Measuring school leadership and management and linkages with literacy: Evidence from rural and township primary schools in South Africa

Abstract

This paper describes a process to develop and trial new metrics in South Africa to quantify school leadership and management practices or processes that are considered theoretically related to literacy outcomes. The predictive validity of these measures is assessed in challenging contexts, including 60 township and rural primary schools in South Africa. We observe a randomness to how better leadership and management practices are distributed across better and worse performing schools. Regression analyses confirm weak and inconsistent linkages between measured leadership and management dimensions and literacy outcomes across the sample. However, we find evidence of stronger linkages with intermediate outcomes, including monitoring curriculum coverage. This research contributes to a burgeoning, yet underdeveloped literature on educational management and leadership in Africa and the challenges of measurement in this context.

Keywords

Leadership management challenging contexts South Africa literacy

Introduction

Since the 1960s, there has been significant growth in the global knowledge base on school leadership and management (SLM). However there is an acknowledgement of the limits of this knowledge, with ‘far less systematic knowledge on how school leaders carry out their roles in developing nations throughout the world’ (Hallinger, 2017: 363). In Africa, educational management and leadership literature is in an emergent phase, with contributions dominated by one or two countries. Even South Africa, as the largest country contributor to the African literature, derives its knowledge base predominantly from qualitative research studies, with case-study locations limited to a few schools (Bush and Glover, 2016; Hallinger, 2017). There is limited understanding of the empirical linkages between SLM and learning outcomes in the country and Africa more broadly (Hallinger, 2017; Hoadley et al., 2009).

Yet there has been a long-held view that much of the unexplained variation in learner performance in South Africa, particularly across historically disadvantaged school contexts, may be attributed to differences in SLM (Crouch and Mabogoane, 1998, 2001). Subsequent to this finding, school effectiveness studies began to include proxy indicators of SLM into education production function models of learning (see Wills, 2019 for a review). These studies suggest there is educational value in protecting and extending learning time (Gustafsson, 2007; Van der Berg and Louw, 2006); managing the procurement and inventory of books and stationery (Kotze, 2017; Shepherd, 2016; Spaull, 2013; Van der Berg, 2008); quality assuring tests; monitoring learner results; and, importantly, managing and monitoring time-on-task, opportunity-to-learn and curriculum coverage (Carnoy et al., 2015; Taylor, 2011; Taylor and Prinsloo, 2005).

These are useful insights, but the inclusion of these indicator variables in production function models has been far from sufficient in accounting for large unexplained variation in learning initially attributed to management. Furthermore, many SLM indicators that one would assume to matter for learning have been found to be insignificant (Gustafsson, 2005; Van Staden and Howie, 2014). Rather than operating through strong direct effects, this raises the possibility that SLM may interact with learning in indirect ways, with small direct associations (Hallinger and Heck, 1998). However, a question arises as to whether this null hypothesis holds if more rigorous efforts are applied to quantifying SLM, including the development of new tools and centring measurement on clearer conceptual frameworks.

In this paper we describe a process to develop and trial new metrics for codifying SLM practices or processes that may be linked to better literacy in poor South African schools. This is centred on a ‘leadership for literacy’ conceptual framework (Hoadley, 2018) where we prioritise six categories of resources available to school leaders in shaping a literacy learning environment in South Africa. Our first objective is to quantify the availability, use and deployment of these resources by school leaders. The second objective is to assess the reliability and validity of derived ‘leadership for literacy’ resource indices. The data for this process were gathered from 60 township and rural public schools in three South African provinces. In addition to SLM data, we collected reading and literacy scores for grade 3 and 6 students to establish whether our ‘leadership for literacy’ indices are predictive of literacy levels and gains using multivariate regression modelling. At this outset, we acknowledge that we can only identify correlational rather than causal relationships in the analysis. The qualitative case-study component of the wider mixed-methods study drilled further into the generative mechanisms underlying the quantitative findings (as discussed extensively in Taylor et al. (2019)). This paper aims to give specific treatment to the quantitative research process, analytical strategy and quantitative results.

The paper contributes to a knowledge base on leadership and management in challenging primary school contexts in Africa. It highlights that, despite efforts to measure SLM more rigorously, weak and inconsistent associations with literacy are still found in poor primary school contexts. Nevertheless, somewhat stronger associations are evident with intermediate outcomes such as monitoring curriculum coverage.

The next section provides background on measuring SLM and its linkages with learning. We consider three different approaches to instrument development and identify the implications of existing study findings for our methodological approach. We then provide more details on our methodology and data collected. This is followed by a discussion of the reliability and validity of the six ‘leadership for literacy’ indices, particularly focusing on predictive validity, before the final section concludes.

Background

Internationally, qualitative studies yield support for the educational value of leadership, particularly when framed from an instructional leadership perspective (Leithwood et al., 2004; Robinson et al., 2008). Paradoxical results are however found in quantitative studies. A meta-analysis by Robinson et al. (2008) and Marzano et al. (2005) find moderate to strong correlations with learning. However, in a well-cited meta-analysis of 37 multi-national studies by Witziers et al. (2003), correlations between leadership and student outcomes are considered to be weak. This echoes earlier reflections by Hallinger and Heck (1998) that the effect of SLM on learning is small and indirect, mediated through alternate mechanisms such as the work of teachers.

Robinson et al. (2008), in attempting to explain these paradoxical results, argue that they should be examined against various other possible explanations for the predominance of null or weak associations. These include the use of narrow theoretical frameworks in conceptualising SLM; small sample sizes; and limited detectable variation in SLM practice that compromises the identification of direct effects (Robinson et al., 2008). Instruments may also be inadequate to measure SLM, particularly in low-income settings (Lemos and Scur, 2017). The limits of existing instruments to capture leadership and management quality have been suggested as econometricians begin to isolate out large causal ‘principal effects’ (Branch et al., 2012; Grissom et al., 2015a) or identify significant and large positive causal impacts of SLM interventions on learning (Fryer, 2017; Tavares, 2015).

Despite burgeoning empirical work on the determinants of learning outcomes in South Africa, studies have not aimed to capture SLM as a fuller construct (one exception is a small qualitative study by Taylor et al. (2013)). It has not been possible to determine whether the weak explanatory power of SLM indicators in predicting learning outcomes in historically disadvantaged schools is due to poor measurement of SLM or a lack of variation in existing SLM proficiencies or learning outcomes across schools. In general, too little attention is given to research methodology and the development of effective assessment tools in establishing a valid knowledge base underlying the practice of educational leadership and management (Heck and Hallinger, 2005: 232).

Approaches to quantifying SLM

Three broad sets of instrument-based approaches are typically used in existing studies to quantify SLM, including:

detecting leadership behaviours or effectiveness with self-administered questionnaires;

quantifying school leaders’ allocations of time to different tasks; and

assessing SLM competencies using scoring rubrics.

Here we identify the advantages and disadvantages of each approach and how findings from existing studies adopting these approaches informed our study design.

Self-administered questionnaires are commonly used to evaluate leadership behaviours and effectiveness (for a list of instruments see Condon and Clifford (2012)). Instruments may vary in terms of their reliability and validity (Goldring et al., 2009) and even rigorously developed instruments have shortcomings. Yet there is still evidence that many of these instruments provide measures associated with learning.

Two key reflections from these studies have implications for our design. First, ratings tend to be inflated, even when provided by supervisors or peers, so that less than desirable variation is detected in leadership across schools (Grissom et al., 2015b: 22; McCullough et al., 2016). But data collected from teachers are typically more reliable compared with principal self-reports or supervisor reports (Hallinger, 2008: 16). Collecting responses from teachers in addition to school leaders is therefore a preferred approach. Second, the strength of associations with learning may vary depending on the subject and grade level in question (Grissom et al., 2015b; McCullough et al., 2016). This calls for learning-centred leadership theories where leadership is conceptualised with a subject-specific and grade-specific end in mind (Stein and Nelson, 2003).

Existing studies also quantify the work of school leaders in terms of time allocations to different tasks measured through work activity analyses, structured observations or self-reported activity logs (Camburn et al., 2010; Grissom et al., 2015b; Horng et al., 2009; Lee and Hallinger, 2012). These studies highlight that SLM tasks comprise a diverse set of functions, but allocating more time to some tasks over others may yield higher returns to learning (Grissom et al., 2015b). Time use and allocations to tasks may also vary within leaders in the same school, within the same leaders over time (Camburn et al., 2010) and across societies as a result of differential economic, socio-cultural and institutional dynamics (Lee and Hallinger, 2012). In response, our SLM assessment was designed to be contextually appropriate for the South African public schooling context. Data are collected about more than one leader or manager in a school and at two different time points.

But how does one combine data from different times, several respondents and different questionnaires? A predetermined scoring rubric was developed for this purpose to facilitate data triangulation. Rubrics are increasingly used internationally to quantify competencies in education management, assessment, or systems technologies (Arcia et al., 2015; Bloom and Van Reenen, 2010; Lemos and Scur, 2017). Guided by rubric descriptions, different data sources can be combined to assess how institutional practices or processes compare to described proficiency standards.

A well-known scoring rubric is the World Management Survey (WMS). Its application in large studies reveals that higher levels of formalisation of school management practices and processes are associated with better learning outcomes and teacher practices, even in developing countries (Bloom et al., 2015; Crawfurd, 2017; Scur, 2017). But this approach depends on using high-level researchers to conduct interviews with head teachers and in turn establishing SLM proficiency scores. This may be costly and infeasible in environments with few available researchers speaking indigenous languages. We aimed to use a rubric approach but to limit costs and improve scalability by using fieldworkers rather than experienced researchers to collect data. Close-ended questions are relied on more heavily in our approach to reduce the need for value-judgements.

Method and data

SLM measurement approach

Our SLM measurement approach was operationalised in four phases as shown in Figure 1, including the development of a theoretical framework, its empirical application across interconnected qualitative and quantitative fieldwork components, and final index construction.

Phase 1 – Theoretical framework

A ‘leadership for literacy’ framework underpinned our study design (Hoadley, 2018), where its literacy focus is informed through two considerations. The first is the notion that leadership effectiveness is deeply connected to leaders’ subject-specific knowledge resources (Burch and Spillane, 2003; Stein and Nelson, 2003). Second, the literacy focus is contextually appropriate where an inability to read for meaning pervades the South African schooling system, compromising individual and economic advancement. Almost eight of every ten South African grade 4 students cannot read for meaning in any language, as revealed in the 2016 Progress in International Reading and Literacy Study (PIRLS) (Howie et al., 2017). South Africa was ranked last of the 50 countries participating in PIRLS 2016 (Mullis et al., 2017). As Murphy (2004) emphasises, in low literacy contexts leadership should be leveraged to move new knowledge on literacy into classrooms and schools. In this vein, our framework – informed through a review of school effectiveness and leading for reading literature (see Hoadley, 2018) – identifies six clusters of resources available to leaders in promoting literacy in disadvantaged South African schools:

material resources (time): the allocation of time for teaching language and reading, as well as protecting and maximising the use of time for learning;

material resources (presence of text): the prioritisation and procurement of textbooks,¹ graded readers and library resources to support a programme of reading;

material resources (use of text): the optimal utilisation of these text resources for reading;

knowledge resources: within-school expertise in understanding the value of reading and the methods to teach it effectively, as well as the extent to which this understanding is shared by school leaders and teachers;

human resources: the effective recruitment, utilisation and development of teaching expertise, particularly in reading and language; and

strategic resources: the mobilisation of available resources to drive a coherent literacy programme.

Each of these resources is expanded on in Table 1. This is by no means a comprehensive list of factors that may be important for literacy development. Murphy (2004) for example considers ten sets of resources in his ‘leadership for literacy’ meta-analysis. But in developed countries, with different schooling systems and persistent shortages of resources and skills, prioritising a few factors that may promote baseline levels of functionality and learning is instructive. Furthermore, where much of the research on leading for literacy emanates from developed countries, the issue of basic resources tends to be downplayed (Lee and Zuze, 2011). Pretorius and Machet (2004) for example draw attention to three obvious resources that define the context for literacy accomplishment in disadvantaged schools: instructional time, books and good teachers.

Table 1.

Six ‘leadership for literacy’ dimensions with sub-dimension descriptions.

LL index dimensions	Sub-dimension	N elements
	Allocation / structure of time for teaching of language and reading	5
Material resources: time	Maximum use of teaching time (limited disruptions and few free periods)	4
Material resources: time	Low absenteeism and teacher presence in classroom	5
	Additional time for reading beyond class	5
Material resources: presence of text	Presence of text in Grade 3 classroom	10
Material resources: presence of text	Presence of text in Grade 6 classroom	9
	Use of text in grade 3 classroom	3
Material resources: use of text	Use of text in grade 6 classroom	3
Material resources: use of text	School-wide management of resources to promote the availability and use of text	3
	Culture of reading among teachers	9
Knowledge resources	Knowledge of teaching reading	7
Knowledge resources	Knowledge of remediation	2
	Knowledge sharing – professional collaboration	4
	The presence of managers and leaders in the school to promote reading	4
	Qualification levels; teacher and school management team (SMT) alignment of specialisations to subjects and phase taught	4
	Presence of a reading expert in the school	2
Human resources	Acknowledging and rewarding teacher performance	2
	Professional development – teacher exposure to professional development opportunities including workshops on reading instruction	6
	Managing poor performance and consequence management	4
	Appointing staff and attracting talent to promote improved teacher quality	3
	Use of networks and financial management to support a reading programme	4
	Evidence of reading assessment practices	3
	Performing tracking of parameters, including reading and reviews of performance (whole staff meetings and one on one meetings)	3
Strategic resources	Monitoring of lessons and curriculum coverage	4
	Clear strategies to create a reading programme (programmes implemented, celebration of reading, promoting enjoyment of reading)	5
	Vision, goal setting and expectations – school goals incorporate ‘improved reading’	4

We were interested in quantifying both leadership and management functions despite these being distinct concepts. The theoretical framework assumes these functions are shared across individuals in the school organisation (Spillane et al., 2004). This is in line with South African school policy which envisages SLM as distributed across school management team members comprising the principal and middle managers (deputy principal and heads of department) (Department of Basic Education (DBE), 2016).

Phase 2 – Qualitative process

While we do not report on the qualitative findings in this paper, this process was critical to aligning the quantitative instrument development process with the theoretical framework. Case studies were conducted in four pairs of high and lower achieving schools selected from the larger 60-school sample in township and rural areas. Three to four days were spent in each case-study school, where visits were led by a team of four South African-based education specialists. Semi-structured interviews with principals, teachers and middle managers were conducted in these schools. Interview findings were augmented with classroom and school observations, focusing on identifying factors that support or hinder a literacy-learning environment. This case-study process generated thick descriptions to augment the development of rubric descriptions as described in the next phase, and in turn the writing of questions incorporated into quantitative instruments.

Phase 3 – Quantitative process

Developing the scoring rubric involved mapping each of the six resource dimensions from the theoretical framework into descriptions of SLM proficiency. These dimensions were further broken down into sub-dimensions which comprise individual rubric elements (see Table 1). In total, scoring descriptions were written for 114 rubric elements. As an example, Table 2 illustrates how rubric descriptions relate to quantitative scores of 1 (low) to 5 (high) for three elements under the human resource dimension.

Table 2.

Example of scoring rubric descriptions for three elements under the human resource dimension.

	Score 1 (low)	Score 2	Score 3 (middle)	Score 4	Score 5 (high)
Use of a reading expert – assesses whether reading experts provide reading instruction to teachers or learners	NA – no reading expert.	There is a reading expert but no respondents indicate that the reading expert helps teachers with how to teach reading. But may indicate that the teacher supports students with their reading ‘quite a lot’ or ‘a lot’.	There is a reading expert and 1 of 3 respondents indicate that the reading expert helps teachers how to teach reading.	There is a reading expert and 2 of 3 respondents indicate that the reading expert helps teachers with how to teach reading.	There is a reading expert and either all 3 respondents indicate that the reading expert helps teachers with how to teach reading OR 80% or more of language educators indicate they go to the reading specialist for help at least once a term.
The school has a system for acknowledging its best teachers through rewards/awards	Other than the Integrated Quality Management System (IQMS) – a weak evaluative system – no system exists for acknowledging best teachers through rewards/awards.	Other than IQMS, there is a system for acknowledging the best teachers through rewards/awards.	Other than IQMS, there is a system for acknowledging the best teachers through rewards/awards. It singles out individuals rather than just awarding all or most teachers.	Other than IQMS, there is a system for acknowledging its best teachers through rewards/awards that singles out individuals rather than just awarding all or most teachers. Awards/rewards happen regularly (once or more than once a year).	Other than IQMS, there is a system for acknowledging its best teachers through rewards/awards that singles out individuals rather than just awarding all or most teachers. Awards/rewards happen regularly (once or more than once a year). Clear criteria used to determine who gets an award.
The School Governing Body (SGB) supports effective hiring as indicated by the principal	The SGB does not have necessary competencies and skills to make good recommendations on staff appointments. The SGB does not have necessary competencies and skills to fulfil their functions. The school would be ‘much better’ or ‘a little better off’ if they had no SGB.	The SGB does not have necessary competencies and skills to make good recommendations on staff appointments but may have competencies in other areas.	The SGB has necessary competencies and skills to make good recommendations on staff appointments.	The SGB has necessary competencies and skills to make good recommendations on staff appointments and has necessary competencies and skills to fulfil their functions. The school would be ‘worse off’ if they had no SGB.	The SGB has necessary competencies and skills to make good recommendations on staff appointments and has necessary competencies and skills to fulfil their functions. The school would be ‘much worse off’ if they had no SGB.

Close-ended questions were written to generate variables necessary to evaluate how a school should be scored on each rubric element. Relying on close-ended rather than open-ended questions reduces the need for probing and research judgements from fieldworkers. The questions were allocated across six questionnaires, which can be administered during one school day. These include interviews with the principal, deputy principal, a grade 3 teacher and grade 6 teacher; a self-administered survey for all teachers; and a school observational instrument. Observational questions required fieldworkers to verify the existence of factors such as book resources in classrooms, the state or existence of school libraries, the availability of management documents (e.g. timetables, assessment plans and financial documents) and the presence of teachers in classrooms.

Before finalising instruments, they were subjected to multiple stages of piloting. We initially trialled instruments in four schools located across three provinces, then revisions were made and we re-piloted instruments in another three schools. Additionally, all instruments were reviewed by four former principals and two district officials working in rural school contexts. We also engaged in cognitive piloting of the teacher survey in two peri-urban schools in KwaZulu-Natal.

A dedicated fieldworker, trained over a four-day process on the instruments, was responsible for conducting SLM interviews and collecting observation data for each school. The training of fieldworkers was supported through an in-school simulation day. Senior project researchers were present during the school simulation to monitor the quality of interviews and data collection process.

Phase 4 – Index construction

Once data were collected, a data-coding process combined variables to ‘objectively’ score each rubric element. The scoring is objective in the sense that the data determine each school’s score on a rubric element rather than a researcher using a value-judgement. In total, over 500 variables collected across the six instruments were used to code 114 rubric elements. Finally, the six ‘leadership for literacy’ dimensions were obtained using principal components analysis (PCA) to weight each element in terms of the variation each element explains in an underlying unobserved factor.

Reliability

Reliability was embedded into the measurement process by collecting data from different sources and respondents. Almost half of the elements are coded using data that are triangulated in some way; for example, using responses from multiple respondents. Of the 114 elements, 42 rely on data that require the fieldworker to observe evidence of practice. Due to potential inaccuracies in self-reported data from teachers and school leaders, we also constructed a second set of combined indices only using elements coded from observational or ‘evidence-based’ data. (The supplementary materials provide distributional plots of these two sets of ‘leadership for literacy’ indices.)

Although we collected data at two points in time, we did not administer two rounds of the same instruments, so that evaluating the test–retest reliability is not possible. However, we do test whether the indices discriminate in ways that we would expect, such as observing better practices in urban as opposed to rural schools or in the least poor schools in the sample.

Our theoretical framework is multi-dimensional in its conceptualisation. Conducting tests of internal consistency, such as Cronbach’s alpha, which assume construct uni-dimensionality, are not appropriate (Peters, 2014).

Content and face validity

Methods to establish the validity of the measurement approach were incorporated into each of the four phases in Figure 1. The literature review, its qualitative application in schools, the rubric development process and several rounds of instrument piloting were critical to establishing content validity. The mixed methods design also allowed for face-validity checks (i.e. at face value, do our metrics measure what they claim to measure?). We also evaluated the predictive validity of the SLM measures. The method and analytical strategy to do this are now discussed.

School sampling

A possible reason for weak associations between SLM and learning is that study samples lack detectable variation in both student learning and SLM practices. In addressing this concern, the 60 schools were purposively selected to artificially add as much student performance variation as exists in the sampling frame of public schools reaching poorer populations in three of nine South African provinces (KwaZulu-Natal, Gauteng and Limpopo). As described in Wills (2017), we initiated an intensive search for the 30 best performing schools in these provinces that charge minimal or no school fees.² This involved identifying top-performing schools in system-wide testing data – namely the Annual National Assessments. We corroborated these data against several recommendations of ‘good’ no-fee schools collected from various sources. Then 30 lower performing schools, in similar geographic locations, were matched to the higher performing schools. School selection was also informed by whether the dominant student language in the school was Zulu, Sepedi or Xitsonga.

Literacy data

Grade 3 and 6 students were tested in these 60 schools at the beginning of 2017. Pre-testing was necessary to verify school literacy performance levels at project inception and to guide the selection of matched school pairs for the qualitative case-study work. We then re-tested the same learners towards the end of the same school year (i.e. post-test).

The grade 6 English literacy test, administered to an entire class of grade 6 students in each school, consisted of two silent reading comprehension passages (released items from grade 4 PIRLS assessments³) and a vocabulary test. From the original sample of 2656 grade 6 students, 2379 also wrote the post-test, implying a low attrition rate of 11%.

Additionally, pre- and post-test English Oral Reading Fluency (ORF) scores are available for 599 grade 6 students (drawn from roughly 10 randomly selected students in each school). We have African language ORF scores for these students but only for the end of 2017. A battery of reading and literacy tasks in English and African language were also administered to 10 randomly selected grade 3 students in each school. Pre- and post-test scores are available for 631 grade 3 students.

Analytical strategy

The predictive validity of the ‘leadership for literacy’ indices is assessed using an education production function framework. Literacy outcomes are expressed as a function of each ‘leadership for literacy’ index, $L L_{i s}$ , and a set of individual or home, and school characteristics:

Y_{i s t} = β_{0} + β_{1} L L_{i s} + β_{1} I_{i s} + β_{2} S_{s} + ∊_{i s}

Here $Y_{i st}$ represents the post-test scores of the i_th student in school s. $I_{i s}$ is a vector of the student’s background characteristics including a measure of wealth⁴ and its square, the student’s age, gender and whether the student attended grade R (equivalent to kindergarten). Additional background controls include whether the student gets help with homework, living arrangements with parents, parents’ employment status, rural versus urban location, and speaking English at home ‘sometimes’ or ‘always’ rather than ‘never’. School controls, S_s , include average school wealth, class size, low-fee versus no-fee status and whether a school’s language of learning or teaching (LOLT) in the foundation phase is English. To minimise measurement error in the controls, averages are used across data points collected at the beginning and end of the year.

Equation (1) estimates levels of learning and is unlikely to control for unobserved factors such as student ability or historical exposure to different factors which may contribute more to the learning process than observable school factors. For this reason, a value-added model which includes a pre-test score $Y_{i s t - 1}$ is also used to estimate whether the ‘leadership for literacy’ indices explain any differences in literacy skills gained within a year (see equation 2). While value-added models are prone to measurement error, we are quite confident that the outcome measures are reliable with very high correlations across pre-test and post-test scores. For example, the Pearson correlation statistic was almost 0.90 across the grade 6 English literacy pre- and post-test scores.

Y_{i s t} = β_{0} + β_{1} Y_{i s t - 1} + β_{2} L L_{i s} + β_{3} I_{i s} + β_{4} S_{i s} + ∊_{i s}

Finally, we also explore whether associations exist between our indices and intermediate outcomes.

I n t e r m e d i a t e_{s} = β_{0} + β_{1} L L_{s} + β_{2} S_{s} + ∊_{s}

In equation 3, intermediate variables vary at the school level (not the student level), so we also average teacher and student characteristics (S_s ) to the school level. The four intermediate variables considered include:

coverage of work in the best grade 6 learner’s exercise or workbook (expressed in centiles);

the percentage of classrooms with a present teacher and students engaged in a learning activity;

the percentage of teachers indicating their head of department conducts at least biweekly checks of their curriculum coverage; and

an index of average teacher ‘employee engagement’. An engagement score is generated for teachers in each school, averaged to the school level and normalised. The score is constructed from eight questions related to work enjoyment, satisfaction with the school management team, sense of safety, stress levels, receiving praise or recognition, and feeling valued.

Results

Reliability

One way to assess whether the ‘leadership for literacy’ indices are reliably measuring what we intend them to is to compare index scores across rural and urban schools, and across the poorest 20% of schools and the least poor 20% of schools. In general, we would expect higher scores for urban and wealthier schools compared to rural and poorer schools.

The expected patterns hold for four of six indices, as seen in Table 3. However, the human resources and strategic resources indices are higher in poorer than wealthier schools – a pattern we would not expect. Limiting the indices to being constructed from ‘evidence-based’ elements yields patterns on the strategic resources index that are somewhat more aligned with expectations.

Table 3.

‘Leadership for literacy’ indices by location and school wealth.

		Rural	Urban	Poorest20%	Least poor20%
Material resources: time (z-score)	PCA	−0.03	0.04	−0.02	−0.15
	Evidence-based	−0.04	0.04	−0.27	0.19
Material resources: text presence (z-score)	PCA	−0.5	0.6***	−0.6***	0.5
	Evidence-based	−0.4	0.5***	−0.4**	0.4
Material resources: text use (z-score)	PCA	−0.16	0.19	−0.05	0.43
	Evidence-based	−0.18	0.22	-0.07	0.47
Knowledge resources (z-score)	PCA	−0.02	0.02	−0.26	0.21
	Evidence-based	−0.10	0.12	−0.33	0.28
Human resources (z-score)	PCA	0.33	−0.4***	0.4***	−0.53
	Evidence-based	0.36	−0.4***	0.4***	−0.70
Strategic resources (z-score)	PCA	0.43	−0.5***	0.4*	−0.25
	Evidence-based	−0.38	0.5***	−0.4***	0.60
N schools		33	27	18	13

Note: Differences are statistically significantly different at *** p < 0.01, ** p < 0.05, * p < 0.1.

Face validity

In a mixed-methods paper we examine convergences and divergences between the quantitative data collected through our instruments and the qualitative findings for eight topics that were probed in the case-study schools (Taylor et al., 2019 ). Qualitative and quantitative findings only converged for two of eight topics compared. The convergence on these two topics was not surprising as they could be assessed using observational data. Measuring the other six topics requires collecting self-reported recall or perception-based responses as they cannot be directly observed. This opens the door for less reliable socially acceptable responses. Given these face-validity concerns, we assess whether predictive validity improves if we only construct indices from observational data (i.e. excluding elements derived from self-reported or recall responses).

Predictive validity

Descriptive findings

Overall, systematically better SLM practices are not detected in better performing schools in our sample. This is initially seen in Figure 2, which plots the raw correlation between the six ‘leadership for literacy’ dimensions and grade 6 English literacy test scores in reading comprehension and vocabulary (expressed in z-scores) at the 50th and 90th percentile. The lack of any systematic relationship for at least five of the six dimensions is evident. While the material presence of text in the school shows a positive association, in the multivariate estimations this is absorbed by controls for socio-economic status.

Multivariate estimation of literacy outcomes

The overall impression of no or weak associations between the ‘leadership for literacy’ indices and learning is confirmed in the multivariate analyses, seen in Tables 4 and 5. If positive and significant results do emerge, these are inconsistent across grades and different test outcomes.

Table 4.

Estimating grade 6 literacy outcomes.

	A) Estimating Gr. 6 English literacy (reading comprehension and vocabulary test, z-scores)				B) Estimating Gr. 6 ORF (z-scores)				C) Estimating Gr. 6 African language ORF (% of WCPM)
	Levels		Value-added		Levels		Value-added		Levels		Value-added
	Model 1	Model 2	Model 1	Model 2	Model 1	Model 2	Model 1	Model 2	Model 1	Model 2	Model 1	Model 2
Material resources Time (z-score)	0.05	0.06	0.01	0.00	0.00	−0.01	0.01	0.00	0.16	0.14	0.30	0.21
Material resources Time (z-score)	(0.08)	(0.06)	(0.02)	(0.02)	(0.08)	(0.05)	(0.02)	(0.02)	(0.67)	(0.55)	(0.54)	(0.33)
Pre-test control*			1.160***	1.134***			0.924***	0.895***			0.404***	0.427***
Pre-test control*			(0.03)	(0.02)			(0.04)	(0.04)			(0.03)	(0.03)
Adjusted R-squared	0.00	0.25	0.80	0.80	0.00	0.20	0.77	0.77	0.00	0.16	0.48	0.59
Material resources: text presence (z-score)	0.181**	−0.06	0.02	0.01	0.0987*	−0.09	−0.01	−0.0549*	−0.13	0.07	−1.003*	0.35
Material resources: text presence (z-score)	(0.07)	(0.06)	(0.02)	(0.02)	(0.06)	(0.07)	(0.02)	(0.03)	(0.67)	(0.75)	(0.56)	(0.48)
Pre-test control*			1.158***	1.134***			0.925***	0.893***			0.410***	0.427***
Pre-test control*			(0.03)	(0.02)			(0.04)	(0.04)			(0.03)	(0.03)
Adjusted R-squared	0.02	0.25	0.80	0.80	0.01	0.20	0.77	0.77	0.00	0.16	0.48	0.59
Material resources: text use (z-score)	0.05	−0.03	−0.03	−0.03	0.04	−0.01	−0.02	−0.02	0.12	0.44	−0.41	0.38
Material resources: text use (z-score)	(0.11)	(0.08)	(0.03)	(0.03)	(0.09)	(0.07)	(0.03)	(0.03)	(0.82)	(0.61)	(0.73)	(0.38)
Pre-test control*			1.162***	1.134***			0.925***	0.896***			0.405***	0.427***
Pre-test control*			(0.03)	(0.02)			(0.04)	(0.04)			(0.03)	(0.03)
Adjusted R-squared	0.00	0.25	0.80	0.80	0.00	0.19	0.77	0.77	0.00	0.17	0.48	0.59
Knowledge resources (z-score)	0.06	0.00	0.02	0.02	0.03	0.01	0.00	0.01	−0.15	0.42	−0.38	0.46
Knowledge resources (z-score)	(0.10)	(0.06)	(0.02)	(0.02)	(0.08)	(0.06)	(0.03)	(0.02)	(0.66)	(0.65)	(0.63)	(0.35)
Pre-test control*			1.159***	1.134***			0.924***	0.895***			0.404***	0.427***
Pre-test control*			(0.02)	(0.02)			(0.04)	(0.04)			(0.03)	(0.03)
Adjusted R-squared	0.00	0.25	0.80	0.80	0.00	0.20	0.77	0.77	0.00	0.17	0.48	0.59
Human resources (z-score)	0.01	0.181**	0.0783**	0.100***	−0.01	0.162**	0.01	0.05	1.03	1.066*	1.527**	0.727*
Human resources (z-score)	(0.07)	(0.06)	(0.03)	(0.03)	(0.06)	(0.07)	(0.03)	(0.03)	(0.66)	(0.57)	(0.57)	(0.39)
Pre-test control*			1.163***	1.126***			0.924***	0.889***			0.409***	0.425***
Pre-test control*			(0.03)	(0.02)			(0.04)	(0.04)			(0.03)	(0.03)
Adjusted R-squared	0.00	0.26	0.80	0.81	0.00	0.21	0.77	0.77	0.01	0.17	0.49	0.59
Strategic resources (z-score)	−0.03	0.10	0.02	0.03	−0.10	0.07	0.01	0.04	0.80	1.083*	1.657**	0.74
Strategic resources (z-score)	(0.09)	(0.08)	(0.03)	(0.04)	(0.07)	(0.06)	(0.03)	(0.04)	(0.79)	(0.64)	(0.66)	(0.48)
Pre-test control*			1.161***	1.132***			0.925***	0.894***			0.413***	0.425***
Pre-test control*			(0.02)	(0.02)			(0.04)	(0.04)			(0.03)	(0.03)
Adjusted R-squared	0.00	0.25	0.80	0.80	0.01	0.20	0.77	0.77	0.00	0.17	0.50	0.59
Student observations	2541	2541	2379	2379	599	599	599	599	589	589	589	589
Controls
Individual and home		X		X		X		X		X		X
Province		X		X		X		X		X		X
School		X		X		X		X		X		X

Source: Leadership for literacy dataset, 2017.

Notes: Standard errors are in parentheses and clustered at the school level. Significant at *10% level, **5% level, ***1% level.

Each cell represents a regression of each individual index on the outcome variable.

The pre-test control in estimating A) English literacy scores and B) English ORF scores is the students’ pre-test results on the same tests. The pre-test control in estimating C) African language ORF is the student’s pre-test English ORF score because no pre-test is available in African language ORF. In estimation C) a control is also added to reflect the African language in question. For a description of individual, home and school controls see the discussion under ‘analytical strategy’.

WCPM = words read correctly per minute.

Table 5.

Estimating grade 3 literacy outcomes.

	Gr. 3 English ORF (% of WCPM)				Gr. 3 African language ORF (% of WCPM)
	Levels		Value-added		Levels		Value-added
	Model 1	Model 2	Model 1	Model 2	Model 1	Model 2	Model 1	Model 2
Material resources: time (z-score)	2.45	1.840**	1.338**	1.330**	3.708**	3.037**	2.040**	2.018**
Material resources: time (z-score)	(1.64)	(0.89)	(0.64)	(0.55)	(1.41)	(1.30)	(0.72)	(0.72)
Pre-test control*			1.084***	0.989***			1.040***	0.992***
Pre-test control*			(0.03)	(0.03)			(0.03)	(0.04)
Adjusted R-squared	0.01	0.32	0.78	0.80	0.02	0.18	0.73	0.74
Material resources: text presence(z-score)	−3.063**	−0.51	−0.11	0.47	−1.45	−0.50	0.55	0.75
Material resources: text presence(z-score)	(1.32)	(0.91)	(0.61)	(0.62)	(1.35)	(1.53)	(0.85)	(0.92)
Pre-test control*			1.087***	0.993***			1.048***	0.997***
Pre-test control*			(0.03)	(0.03)			(0.03)	(0.04)
Adjusted R-squared	0.02	0.31	0.78	0.79	0.00	0.17	0.73	0.74
Material resources: text use (z-score)	−3.407**	−1.741**	−0.33	0.23	−1.84	−2.777**	0.19	0.14
Material resources: text use (z-score)	(1.46)	(0.79)	(0.60)	(0.54)	(1.60)	(1.37)	(0.78)	(0.83)
Pre-test control*			1.085***	0.993***			1.047***	0.997***
Pre-test control*			(0.03)	(0.03)			(0.03)	(0.04)
Adjusted R-squared	0.03	0.32	0.78	0.79	0.00	0.17	0.73	0.74
Knowledge resources (z-score)	−0.66	−1.516*	0.01	−0.21	−1.48	−2.625**	−0.09	−0.34
Knowledge resources (z-score)	(1.81)	(0.86)	(0.62)	(0.52)	(1.75)	(1.27)	(0.76)	(0.71)
Pre-test control*			1.088***	0.990***			1.046***	0.995***
Pre-test control*			(0.03)	(0.03)			(0.03)	(0.04)
Adjusted R-squared	0.00	0.32	0.78	0.79	0.00	0.18	0.73	0.74
Human resources (z-score)	2.913*	−0.01	0.28	−0.33	1.80	−0.11	0.17	−0.12
Human resources (z-score)	(1.57)	(0.84)	(0.64)	(0.56)	(1.66)	(1.28)	(0.78)	(0.73)
Pre-test control*			1.086***	0.992***			1.046***	0.996***
Pre-test control*			(0.03)	(0.03)			(0.03)	(0.04)
Adjusted R-squared	0.02	0.31	0.78	0.79	0.00	0.17	0.73	0.74
Strategic resources (z-score)	3.124**	2.485**	−0.63	−0.75	4.312**	3.872**	−0.63	−0.81
Strategic resources (z-score)	(1.55)	(1.05)	(0.49)	(0.52)	(1.51)	(1.66)	(0.63)	(0.73)
Pre-test control*			1.096***	0.999***			1.051***	1.002***
Pre-test control*			(0.03)	(0.03)			(0.03)	(0.04)
Adjusted R-squared	0.02	0.32	0.78	0.79	0.02	0.18	0.73	0.74
Student observations	631	631	631	631	631	631	631	631
Controls
Individual and home		X		X		X		X
Province		X		X		X		X
School		X		X		X		X

Source and notes: See Table 4. In estimations of African language outcomes, we control for the language in question.

In Table 4 we estimate three grade 6 post-test outcomes: English literacy test scores expressed in z-scores; English ORF scores expressed in z-scores; and African language ORF scores expressed as the percentage of words read correctly per minute (% of WCPM). Table 5 shows estimates for two grade 3 literacy post-test outcomes: English and African language ORF (% of WCPM).

In each table we only show the coefficient estimate on each index entered individually in the regressions. We also show coefficients on pre-test literacy outcomes in the value-added models. Model 1 reflects the raw correlation between each index and literacy outcomes (i.e. excludes controls) while model 2 includes individual, classroom and school-level controls.

The limited explanatory power of the indices in estimating grade 6 literacy outcomes is evident in the low R-squared values in Table 4, model 1. Most of the time, no significant coefficients are observed on any of the indices. An exception is the significant and positive coefficient of 0.1 standard deviations on the human resources index in the fully controlled, value-added estimations of grade 6 literacy outcomes.

Returning to the issue of reliability, the exclusion of self-reported data in index construction does not significantly improve the predictive validity of the indices. We tested this by re-running all estimations in Tables 4 and 5 but using indices constructed from observational or evidence-based data only (results are shown in supplementary tables). We also find no evidence that the effect of the indices is mediated through the presence of knowledge resources.

Each rubric element is also included individually in estimations as they may be predictive of outcomes as stand-alone measures rather than in aggregation. A striking result is how few of the 114 rubric elements are individually predictive of selected literacy outcomes. Only 13% of the rubric elements are positively and significantly associated with grade 6 English literacy outcomes in the fully controlled model – only 12% if we include pre-test scores. Roughly 10% of all elements are statistically significantly negatively associated with each of the grade 3 and 6 literacy outcomes (see the supplementary tables for results). While unexpected, the identification of negative associations is not uncommon in the literature on ‘leadership effects’ (Robinson et al., 2008: 655).

Multivariate estimation of intermediate outcomes

Our ‘leadership for literacy’ indices yield stronger and more consistent associations with intermediate outcomes than with student outcomes, as seen in Table 6. The human resources as well as the strategic resources indices are positively associated with indicators of curriculum coverage. A one standard deviation increase in the strategic resources index, for example, is associated with an 8% point increase in the percentage of teachers who report their curriculum coverage is checked biweekly by a head of department (middle manager), and a 15% point jump in the amount of work in students’ exercise books and workbooks (expressed in centiles).

Table 6.

Estimating intermediate outcomes.

		Coverage of work in exercise / workbooks (in centiles)			% of utilised classrooms with present teacher and students engaged in learning activity			% of educators indicating their curriculum coverage is checked biweekly			Index of teacher engagement (z-score)
		Coef.	P-value	R2	Coef.	P-value	R2	Coef.	P-value	R2	Coef.	P-value	R2
Material resources: time (z-score)	PCA	2.90	0.47	−0.03	1.47	0.52	0.02	2.56	0.17	0.51	0.06	0.14	0.24
Material resources: text presence (z-score)	PCA	5.18	0.35	−0.02	−0.50	0.91	0.02	−0.08	0.98	0.50	0.01	0.81	0.21
Material resources: text use (z-score)	PCA	−10.91	0.00	0.10	2.79	0.36	0.04	−0.69	0.81	0.50	−0.07	0.19	0.24
Knowledge resources (z-score)	PCA	−2.52	0.57	−0.03	2.17	0.44	0.03	−1.21	0.62	0.50	0.02	0.62	0.21
Human resources (z-score)	PCA	7.04	0.08	0.01	−0.09	0.97	0.02	5.28	0.10	0.54	0.20	0.00	0.45
Strategic resources (z-score)	PCA	15.41	0.00	0.18	−3.36	0.12	0.04	7.60	0.01	0.58	0.21	0.00	0.43
SLM all combined (z-score)	Mean	8.52	0.03	0.04	0.25	0.93	0.02	2.84	0.21	0.51	0.16	0.00	0.37
Controls
Student characteristics		X			X			X			X
School characteristics		X			X			X			X
Province		X			X			X			X

Notes: N = 60 for all regressions. Each cell represents a regression of each individual index and the outcome variable.

Average student characteristics of grade 6 class include % overage, % who attended grade R, % who always or almost always speak English at home, % whose parents are not employed, % with own story books at home. School controls include average school wealth, grade 6 class size, English LOLT and low-fee paying. Cells are highlighted where p-values are less than 0.1.

There is also evidence that the allocation and deployment of the six resources may matter for the extent to which teachers enjoy their work and feel connected to their jobs and the school organisation. However, these findings are treated with caution because one would expect higher teacher engagement levels to be met with evidence of more teaching in the classroom. This is not the case.

Discussion

This paper described a process to construct composite measures of school leadership and management that may be linked to literacy outcomes in township and rural South African schools. Centred on a ‘leadership for literacy’ theoretical framework and in-depth qualitative work, a pre-determined rubric guided the SLM assessment approach. Despite giving more attention to quantifying SLM than has previously been attempted in South Africa, we were unable to reject a null hypothesis that SLM has a direct connection with literacy outcomes.

Only one of six ‘leadership for literacy’ dimensions was significantly positively associated with grade 6 literacy outcomes. That is, the human resource dimension, which quantifies the extent to which leaders develop, recognise and deploy expertise to build a literacy environment. Similarly, at the grade 3 level, only one resource dimension – the efficient use and allocation of time for reading and language – has a significant but small positive association with grade 3 reading outcomes. But somewhat stronger linkages are found with intermediate outcomes. Schools that more effectively deploy and utilise human or strategic resources have more evidence of curriculum management and teacher work engagement. But the overall finding is one of few associations between six ‘leadership for literacy’ dimensions and literacy outcomes.

Most of the time, the expected relationships are not found. This echoes earlier reflections in South African school effectiveness research (Gustafsson, 2007; Van Staden and Howie, 2014), and challenges prior beliefs that leadership and management accounts for the lion’s share of unexplained variation in learner performance across historically disadvantaged South African schools (see Crouch and Mabogoane, 1998).

Despite these results, the study highlights the limits and challenges of quantifying SLM in these contexts which may constrain the identification of effects. Reliability and validity concerns still taint our quantitative data collection process. Close-ended questions were used to limit high-level judgements that characterise semi-structured interviews. While this improves the scalability of the approach – reducing the need for experienced researchers – it reduces the opportunity to probe through socially desirable responses. The case studies revealed numerous instances where initial responses in interviews were biased through social desirability concerns, with respondents telling the interviewer what they believed the latter wanted to hear, or what they perceived to be accurate (Mertler, 2019). Without effective probing, these socially acceptable responses bias the data and reduce the detectability of variation in SLM proficiency across schools. Incorporating triangulation into the measurement approach was insufficient to compensate for these reliability concerns. Even observational data did not significantly improve predictive validity over measures derived from self-reported data. We argue that substituting fieldworkers for experienced researchers to collect quantitative (not just qualitative) data on SLM is a sub-optimal approach, limiting the identification of variation in SLM outcomes.

Additionally, limited variation in literacy outcomes across the rural and township school sample may constrain the detection of associations. This occurs despite exhausting available opportunities to add more variation through a purposeful sampling approach. In future studies, it may help to extend sample sizes both vertically and horizontally – adding more schools and classes or grades within schools. Notably, we found that teachers’ perceptions and experiences of their managers and school leaders varied more within than across schools (Taylor et al., 2019 ). More predictive power may also be gained through conceptualising SLM with additional resource dimensions or using another subject lens, such as mathematics.

Notwithstanding the merits of advancing research efforts to improve the detectability of SLM effects in South African schools, this is subsidiary to the real issue of building SLM capacity and leadership content knowledge. Particular attention needs to be given to developing the literacy knowledge of leaders to aggressively advance the teaching of reading in schools and support the implementation of coherent reading programmes (Murphy, 2004). In the qualitative findings from eight case study schools, weak practices in all ‘leadership for literacy’ domains were typically observed even in the best performing schools in the sample (Taylor et al., 2019). Where good leadership and management practices were discerned in the deployment of one resource area, this was juxtaposed against weaknesses in how one or more of the other resources were deployed. Furthermore, where material, human and strategic resources were effectively deployed this was strongly mediated through the presence of knowledge resources of not only leaders, but teachers. Addressing the South African literacy crisis is unlikely to happen in isolation of hiring and developing leaders and teachers with the requisite skills to advance literacy knowledge in schools.

Supplemental material

Supplementary_Tables_EMAL_SLM_in_SA - Measuring school leadership and management and linkages with literacy: Evidence from rural and township primary schools in South Africa

Supplementary_Tables_EMAL_SLM_in_SA for Measuring school leadership and management and linkages with literacy: Evidence from rural and township primary schools in South Africa by Gabrielle Wills and Servaas van der Berg in Educational Management Administration & Leadership

Footnotes

Acknowledgements

We are grateful for the collaborative work of the ‘Leadership for literacy’ study team including Nick Taylor, Ursula Hoadley, Jaamia Galant, Francine de Clercq, Nic Spaull, Nompumelelo Mohohlwane and Lilli Pretorius, and especially David Carel for overseeing quantitative data collection. Thank you to Marie-Louise Schreve, Carine Brunsdon and Silke Rothkegl-Van Velden for their administrative support. We also acknowledge the efforts of fieldworkers, data capturers and test markers.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research is funded by the Economic and Social Research Council [grant ES/N01023X/1].

ORCID iD

Gabrielle Wills

Supplemental material

Supplemental material for this article is available online.

Notes

References

Arcia

Macdonald

Patrinos

Porta

(2015) School Autonomy and Accountability (No. 9), System assessment and benchmarking for education results, systems approach for better education results (SABER working paper series), Washington, DC.

Bloom

Van Reenen

(2010) New approaches to surveying organizations. The American Economic Review 100: 105–109.

Bloom

Lemos

Sadun

Van Reenen

(2015) Does management matter in schools? The Economic Journal 125: 647–674.

Branch

Hanushek

Rivkin

(2012) Estimating the Effect of Leaders on Public Sector Productivity: The Case of School Principals (No. 17803), National Bureau of Economic Research Working Paper Series, Cambridge, MA. doi:10.3386/w17803.

Burch

Spillane

(2003) Elementary school leadership strategies and subject matter: Reforming mathematics and literacy instruction. Elementary School Journal 103(5): 519–535.

Bush

Glover

(2016) School leadership and management in South Africa: Findings from a systematic literature review. International Journal of Educational Management 30: 211–213.

Camburn

Spillane

Sebastian

(2010) Assessing the utility of a daily log for measuring principal leadership practice. Educational Administration Quarterly 46: 707–737.

Carnoy

Ngware

Oketch

(2015) The role of classroom resources and national educational context in student learning gains: Comparing Botswana, Kenya, and South Africa. Comparative Education Review 59: 199–233.

Condon

Clifford

(2012) How rigorous are commonly used principal performance assessment instruments? A quality school leadership issue brief. Revised. Washington, DC: American Institutes for Research.

10.

Crawfurd

(2017) School management and public–private partnerships in Uganda. Journal of African Economies 26: 539–560.

11.

Crouch

Mabogoane

(1998) When the residuals matter more than the coefficients: An educational perspective. The Review of Economic Studies 22: 1–14.

12.

Crouch

Mabogoane

(2001) No magic bullets, just tracer bullets: The role of learning resources, social advantage and education management in improving the performance of South African schools. Social Dynamics 27: 60–78.

13.

Department of Basic Education (DBE) (2016) Personnel Administrative Measures (PAM) Government Gazette No. 39684. 12 February 2016. Pretoria.

14.

Fryer

(2017) Management and student achievement: Evidence from a randomized field experiment (no. 23437), NBER Working Paper Series, Cambridge, MA.

15.

Goldring

Cravens

Murphy

Porter

Elliott

Carson

(2009) The evaluation of principals: What and how do states and urban districts assess leadership? The Elementary School Journal 110: 19–39.

16.

Grissom

Kalogrides

Loeb

(2015a) Using student test scores to measure principal performance. Educational Evaluation and Policy Analysis 37: 3–28.

17.

Grissom

Loeb

Mitani

(2015b) Principal time management skills: Explaining patterns in principals’ time use, job stress and perceived effectiveness. Journal of Educational Administration 53: 773–793.

18.

Gustafsson

(2005) The relationships between schooling inputs and outputs in South Africa: Methodologies and policy recommendations based on the 2000 SACMEQ dataset, SAQMEQ paper series, Botswana.

19.

Gustafsson

(2007) Using the hierarchical linear model to understand school production in South Africa. South African Journal of Economics 75: 84–98.

20.

Hallinger

(2008) Methodologies for studying school leadership: A review of 25 years of research using the Principal Instructional Management Rating Scale. Paper presented at the annual meeting of the American Educational Research Association, New York.

21.

Hallinger

(2017) Surfacing a hidden literature: A systematic review of research on educational leadership and management in Africa. Educational Management Administration & Leadership 46: 362–384.

22.

Hallinger

Heck

(1998) Exploring the principal’s contribution to school effectiveness: 1980–1995. School Effectiveness and School Improvement 9: 157–191.

23.

Heck

Hallinger

(2005) The study of educational leadership and management: Where does the field stand today? Educational Management Administration & Leadership 33: 229–244.

24.

Hoadley

(2018) Leading for literacy: A review of the research. Research on Socio-economic Policy Working Paper No. WP19/2018, Stellenbosch University.

25.

Hoadley

Christie

Ward

(2009) Managing to learn: Instructional leadership in South African secondary schools. School Leadership 29: 373–389.

26.

Horng

Klasik

Loeb

(2009) Principal Time-Use and School Effectiveness (School Leadership and Research Report No. 09-3). Stanford, CA.

27.

Howie

Combrink

Roux

Tshele

Mokoena

McLeod Palane

(2017) PIRLS Literacy 2016: South African Highlights Report. Pretoria.

28.

Kotze

(2017) Social Gradients, Early Childhood Education and Schools Performing Above the Demographic Expectation: Empirical Insights Into Educational Issues. Stellenbosch: Stellenbosch University.

29.

Lee

Hallinger

(2012) National contexts influencing principals’ time use and allocation: Economic development, societal culture, and educational system. School Effectiveness and School Improvement 23: 461–482.

30.

Lee

Zuze

(2011) School resources and academic performance in sub-Saharan Africa. Comparative Education Review 55(3): 369–397.

31.

Leithwood

Louis

Anderson

Wahlstrom

(2004) Review of Research. How Leadership Influences Student Learning, Learning from Leadership Project. New York: The Wallace Foundation.

32.

Lemos

Scur

(2017) Developing Management: An expanded evaluation tool for developing countries (No. 16/007), RISE Working Paper Series, London.

33.

Marzano

Waters

McNulty

(2005) School Leadership That Works: From Research to Results. Aurora: ASCD and McREL.

34.

McCullough

Lipscomb

Chiang

Gill

Cheban

(2016) Measuring school leaders’ effectiveness: Final report from a multiyear pilot of Pennsylvania’s framework for leadership (REL 2016–106), Washington, DC.

35.

Mertler

(2019) Introduction to Educational Research. Los Angeles: SAGE.

36.

Mullis

IVS

Martin

Goh

Prendergast

(eds) (2017) PIRLS 2016 Encyclopedia: Education policy and curriculum in reading. Retrieved from Boston College, TIMSS & PIRLS International Study Center. Available at: http://timssandpirls.bc.edu/pirls2016/encyclopedia/ (accessed July 2018).

37.

Murphy

(2004) Leadership for literacy: A framework for policy and practice. School Effectiveness and School Improvement 15(1): 65–69.

38.

Peters

(2014) The alpha and the omega of scale reliability and validity. European Health Psychologist 1: 54–72.

39.

Pretorius

Machet

(2004) Literacy and disadvantage: Learners’ achievements in the early primary school years. Africa Education Review 1(1): 128–146.

40.

Robinson

VMJ

Lloyd

Rowe

(2008) The impact of leadership on student outcomes: An analysis of the differential effects of leadership types. Educational Administration Quarterly 44: 635–674.

41.

Scur

(2017) Organization Matters: The Causes and Consequences of Organizational Choices in the Private and Public Sectors. Oxford: University of Oxford.

42.

Shepherd

(2016) The Open Door of Learning – Access Restricted: School Effectiveness and Efficiency Across the South African Education System. Stellenbosch: Stellenbosch University.

43.

Spaull

(2013) Poverty and privilege: Primary school inequality in South Africa. International Journal of Educational Development 33: 436–447.

44.

Spillane

Halverson

Diamond

(2004) Towards a theory of leadership practice: A distributed perspective. Journal of Curriculum Studies 36(1): 3–34.

45.

Stein

Nelson

(2003) Leadership content knowledge. Educational Evaluation and Policy Analysis 25: 423–428.

46.

Tavares

(2015) The impact of school management practices on educational performance: Evidence from public schools in São Paulo. Economics of Education Review 48: 1–15.

47.

Taylor

Prinsloo

(2005) The Quality Learning Project Lessons for High School Improvement in South Africa. Johannesburg: JET Education Services.

48.

Taylor

Wills

Hoadley

(2019) Addressing the ‘leadership conundrum’ through a mixed methods study of school leadership for literacy. Research in Comparative and International Education 14(1): 30–53.

49.

Taylor

Gamble

Spies

Garisch

(2013) Chapter 4: School leadership and management. In: Taylor

van der Berg

Mabogoane

(eds) Creating Effective Schools: Report of South Africa’s National School Effectiveness Study. Cape Town: Pearson Education.

50.

Taylor

(2011) Uncovering indicators of effective school management in South Africa using the national school effectiveness study (no. 10/11), Stellenbosch Economic Working Papers, Stellenbosch.

51.

Van der Berg

(2008) How effective are poor schools? Poverty and educational outcomes in South Africa. Studies in Educational Evaluation 34: 145–154.

52.

Van der Berg

Louw

(2006) Unravelling the mystery: Understanding South African schooling outcomes in regional context. Paper presented at the Conference for the Study of African Economics, Oxford University.

53.

Van Staden

Howie

(2014) Reflections on Creemers’ comprehensive model of educational effectiveness for reading literacy: South African evidence from PIRLS 2006. Perspectives in Education 32: 172–192.

54.

Wills

(2017) What do you mean by ‘good’? The search for exceptional primary schools in South Africa's no-fee school system. Stellenbosch Economics Working Paper Series No16/2017, Stellenbosch University. Available at: https://www.ekon.sun.ac.za/wpapers/2017/wp162017 (accessed August 2018).

55.

Wills

(2019) A quantitative perspective on school leadership and management in South Africa. In: Spaull

Jansen

(eds) South African Schooling: The Enigma of Inequality. A Study of the Present Situation and Future Possibilities. Berlin: Springer International, 301–320.

56.

Witziers

Bosker

Krüger

(2003) Educational leadership and student achievement: The elusive search for an association. Educational Administration Quarterly 39: 398–425.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.17 MB