What difference does school leadership make to student outcomes?

ABSTRACT

Although it is widely believed by parents and policy makers that the quality of school leadership makes a difference to what students learn at school, testing this hypothesis has proved very complex. Part of the problem is the unreliability of many measures of leadership quality. In addition, the field has been dominated until quite recently by abstract theories of leadership which are not closely aligned to the specific work of educational leaders. While recent research by economists has confirmed that school leadership makes a significant difference to student outcomes, their analyses beg the question of how these effects are produced. In this article we address this question by updating the first author’s 2008 model of student-centred leadership which identified the average impact of five sets of leadership practices on student outcomes. We argue that future research programmes should develop theories and measures of educational leadership which are based on expert judgments of the quality of enactment of these five sets of practices, and test their impact on a wider range of student outcomes.

KEYWORDS:

• Educational leadership
• leadership effects
• school principals
• school improvement
• instructional leadership
• student-centred leadership
• transformational leadership

Parents, policy makers and politicians are convinced that school leaders, particularly principals, make a difference to the social and academic outcomes of the learners for whom they are responsible. Stories of new principals who succeed in turning around poor performing schools reinforce this belief (Murphy 2008; Duke and Salmonowicz 2010), as does our shared tendency to attribute the performance of all organisations, not just educational ones, to the quality of its leadership (Meindl 1998). Yet, paradoxically, it has proved extremely difficult to empirically test the difference that the quality of school leadership makes to the learning of students. While it is now generally accepted by leadership researchers that leaders do matter (Leithwood et al. 2008; Leithwood et al. 2010; Grissom and Loeb 2011), the estimates of the size of their average effect on student outcomes vary from miniscule to quite large (Marzano et al. 2005; Robinson et al. 2008).

In this paper we review research which has investigated links between student outcomes and those who hold formal leadership positions. While some studies include all of a school’s senior leaders in their leadership sample, the majority confine their assessments to the school principal. In the former case we refer to school-wide leadership and in the latter to principalship. While we recognise that leadership may also be exercised by those who do not hold formal leadership positions, there is scant evidence about the impact of distributed leadership on student outcomes and it is that impact which is the focus of this article.

There are three major reasons for the difficulty in researching the links between school leadership and student outcomes. The first concerns the difficulty of measuring the relevant variables. Gaining reliable data on the quality of school leadership, based on observations of leadership practices or detailed behavioural interviews, is very expensive. Cheaper methods, such as self-assessment ratings, typically produce high scores and limited variance (Grissom and Loeb 2011; Di Liberto et al. 2015) and third-party ratings by teachers may also be subject to various sorts of perceptual biases such as halo effects (Favero et al. 2016). Some researchers use measures of the frequency of certain leadership behaviours, such as observations of teaching, on the assumption that greater frequency means higher quality leadership, when, in fact, there is little or no correlation between the two (Di Liberto et al. 2015). All such measurement errors reduce the chance of finding leadership effects.

Qualification rates and results on standardised tests of literacy and math are the predominant student outcome measures employed. These academic outcome measures do not cover the full social and emotional outcomes that are also valued in most school systems. We acknowledge the paucity of studies linking social, emotional and cultural student outcomes to leadership quality as a limitation of the current body of research.

Statistical analyses of leadership effects on student outcomes require separation of the effects of leadership from the effects of other confounding variables such as differences in students’ socio-economic background, school composition, and prior achievement levels. In the United States, federal, state and school district administrations take responsibility for the collection of standardised and repeated assessments on many of these variables, and the resulting data sets are made available, under certain conditions, to researchers. The New Zealand Ministry of Education takes no responsibility for the creation and use of such data sets, and this is one major reason why there are no large-scale New Zealand studies of the impact of any school-level variable, including leadership, on student outcomes.

A second reason for the difficulty in researching links between leadership and student outcomes is that, unless they are teaching principals in very small schools, leaders influence learning indirectly through the work of their teachers. The role of the senior leadership team is to create the conditions that enable their teachers to teach in ways that ensure adequate levels of student growth on important learning outcomes. Currently, there is considerable debate about which conditions are the most important and about the pathways through which they are established One argument is that, since the quality of teaching and teachers is the most important school-level influence on student learning (Rivkin et al. 2005; Hattie 2009), leaders’ primary focus should be on the improvement of the quality of teaching and the curriculum. While there is good evidence for the importance of such ‘instructional’ leadership (Robinson et al. 2008) there is also empirical support for leadership effects that flow from a focus on the quality of school organisation and management (Grissom and Loeb 2011; Di Liberto et al. 2015) and from the creation of school cultures in which trust and collective responsibility for students’ learning is high (Louis and Murphy 2017). Rather than test the effects of separate pathways, some researchers are now developing integrated models which test the relative effects of multiple paths through which leaders can exercise influence (Leithwood et al. 2010; Li et al. 2016). Such multivariate studies should yield more robust theories about the combination of conditions, such as features of school organisation, the teaching programme, family relationships and school culture, that are needed to improve student outcomes.

A third reason for the difficulty of linking leadership to student outcomes is the time it takes for the effects of leadership decisions to be felt. For example, a principal’s decision to introduce teacher professional learning communities, requires major changes in timetabling so teachers can meet. These organisational changes must then be accompanied by changes in teacher culture, so that meetings in which evidence-based discussion of the connections between teaching practice and student learning becomes the norm. Assuming such interaction is established, teachers must then apply their learning to their own teaching in ways that improve the learning of their students, and, only then, will the impact of the leader’s original decision on student outcomes be discernible. In a longitudinal study of principal effects, Coelli and Green (2012) estimated that it may take as many as five years for the full effect of a principal to be evident in student outcome data. This means that studies which use correlational designs (i.e. they measure leadership quality and student outcomes at one point in time), may underestimate leadership effects. While longitudinal research designs are now more common (May et al. 2012; Heck and Hallinger 2014), they bring their own set of challenges, not the least of which is the high turnover of school leaders, particularly in economically-disadvantaged communities (Coelli and Green 2012).

Despite these difficulties, leadership researchers have now been able to confirm what parents and politicians suspected all along – the quality of school leadership really does matter (Grissom and Loeb 2011; Coelli and Green 2012). Perhaps the most compelling evidence comes from economic analyses which have eschewed subjective measures of leadership quality and relied instead on longitudinal designs which track the relationship between shifts in student outcomes and principal turnover and use rigorous statistical controls over potentially confounding background variables (Coelli and Green 2012; Branch et al. 2013; Bloom, Remos et al. 2014; Dhuey and Smith 2014; Böhlmark et al. 2016). While these analyses tell us that principals make a difference, they do not tell us how they make a difference and it is to this question we now turn.

From leadership styles to leadership practices

Two leadership theories, describing distinctive styles of leadership, have dominated research on the impact of school leaders on student outcomes. The first, transformational leadership, has its origins in James McGregor Burns’s work on leader-follower relations (Burns 1978). Burns was interested in how some leaders were able to motivate followers to move beyond self-interest to pursue the larger goals of the group or organisation. Burns’s theory was developed further in the nineteen eighties and nineties by Bass and his colleagues (Bass 1985; Bass and Avolio 1994). They argued that transformational leaders inspired their followers through four influence processes: giving personal attention to individual staff so they feel uniquely valued (individualised consideration); encouraging creativity and new ways of thinking about old issues (intellectual stimulation); communicating optimism and high expectations (inspirational motivation); and providing a vision and sense of purpose which elicits trust and respect from followers (idealised influence).

The early versions of transformational leadership theory were solely concerned with leader-follower relations and, as such, applicable to leaders in any type of organisation. Partly in response to criticism that the theory lacked content related to the specifics of the educational leadership role, educational researchers added constructs, such as instructional support, that reflected some of the unique features of the role (Leithwood and Jantzi 2005). Individual studies of the relationship between transformational leadership and student outcomes have yielded very mixed results (Leithwood and Jantzi 2005) with meta-analyses showing the average effect size is small (Robinson et al. 2008; Sun and Leithwood 2012).

Instructional leadership is the second theory that dominates research on leadership and student outcomes. Unlike transformational leadership, instructional leadership theory is firmly grounded in educational work. While there are variations in the concept, the common themes are the close involvement of leadership in establishing an academic mission, monitoring and providing feedback on teaching and learning, and promoting the importance of professional development (Alig-Mielcarek and Hoy 2005; Hallinger 2005).

Instructional leadership has its origins in the early eighties in studies of successful schools in poor urban communities. Such schools usually had strong instructional leadership which established a learning climate free of disruption, a system of clear teaching objectives, and high teacher expectations for students (Bossert et al. 1982).

As for studies on transformational leadership, individual studies of the relationship between instructional leadership and student outcomes yield widely varying effect sizes. In the first author’s meta-analyses, completed for the Ministry of Education’s Best Evidence Synthesis programme, Robinson estimated the average effect of instructional leadership to be three to four times that of transformational leadership (Robinson et al. 2009). The results of some subsequent studies, however, have not confirmed the importance of instructional leadership (Grissom and Loeb 2011), or have argued that it is the integrated mix of transformational and instructional leadership styles that is most powerful (Day et al. 2016).

Despite several decades of research on theoretically-derived leadership styles, there is still no consensus about their relative impacts on student outcomes. That may be why educational leadership researchers are increasingly studying the impact of leadership practices and turning to analyses of the work, rather than abstract formulations of leadership style, (Grissom and Loeb 2011; May et al. 2012; Di Liberto et al. 2015).

If leadership theory and research is firmly grounded in the work, one quickly recognises its holistic and context-specific nature. Academic debates that set up oppositions between transformational and instructional leadership, for example, are unhelpful because they partition aspects of leadership which must be integrated. The job of educational leaders is not only to inspire and stimulate followers (their staff) by creating a compelling vision and communicating high expectations. This must be done in ways that serve the primary responsibility of the role, which is to ensure adequate growth in the learning of their students. Leaders, in other words, must build relationships with their staff in ways that maintain a focus on the core purpose and on the well-being and learning of students. Integrating the pursuit of this increasingly complex and specialised educational work while building relationships of trust is the job of educational leaders. What do we know about the practices that enable leaders to succeed in that integrated work?

The impact of leadership practices

The role of the principal is vast in scope, highly demanding and time pressured (Riley 2015). It is important, therefore, that leaders give priority to those practices that are likely to make the most impact on student learning and well-being. Some guidance about what those practices are is available from meta-analyses of published research from which the impact of different practices can be derived. We organise the remainder of this section according to the five dimensions of practice that the first author derived from a meta-analysis of such research (Robinson et al. 2008). The published studies included in the meta-analysis measured principal or school-wide leadership, were conducted in primary and secondary schools, and used student qualification rates and scores on standardised literacy and maths assessments as measures of student outcome (Robinson et al. 2008). Robinson called the five dimensions of practice, together with their associated knowledge and skill capabilities, student-centred leadership (Robinson 2011). The model has been independently evaluated in 127 Queensland Catholic schools (Avenell 2015). Results showed a strong relationship between scores on each of the five leadership dimensions and school performance data derived from a national standardised assessment of literacy and numeracy.

In the following we expand on the practices involved in each of the five dimensions, and include relevant research published since completion of the original meta-analysis. To summarise the model, in schools where students achieve well above expected levels, the leadership looks quite different from the leadership in otherwise similar lower performing schools. In the higher performing schools, it is much more focused on the business of improving learning and teaching, through quality enactment of goal-setting (Dimension 1), strategic resourcing of those priority goals (Dimension 2) and ensuring the quality of teaching (Dimension 3). In ensuring such quality, leaders learn what they and their teachers need to learn to achieve their priority goals. Leaders in high performing schools will not only organise and resource teacher professional learning and development but are more likely to lead it (Dimension 4). The fifth dimension of student-centred leadership (ensuring an orderly and safe environment) provides a foundation for all the rest.

Setting goals and expectations (dimension 1)

This dimension includes deciding what goals to set, gaining the commitment of those responsible for achieving them, and communicating them to all those with an interest in their achievement. The quality of goal-setting cannot be separated from the quality of relationships. Leaders can set goals, but they will remain empty words unless they motivate those whose efforts are required to achieve them. This dimension of leadership has, on average, a small to moderate impact on student outcomes (Robinson et al. 2008). It works indirectly, by focusing and coordinating the work of adults around promoting the learning and achievement of students.

In schools where teachers report strong goal-setting activity by leaders, students will, on average, achieve significantly more than in otherwise similar schools (Robinson et al. 2008). The leaders of higher performing schools give more emphasis than do their counterparts in lower performing schools, to communicating goals and expectations, informing the community of academic accomplishments and recognising academic achievement. In higher performing schools, teachers also report that there is more consensus about school goals.

The challenge in this dimension is not just to set goals, for nearly all school leaders are required to do this as part of their school’s strategic and annual planning and for their own professional development planning. Rather, the challenge is to set them in ways that meet the conditions required to make goal-setting work (Locke and Latham 1990; Latham and Locke 2006, 2007). Goal theory tells us that goals will increase performance when they are specific, when people are committed to them and when they believe they already have, or will gain, the capacity to achieve them (Latham and Locke 2006).

A more recent meta-analysis of the direct and mediated effect of goal-setting practices on student achievement showed that while these leadership practices impacted teacher empowerment and sense of efficacy, they did not flow through to student achievement (Sun and Leithwood 2015). Similarly, a study of a range of leadership practices conducted in a national sample of Italian high schools, did not show significant effects of goal-setting on math achievement (Di Liberto et al. 2015). In contrast, the quality of goal-setting and its communication had a significant positive impact on student attendance, achievement, and rates of violence in a large sample of New York schools (Favero et al. 2016).

The contrary evidence could be explained by examining the extent to which measures of leaders’ goal-setting capture the qualities that goal theory has identified as central to its effectiveness. That theory, and the long history of goal-setting research in social psychology (Latham and Locke 2006, 2007), suggest that school leaders’ goal-setting will drive improved student outcomes if it creates a motivating discrepancy between current reality and the desired future state. For student-centred leaders, the most important indicator of the current reality is evidence about the wellbeing and learning of students. Goals that are clear, specific, and perceived as challenging but attainable, focus attention and effort and motivate persistent goal-relevant behaviour. School-level goals are vertically integrated into the goals of subject departments, year-level teams and individual teachers so that the disparate contributions of groups and individuals are coordinated in ways that serve goal attainment (Robinson et al. 2017). These are the types of qualities that goal theory suggests should be measured in future studies of the impact of this leadership dimension on student outcomes.

Resourcing strategically (dimension 2)

Once clear goals are established, then Dimension 2 – resourcing strategically – comes into play. In Robinson’s 2008 meta-analysis, this dimension had a small average effect on student outcomes (Robinson et al. 2008). Leaders in high performing schools allocate scarce resources such as money, time on the timetable, teaching materials and instructional expertise in ways that give priority to key goals. For example, if a school has committed to the goal of improving boys’ reading, then that commitment should be reflected in such things as the focus of teachers’ professional development, the choice of texts and supplementary readers, and the criteria for hiring a new teacher of English. Staffing, teaching resources and teacher professional development are orchestrated in ways that increase the chance of achieving the goal.

From a leadership point of view the key research question is not how many resources are available, but how they are used (Miles and Frank 2008). Research on use requires investigation of the links between goals and resource acquisition and allocation, and these links are much more difficult to investigate than leaders’ access to resources. It is the alignment and coordination of resources with strategic priorities that is key to understanding how leaders drive improvement in student outcomes (Bryk et al. 2010).

Differences in principal discretion over resource acquisition and allocation further complicate the picture. Substantial governmental and union constraints in some countries on the hiring, promotion, remuneration and firing of teachers may account for the non-significant relationship between principals’ personnel management practices and student achievement in a large-scale cross-national study (Di Liberto et al. 2015). In contrast, New Zealand’s self-managing school system accords principals and boards considerable discretion over who they hire, reward and fire, but no research is available on the use and consequences of this discretion.

Perhaps the most important strategic resource to be managed is students’ time. Establishing a school culture in which students’ time is treated as a strategic resource is an important responsibility of school leadership, because achievement is a function of both the amount and the quality of time students spend in learning the curriculum (Sebastian and Allensworth 2012; Sebastian et al. 2017). Without quality, longer school days, summer schools and catch-up classes waste student time. Given adequate quality, however, increased time in a subject will lift achievement (Scheerens and Bosker 1997).

While goals specify what should be achieved by when and with whom, it is the strategic alignment of goals with student and teacher time, and relevant expertise and materials, that enable their achievement (Miles and Frank 2008). Principals in high performing schools make the tough decisions required to allocate and reallocate resources to support goal achievement. They also decline or postpone resources and opportunities, such as externally funded initiatives, that distract from their priority goals (Reid 2014). To be a strategic leader is to be clear and consistent about what the school organisation is and is not currently pursuing, and to resource accordingly.

Ensuring the quality of teaching (dimension 3)

In Robinson’s original meta-analysis, the average effect of this leadership dimension was a moderate .47. Since its publication in 2008, more recent individual studies have added to our understanding of the qualities which explain the power of the leadership practices that are included in this dimension. In the following, we focus on three sets of Dimension 3 practices: developing a coherent instructional framework, monitoring a variety of evidence about the impact of teaching on student outcomes and supporting teachers by visiting classrooms and providing feedback.

Developing a coherent instructional framework

A school has a coherent instructional framework if the curriculum and associated teaching strategies and assessments are coordinated within and between year levels. This ensures a progression of increasingly difficult subject matter and students’ access to the same content regardless of teacher assignment. Leaders foster coherence by supporting and requiring the use of the framework, and by its inclusion in staffing, resourcing and professional development policies and practices. Students achieve more in schools with more coherent instructional programmes (Newmann et al. 2001) and teachers develop stronger professional communities when they have a common approach to teaching and learning (Bryk et al. 2010).

Coherence works because it supports how students learn. Students learn and remember more when key ideas are presented in ways that connect with their prior knowledge and experience and when they are given multiple, varied opportunities to gain an in-depth understanding of new concepts (Bransford et al. 2000). We also know that exposure to multiple representations of the same idea over a relatively short period of time – for example a unit of work spanning ten days – promotes students’ learning (Nuthall and Alton-Lee 1993). Learning opportunities that meet these conditions are more likely to be found in teaching programmes that are planned around a progression of learning objectives and associated informal assessments. A coherent instructional framework means that teachers reinforce the same ideas, use similar vocabulary for communicating those ideas, know how to make links with what has gone before, and are guided in their efforts by common assessments. If learning opportunities are integrated and cumulative, rather than fragmented and rushed, students are more likely to be engaged and successful.

Monitoring and using evidence

A second set of leadership practices that sits under Dimension 3 involves the monitoring and use of evidence about a variety of student outcomes. Principals and senior leaders are responsible for monitoring progress on school-wide goals, and heads of department and team leaders for monitoring social and academic outcomes for their groups of students.

The quality of principals’ engagement in this work was found to be significantly correlated with student achievement in a cross-national study of the impact of principals’ management practices on math achievement (Di Liberto et al. 2015).

For the last two decades there has been a considerable emphasis in OECD countries, including New Zealand, on building the data management infrastructure and the capability required to support this work (Education Review Office 2018). The task is a complex one involving establishing organisational routines for collecting, storing, analyzing and reflecting on assessment or exam data so that improvements can be planned and subsequently evaluated (Halverson et al. 2007).

Structures and routines, however, are not enough to create the evidence-based inquiry and organisational and teacher learning that is required to make best use of data. In addition, leaders need to create a culture in which teachers trust the data, are willing to open up their practice for critical scrutiny and in which the expertise required to improve can be shared and developed (Marsh 2012; Hattie 2015; Datnow and Hubbard 2016).

Evaluating teachers and teaching

The third set of practices we consider under Dimension 3 leadership comprises leaders’ formal evaluation of teachers. This set of practices is closely related to the second, for teacher evaluations should be based, at least in part, on evidence about the impact of the teacher on their students’ learning and wellbeing.

Research on teacher evaluation is consistent in showing its negligible effects on the practice of evaluated teachers, whether evaluation practices are situated within high stakes accountability policies, as they are in England and the United States, or within more developmentally-oriented policies as they are in New Zealand (Sinnema and Robinson 2007; Murphy et al. 2013; Firestone 2014). Rather than improve teaching and student outcomes, high stakes accountability policies in which teachers are rewarded or sanctioned contingent on student test gains, have damaged teacher morale (Mintrop and Sunderman 2009) and narrowed the curriculum as teachers focus on examined subjects (Brown and Clift 2010). In addition, the measurement of teaching quality, whether through classroom observations of teaching or value-added measures of student growth, is not yet sufficiently valid to justify high stakes decisions about teacher performance (Campbell and Ronfeldt 2018).

While more developmentally-oriented teacher evaluation policies and practices avoid some of the harms of high stakes approaches, there is also little positive evidence of their impact on the quality of teaching and learning (Murphy et al. 2013). One reason is the reluctance of leaders to give negative feedback – a reluctance that produces a major mismatch between principals’ private evaluation of the proportion of their teachers who are unsatisfactory and their documented formal evaluations of those same teachers (Kraft and Gilmour 2017). Major reform efforts in the United States have invested heavily in new evaluation tools on the assumption that a key reason for the mismatch is the difficulty in providing rigorous evidence of the quality of teaching (Kraft and Gilmour 2017). A recent evaluation of this reform showed that, despite substantial investment in a suite of validated evaluation tools, the proportion of teachers receiving a less than satisfactory rating changed little (Kraft and Gilmour 2017).

The causes of the discrepancy between leaders’ public and private evaluations of their teachers reach far deeper than technical measurement problems. First, the incentive structure of educational personnel policies encourages leaders to stay silent about poor performance rather than to risk the considerable time, cost and stress that comes with tackling poor performance. The adult-centric nature of such policies means that the leader who addresses such issues can be sanctioned for acts of commission, that is for violating regulations about the treatment of adults, but there is no such sanction for acts of omission, that is, for failing to act in the interests of students by not conscientiously addressing perceived poor teacher performance.

Second, there is considerable evidence that many leaders struggle with communicating and discussing negative feedback even under benign conditions such as in a confidential and safe workshop (Sinnema et al. 2013). They unwittingly treat their feedback as true instead of fallible, and as complete instead of partial, and this construction creates a dilemma between pursuing the problem or maintaining the relationship (Robinson 2016). Unless pressured to do otherwise, most leaders respond to the dilemma by protecting the relationship and avoiding or downplaying the teaching problem. It is this dynamic that partially explains the discrepancy between some leaders’ private and formal evaluations of their teaching staff. In summary, the important research question is not ‘Does teacher evaluation improve teaching?’ but ‘What are the qualities of teacher evaluation that increase its impact on teaching practice and on the students of the evaluated teachers?’ Leaders who are accessible and who evaluate teachers as part of an ongoing supervision process (Tuytens and Devos 2017), who are judged as having relevant expertise, and who have a good relationship with those they evaluate, are likely to be perceived as having provided useful feedback (Tuytens and Devos 2011). Involvement of teachers in the planning of the evaluation also improves its perceived fairness and utility (Tuytens and Devos 2014). It should be noted, however, that these studies assess teachers’ perceptions of their evaluators and there is some evidence that their perceptions of feedback accuracy and utility may not match those of third-party experts (Rigby et al. 2017).

Leading teacher learning and development (dimension 4)

Of the five dimensions of student-centred leadership, the practices grouped under this dimension had the largest average effect size of all five dimensions (Robinson 2008). This finding makes sense because the quality of teaching is the most powerful school level influence on student outcomes (Muijs and Reynolds 2011), and leaders who focus on this dimension are more likely to improve teaching practice and student outcomes than those who do not. That likelihood is substantially increased, however, if leaders provide teacher learning opportunities that meet the conditions for effectiveness suggested by recent research on the impact of teacher professional learning on student outcomes.

Research suggests the following qualities of professional learning are associated with shifts in teaching practice that flow through to improvements in student achievement, well-being or engagement: the professional learning is anchored in evidence of a specific learning need of students; it focuses on the relationship between teaching and student outcomes, it provides worthwhile content; integrates theory and practice, uses external expertise and provides multiple opportunities to learn (Elmore 2002; Timperley and Alton-Lee 2008; Muijs et al. 2014). In the remainder of this section I give a brief overview of each of these aspects of effective professional learning.

In contrast to most professional learning, which starts by introducing teachers to new inputs, the starting point of effective teacher learning is inquiry into evidence about the achievement or experiences of their students. Once the focus of professional learning is determined, leaders need to make wise evidence-based choices about any programme, initiative or intervention that they introduce to their teachers. Education is dogged by fads and fashions, and it is no easy matter to select interventions that are evidence-based and well matched to student and teacher learning needs (Datnow 2000). Leaders can increase the chance of getting worthwhile content by asking potential providers for any evaluations of their programmes, or by doing their own research on websites which provide robust syntheses of the evidence about the effectiveness of different interventions.

Effective professional learning integrates theoretical principles and practical applications. Professional development comprising practical tips does not work because it leaves teachers without the knowledge of underlying principles that enables them to develop adaptive expertise and create the conditions in their own classrooms that are the key to improved student learning (Hammerness et al. 2005). On the other hand, theoretical content that is not linked to practical applications and rich illustration is also ineffective. When teachers complain about content being too theoretical, they are not asking for less theory, but for less theory that is disconnected from the classroom problems they are trying to solve. Effective professional development communicates clear theoretical principles and provides ample opportunity for participants to explore how they can be applied in their own contexts.

The use of external expertise is another feature of effective professional development. By external, I mean that the leadership of the group, whether a teacher, school leader, coach, facilitator or researcher, has demonstrated a greater capacity to solve or prevent the relevant teaching problems than the remaining group members. Expecting teachers who share similar difficulties to solve their problems without the help of such external expertise is unrealistic. When the professional learning takes place in an ongoing school or network-based group, the external expert needs, in addition to the capacity described above, the ability to foster effective ongoing collaboration between participants. Highly skilled facilitation is needed in which the leader can ‘challenge assumptions and present teachers with new possibilities; challenge the social norms by which collegial groups operate wherever these norms constrain professional learning; and keep the focus on students and their learning’ (Timperley 2008, p. 20).

Since the science of adult learning is no different from that of children’s learning, teacher professional learning should be designed in accord with that science (Bransford et al. 2000; Bransford et al. 2005). Just like students, teachers need multiple opportunities to learn, including to try things out in their own classrooms, inquire into their effects, modify their approach and repeat this cycle until improvement is evident.

Since so much professional learning now takes place in ongoing school or cluster-based professional learning communities, an important challenge for researchers on professional learning is to identify the leadership practices that not only build collegiality but build the type of collegiality that remains focused on the impact of teaching on student learning (Dumay et al. 2013; Allen et al. 2015).

Ensuring an orderly and safe environment (dimension 5)

The focus of Dimension 5 leadership is creating a physical and social environment that makes it possible for teachers to teach and for students to focus on, enjoy and succeed in their learning. Viewed from an educational rather than a narrow behaviour management perspective, the work of creating a safe and orderly school is fundamentally about increasing the physical, emotional and cognitive engagement of students by meeting their needs for caring relationships, and for control over and success in their learning (Wang and Holcombe 2010). Students’ engagement with school, particularly their attendance, is strongly affected by whether they judge it to be physically and psychologically safe and whether they feel that most of their teachers care about them (Bryk et al. 2010). It is also affected by the strength of parent-school ties (Bryk et al. 2010).

The leadership of higher performing schools is distinguished on this dimension by high social expectations incorporated within clear, fair and consistently enforced discipline routines and by high levels of trust between students, teachers, leaders and the parent community (Bryk and Schneider 2002). Although the average effect of this leadership dimension is small (Marzano et al. 2005; Robinson et al. 2008), it provides a foundation for all the rest. In the absence of a school climate that is conducive to learning, educational improvement is unlikely (Sebastian and Allensworth 2012; Sebastian et al. 2017) but in acting to improve order, leaders must keep educational ends constantly kept in mind. If student management policies and procedures are disconnected from the quality of curriculum and instruction, the result is likely to be increasing use of external incentives and sanctions to get students to engage in alienating activities. If, on the other hand, leaders understand such things as how students experience their classes, how trust develops between teachers and students, and how good teaching fosters students’ engagement and success, then student management policies and processes are more likely to serve educational values.

Schools with clear and consistent routines foster student learning by increasing students’ opportunities to learn. The concept of opportunities to learn has a long history in research on teaching and is a consistent predictor of student achievement (Muijs et al. 2014). We use the concept to refer to the amount of time during which students are cognitively engaged with the assessed curriculum. Students’ opportunities to learn are reduced by absence from school or class, by losing time at the start and end of lessons, and by student misbehaviour. They are increased by the quality and consistency of the routines used to foster attendance, prompt starts to lessons, and on-task classroom behaviour, as well as by features of teaching and the curriculum that were outlined under Dimensions 3 and 4.

Future directions for research on leadership impacts

In the last 15 years there have been three main advances in research on the impact of educational leadership on student outcomes. First, the importance of the quality of student leadership for student outcomes is now incontrovertible. This has been established by studies using measures of school-wide and principal leadership derived from teacher surveys and by econometric analyses which use very different measures and analytic techniques. More collaborations between researchers from these different traditions would be invaluable in advancing our understanding of how to reduce the measurement error that bedevils the study of leadership effects.

Second, there has been a move away from research on the impact of various leadership styles to the study of the impact of specific sets of leadership practices or dimensions. These studies embed leadership in the educational work and thereby reveal what leaders do to make a difference to student outcomes. The five dimensions of leadership practice identified in Robinson’s 2008 meta-analysis, or similar versions of them, continue to provide a useful overview of the complexity of the work of educational leaders and of the practices that distinguish the leadership of higher and lower performing schools (Avenell 2015). A replication of the meta-analysis is needed to update the findings on the average effects of each of the five dimensions on student outcomes.

The large body of evidence about the nature and impact of particular leadership dimensions is now informing a third development in the field – the testing of multivariate models of the pathways through which school leaders make direct and indirect impacts on student outcomes. Do leaders make a bigger difference by, for example, attending to school organisation, the quality of teaching, teacher professional learning and development or relationships with families and the community? To date this research has not yielded consistent findings about the relative importance of the various pathways, a result that is probably attributable to the testing of different models and the use of different measures.

In addition to the suggestions for future research included above, we conclude with some specific suggestions about how to discover more about the impacts of school leadership on student outcomes. Measures are needed of the full range of student outcomes that are included in the relevant curriculum documents, so questions can be settled about whether what we know about the impact of leadership on academic outcomes, for example, can be generalised to such non-academic outcomes as students’ wellbeing at school and love of learning.

As indicated in our brief discussion of recent econometric analyses of principal effects, the search for leadership impacts may be hampered by incomplete and inconsistent specification of how leadership practices make an impact on student outcomes. The effectiveness of such practices may depend on their purpose, frequency, or on their specific qualities (Kyriakides and Creemers 2008). For example, the effectiveness of leaders’ feedback to teachers may be contingent on leaders’ purpose in giving it. In one US study, principals whose classroom observations and feedback were motivated by developmental purposes had a more positive impact on student outcomes than principals whose feedback was motivated by accountability purposes (Grissom et al. 2013). This finding suggests that studies of the frequency of feedback to teachers that do not take purpose into account are unlikely to yield valid findings about the impact of this leadership practice.

Similarly, the impact of many leadership practices on student outcomes will depend on the presence of certain qualities rather than on the frequency of the practice. Conceptualizations of quality should be based on the best available theory and evidence about the practice in question. This requires an interdisciplinary approach with leadership researchers engaging with the sociological, psychological and management literature relevant to the focus practices. For example, in Dimension 1 leadership, we need to shift from measures of whether leaders set goals, or of how well they are perceived to set them, to expert judgments of whether their goal setting practices incorporate the qualities that goal theory suggests increases performance. Similarly, in Dimension 4 leadership we need data on teacher professional community and collaboration that capture such qualities as the percentage of teacher talk that is about the connection between what students have learned and how they have been taught. In summary, it is important that future studies specify whether the impact of a leadership practice is attributable to its purpose, frequency or specific qualities, and that measures of the relevant practice are aligned to that specification.

A related challenge is the source of the leadership data. There is some evidence that teachers’ ratings of their leaders, which are based on their tacit theories of effectiveness, do not necessarily match our best objective indicators of leadership quality (Reid 2014; Di Liberto et al. 2015). Leadership surveys which collect data from leaders’ and teachers’ ratings of leadership practices suffer from restricted variance and various types of measurement error (Burkhauser et al. 2012). More studies are needed which use reliable protocols to score the quality of leadership practices based on behavioural or detailed interview data (Di Liberto et al. 2015). If such studies also obtained teacher and leader ratings on the same items, then robust tests could be conducted of the correlation between those ratings and experts’ ratings of leadership quality derived from behavioural or interview data.

As knowledge accumulates about what distinguishes higher and lower quality leadership, leadership development policy and practice will become more focused and evidence-based. The next challenge will then be to develop research programmes on the leadership skills, knowledge and dispositions that are required to enact the full range of high-impact leadership practices.

Disclosure statement

No potential conflict of interest was reported by the authors.