kraft (2017), “Teacher and teaching effects on students’ attitudes and behaviors”, Educational Evaluation and Policy Analysis, Vol. 39/1, pp. 146-170, https://doi.org/10.3102/0162373716670260.

[58] Bloom, N. et al. (2015), “Does management matter in schools”, The Economic Journal, Vol. 125/584, pp. 647-674, https://doi.org/10.1111/ecoj.12267.

[20] Brussino, O. (2021), “Building capacity for inclusive teaching: Policies and practices to prepare all teachers for diversity and inclusion”, OECD Education Working Papers, No. 256, OECD Publishing, Paris, https://dx.doi.org/10.1787/57fe6a38-en.

[67] Burgess, S., S. Rawal and E. Taylor (2021), “Teacher peer observation and student test scores: Evidence from a field experiment in english secondary schools”, Journal of Labor Economics, Vol. 39/4, pp. 1155-1186, https://doi.org/10.1086/712997.

[14] Butler, K. (2020), “The value of direct instruction for at-risk students”, Journal of Education and Development, Vol. 4/2, pp. 10-16, https://doi.org/10.20849/jed.v4i2.741.

[37] Caprara, G. et al. (2006), “Teachers’ self-efficacy beliefs as determinants of job satisfaction and students’ academic achievement: A study at the school level”, Journal of School Psychology, Vol. 44/6, pp. 473-490, https://doi.org/10.1016/J.JSP.2006.09.001.

[13] Caro, D., J. Lenkeit and L. Kyriakides (2016), “Teaching strategies and differential effectiveness across learning contexts: Evidence from PISA 2012”, Studies in Educational Evaluation, Vol. 49, pp. 30-41, https://doi.org/10.1016/j.stueduc.2016.03.005.

[26] Carrington, W. and K. Troske (1997), “On measuring segregation in samples with small units”, Journal of Business & Economic Statistics, Vol. 15/4, pp. 402-409, https://doi.org/10.2307/1392486.

[47] Carroll, J. (1963), “A model of school learning”, Teachers College Record, Vol. 64/8, pp. 723-733, https://www.tcrecord.org/content.asp?contentid=2839.

[3] Chetty, R. et al. (2011), “How does your kindergarten classroom affect your earnings? Evidence”, Quarterly Journal of Economics, Vol. 126/4, pp. 1593-1660, https://doi.org/10.1093/qje/qjr041.

[15] Clark, R., P. Kirschner and J. Sweller (2012), “Putting students on the path to learning: The case for fully guided instruction”, American Educator, pp. 6-11, https://www.aft.org/sites/default/files/periodicals/Clark.pdf.

[7] Dee, T. (2005), “A teacher like me: Does race, ethnicity, or gender matter?”, American Economic Review, Vol. 95/2, pp. 158-165, https://doi.org/10.1257/000282805774670446.

[28] D’Haultfœuille, X., L. Girard and R. Rathelot (2021), “segregsmall: A command to estimate segregation in the presence of small units:”, The Stata Journal, Vol. 21/1, pp. 152-179, https://doi.org/10.1177/1536867X211000018.

[27] D’Haultfœuille, X. and R. Rathelot (2017), “Measuring segregation on small units: A partial identification analysis”, Quantitative Economics: Journal of the Econometric Society, Vol. 8/1, pp. 39-73, https://doi.org/10.3982/QE501.

[59] Di Liberto, A., F. Schivardi and G. Sulis (2015), “Managerial practices and student performance”, Economic Policy, Vol. 30/84, pp. 683-728, https://doi.org/10.1093/EPOLIC/EIV015.

[25] Echazarra, A. and T. Radinger (2019), “Learning in rural schools: Insights from PISA, TALIS and the literature”, OECD Education Working Papers, No. 196, OECD Publishing, Paris, https://dx.doi.org/10.1787/8b1a5cb9-en.

[8] Fairlie, R., F. Hoffmann and P. Oreopoulos (2014), “A community college instructor like me: Race and ethnicity interactions in the classroom”, American Economic Review, Vol. 104/8, pp. 2567-2591, https://doi.org/10.1257/aer.104.8.2567.

[60] Flath, B. (1989), “The principal as instructional leader”, ATA Magazine, Vol. 69/3, pp. 19-22, 47-49.

[9] Gershenson, S. et al. (2018), “The Long-Run Impacts of Same-Race Teachers”, NBER Working Paper, No. 25254, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/W25254.

[10] Gershenson, S., S. Holt and N. Papageorge (2016), “Who believes in me? The effect of student-teacher demographic match on teacher expectations”, Economics of Education Review, Vol. 52, pp. 209-224, https://doi.org/10.1016/J.ECONEDUREV.2016.03.002.

[65] Glazerman, S. et al. (2010), Impacts of Comprehensive Teacher Induction: Final Results from a Randomized Controlled Study, (NCEE 2010-4027), National Center for Education Evaluation and Regional Assistance, Institute of Education Sciences, U.S., Washington, DC, https://ies.ed.gov/ncee/pubs/20104027/pdf/20104027.pdf.

[62] Goddard, R. et al. (2015), “A theoretical and empirical analysis of the roles of instructional leadership, teacher collaboration, and collective efficacy beliefs in support of student learning”, American Journal of Education, Vol. 121/4, pp. 501-530, https://doi.org/10.1086/681925.

[19] Guerriero, S. (ed.) (2017), Pedagogical Knowledge and the Changing Nature of the Teaching Profession, Educational Research and Innovation, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264270695-en.

[57] Hallinger, P. (2011), “Leadership for learning: Lessons from 40 years of empirical research”, Journal of Educational Administration, Vol. 49/2, pp. 125-142, https://doi.org/10.1108/09578231111116699.

[34] Hill, H., B. Rowan and D. Ball (2005), “Effects of teachers’ mathematical knowledge for teaching on student achievement”, American Educational Research Journal, Vol. 42/2, pp. 371-406, https://doi.org/10.3102/00028312042002371.

[43] Hines, C., D. Cruickshank and J. Kennedy (1985), “Teacher clarity and its relationship to student achievement and satisfaction”, American Educational Research Journal, Vol. 22/1, pp. 87-99, https://doi.org/10.2307/1162989.

[36] Holzberger, D., A. Philipp and M. Kunter (2013), “How teachers’ self-efficacy is related to instructional quality: A longitudinal analysis”, Journal of Educational Psychology, Vol. 105/3, pp. 774-786, https://doi.org/10.1037/A0032198.

[35] Honicke, T. and J. Broadbent (2016), “The influence of academic self-efficacy on academic performance: A systematic review”, Educational Research Review, Vol. 17, pp. 63-84, https://doi.org/10.1016/j.edurev.2015.11.002.

[63] Ingersoll, R. and M. Strong (2011), “The impact of induction and mentoring programs for beginning teachers: A critical review of the research”, Review of Educational Research, Vol. 81/2, pp. 201-233, https://doi.org/10.3102/0034654311403323.

[5] Jackson, C. (2018), “What do test scores miss? The importance of teacher effects on non-test score outcomes”, Journal of Political Economy, Vol. 126/5, pp. 2072-2107, https://doi.org/10.1086/699018.

[64] Jackson, C., J. Rockoff and D. Staiger (2014), “Teacher effects and teacher-related policies”, Annual Review of Economics, Vol. 6/1, pp. 801-825, https://doi.org/10.1146/annurev-economics-080213-040845.

[54] Kane, T. et al. (2010), “Identifying effective classroom practices using student achievement data”, The Journal of Human Resources, Vol. 46/3, pp. 587-613, https://doi.org/10.3368/jhr.46.3.587.

[16] Kirschner, P., J. Sweller and R. Clark (2006), “Why minimal guidance during instruction does not work: An analysis of the failure of constructivist, discovery, problem-based, experiential, and inquiry-based teaching”, Educational Psychologist, Vol. 41/2, pp. 75-86, https://doi.org/10.1207/s15326985ep4102_1.

[31] Kraft, M. and J. Papay (2014), “Can professional environments in schools promote teacher development? Explaining heterogeneity in returns to teaching experience”, Educational Evaluation and Policy Analysis, Vol. 36/4, pp. 476-500, https://doi.org/10.3102/0162373713519496.

[50] Lavy, V. (2015), “Do differences in schools’ instruction time explain international achievement gaps? Evidence from developed and developing countries”, The Economic Journal, Vol. 125/588, pp. F397-F424, https://doi.org/10.1111/ECOJ.12233.

[12] Le Donné, N., P. Fraser and G. Bousquet (2016), “Teaching Strategies for Instructional Quality: Insights from the TALIS-PISA Link Data”, OECD Education Working Papers, No. 148, OECD Publishing, Paris, https://dx.doi.org/10.1787/5jln1hlsr0lr-en.

[11] Lim, J. and J. Meer (2017), “The impact of teacher-student gender matches: Random assignment evidence from South Korea”, The Journal of Human Resources, Vol. 52/4, pp. 979-997, https://doi.org/10.3368/jhr.52.4.1215-7585R1.

[41] Lipowsky, F. et al. (2009), “Quality of geometry instruction and its short-term impact on students’ understanding of the Pythagorean Theorem”, Learning and Instruction, Vol. 19/6, pp. 527-537, https://doi.org/10.1016/j.learninstruc.2008.11.001.

[48] Muijs, D. et al. (2014), “State of the art: Teacher effectiveness and professional learning”, School Effectiveness and School Improvement, Vol. 25/2, pp. 231-256, https://doi.org/10.1080/09243453.2014.885451.

[45] Mullis, I. et al. (2020), TIMSS 2019 International Results in Mathematics and Science, Retrieved from Boston College, TIMSS & PIRLS International Study Center website, https://timssandpirls.bc.edu/timss2019/international-results/.

[56] OECD (2021), Positive, High-achieving Students?: What Schools and Teachers Can Do, TALIS, OECD Publishing, Paris, https://dx.doi.org/10.1787/3b9551db-en.

[17] OECD (2020), Global Teaching InSights: A Video Study of Teaching, OECD Publishing, Paris, https://dx.doi.org/10.1787/20d6f36b-en.

[22] OECD (2019), Balancing School Choice and Equity: An International Perspective Based on Pisa, PISA, OECD Publishing, Paris, https://dx.doi.org/10.1787/2592c974-en.

[32] OECD (2019), TALIS 2018 Results (Volume I): Teachers and School Leaders as Lifelong Learners, TALIS, OECD Publishing, Paris, https://dx.doi.org/10.1787/1d0bc92a-en.

[39] OECD (2019), TALIS 2018 Technical Report, OECD, Paris, http://www.oecd.org/education/talis/TALIS_2018_Technical_Report.pdf.

[23] OECD (2017), Educational Opportunity for All: Overcoming Inequality throughout the Life Course, Educational Research and Innovation, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264287457-en.

[24] OECD (2017), The Funding of School Education: Connecting Resources and Learning, OECD Reviews of School Resources, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264276147-en.

[61] OECD (2016), School Leadership for Learning: Insights from TALIS 2013, TALIS, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264258341-en.

[6] OECD (2012), Equity and Quality in Education: Supporting Disadvantaged Students and Schools, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264130852-en.

[29] Papay, J. and M. Kraft (2015), “Productivity returns to experience in the teacher labor market: Methodological challenges and new evidence on long-term career improvement”, Journal of Public Economics, Vol. 130, pp. 105-119, https://doi.org/10.1016/j.jpubeco.2015.02.008.

[21] Peterson, A. et al. (2018), “Understanding innovative pedagogies: Key themes to analyse new approaches to teaching and learning”, OECD Education Working Papers, No. 172, OECD Publishing, Paris, https://dx.doi.org/10.1787/9f843a6e-en.

[52] Pischke, J. (2007), “The impact of length of the school year on student performance and earnings: Evidence from the German short school year”, The Economic Journal, Vol. 117/523, pp. 1216-1242, https://doi.org/10.1111/J.1468-0297.2007.02080.X.

[2] Rivkin, S., E. Hanushek and J. Kain (2005), “Teachers, schools and academic achievement”, Econometrica: The Journal of the Econometric Society, Vol. 73/2, pp. 417-458, https://doi.org/10.1111/j.1468-0262.2005.00584.x.

[51] Rivkin, S. and J. Schiman (2015), “Instruction time, classroom quality, and academic achievement”, The Economic Journal, Vol. 125/588, pp. F425-F448, https://doi.org/10.1111/ECOJ.12315.

[66] Rockoff, J. (2008), “Does Mentoring Reduce Turnover and Improve Skills of New Employees? Evidence from Teachers in New York City”, NBER Working Paper Series, No. 13868, National Bureau of Economic Research, Cambridge, MA, http://www.nber.org/papers/w13868.

[49] Schmidt, W., P. Zoido and L. Cogan (2014), “Schooling Matters: Opportunity to Learn in PISA 2012”, OECD Education Working Papers, No. 95, OECD Publishing, Paris, https://dx.doi.org/10.1787/5k3v0hldmchl-en.

[44] Seidel, T., R. Rimmele and M. Prenzel (2005), “Clarity and coherence of lesson goals as a scaffold for student learning”, Learning and Instruction, Vol. 15/6, pp. 539-556, https://doi.org/10.1016/j.learninstruc.2005.08.004.

[55] Stronge, J. et al. (2007), “What is the Relationship Between Teacher Quality and Student Achievement? An Exploratory Study”, Journal of Personnel Evaluation in Education, Vol. 20/3-4, pp. 165-184, https://doi.org/10.1007/s11092-008-9053-z.

[40] Tschannen-Moran, M. and A. Hoy (2001), “Teacher efficacy: Capturing an elusive construct”, Teaching and Teacher Education, Vol. 17/7, pp. 783-805, https://doi.org/10.1016/S0742-051X(01)00036-1.

[30] Wiswall, M. (2013), “The dynamics of teacher quality”, Journal of Public Economics, Vol. 100, pp. 61-78, https://doi.org/10.1016/J.JPUBECO.2013.01.006.

[38] Woolfolk Hoy, A. and H. Davis (2006), “Teacher self-efficacy and its influence on the achievement of adolescents”, in Urdan, T. and F. Pajares (eds.), Self-Efficacy Beliefs of Adolescents, Information Age Publishing, Greenwich, CT.


← 1. See OECD (2019[22]) for an application of the dissimilarity index in the analysis of students’ segregation across schools according to socio-economic background.

← 2. Appendix tables also present analyses comparing schools according to the share of students with a foreign language background and who have special education needs. These results are not commented on in the text.

← 3. “Socio-economically disadvantaged homes” refers to homes lacking the basic necessities or advantages of life such as adequate housing, nutrition or medical care; the evaluation of whether students lived in disadvantaged home is left to the evaluation of the school principal.

← 4. A privately managed school is a school whose principal reported that it is managed by a non-governmental organisation (e.g. a church, trade union, business or other private institution). In some countries, the privately managed schools category includes schools that receive significant funding from the government (government-dependent private schools). A publicly managed school is a school whose principal reported that it is managed by a public education authority, government agency, municipality, or governing board appointed by the government or elected by public franchise. In the principal questionnaire, this question does not make any reference to the source of the school’s funding, which is reported in the preceding question.

← 5. When a latent (unobservable) construct like self-efficacy is measured by a self-report questionnaire, measurement invariance refers to the property that using the same questionnaire in different groups (such as countries) measures the same construct in the same way. Lack of measurement invariance would imply that the score in the self-efficacy scale for respondents in one country is not comparable to the scores of respondents in a different country.

Metadata, Legal and Rights

This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area. Extracts from publications may be subject to additional disclaimers, which are set out in the complete version of the publication, available at the link provided.

© OECD 2022

The use of this work, whether digital or print, is governed by the Terms and Conditions to be found at https://www.oecd.org/termsandconditions.