The link between school leadership, staff job satisfaction and retention. Longitudinal evidence from England

Abstract

This paper presents new evidence on the link between staff perceptions of school leadership and job satisfaction, drawing on three waves of The Engagement Platform data collected from over 4000 members of school staff in England. We find that when staff have a positive rather than an ambivalent view of school leadership, their job satisfaction is markedly improved. Views of school leadership in the autumn term are also found to correlate with whether staff leave their job by the end of the academic year. Together, our results demonstrate how views of leadership play a pivotal role in the satisfaction and retention of school staff.

Key insights

The main issue that the paper addresses is the link between staff views on school leadership, their job satisfaction and staff retention. Our main finding is that staff views on school leadership in the autumn term are linked with their job satisfaction and whether they have left their job by the end of the academic year.

INTRODUCTION

In almost every industry, one of the most important resources available to an organisation is its staff. For instance, in education, several studies have shown the benefits to children of being taught by a higher quality teacher (Hanushek, 2014); seminal work conducted by Hanushek and Rivkin (2010) suggested that a pupil taught by one of the most ‘effective’ teachers makes between 0.1 and 0.3 standard deviations more progress over the course of an academic year than their peers taught by an ‘average’ teacher. It is therefore vital that schools can keep talented individuals motivated and engaged in their job, and do not lose their best members of staff to other employers (either within or outside the education sector).

Unfortunately, in some countries, schools are struggling to achieve this goal. In England—the empirical setting of this paper—evidence from the Teaching and Learning International Study (TALIS) has long pointed towards comparatively low levels of job satisfaction amongst teachers compared with other countries (Jerrim & Sims, 2019). This has been supported by a host of quantitative and qualitative research suggesting that many teachers are dissatisfied with certain aspects of their job (Cooper-Gibson, 2018; Towers et al., 2022; Zieger et al., 2019). The result is that record numbers are leaving the teaching profession (House of Commons, 2024). Similarly, it is also proving challenging for schools to retain their best auxiliary staff (Lucas et al., 2023).

While many factors are likely to be contributing to the dissatisfaction of school staff and their subsequent defection to other employers, how they view the leadership of the school is likely to be key (Sims, 2017). For instance, when staff have faith in the school leadership—and feel that they can count on them for support—they are likely to believe their working conditions will improve and become more motivated in their job. This will, in turn, increase retention rates amongst school staff and ultimately maintain firm foundations on which pupils’ learning and development can be supported (Kraft et al., 2016). Indeed, previous research has suggested that staff views of school leadership is one of the most important working conditions for job satisfaction and retention amongst education staff (Kraft et al., 2016; Ladd, 2011; Sims, 2020).

The central aim of this paper is to build further evidence on the link between staff views of school leadership, their job satisfaction and whether they decide to leave their job. This includes not only teachers, but also teaching assistants, administrators, catering assistants, facilities workers and those in pastoral support roles, amongst others. Existing evidence on these matters is somewhat limited by the cross-sectional nature of the data available, and the correlational analysis that has been undertaken (Nguyen et al., 2023). This paper will, in contrast, present results from a panel data analysis, focusing on how changes in staff views of school leadership are related to changes in their job satisfaction over the course of an academic year. We also illustrate how staff views on school leadership are associated with rates of staff retention over an 8 month period. Together, this provides one of the most comprehensive analyses of the link between school leadership, staff job satisfaction and employee retention conducted in England to date.

Thus, in summary, this paper will:

• Use a largescale dataset to track how staff perceptions of school leadership are related to their job satisfaction.
• Investigate whether staff who are dissatisfied with the leadership in the autumn term are more likely to have left their job by the end of the summer term.
• Establish the role of school leadership in determining the satisfaction of staff in their job and their subsequent retention.

School leadership in England's school system

In England, there are many different types of school, with different management structures and associated styles of leadership. Some are state maintained schools which operate under local authority control and are often run by a single, standalone headteacher (particularly in the primary sector). Maintained schools in England can, however, also federate together under one governing body. Other schools are ‘academies’ which may also be standalone schools or run by multi-academy trusts (MATs), where leaders are responsible for more than one school (this is more common in England's secondary sector). Multi-academy trusts vary in size, can include both primary and secondary schools and may be located in various parts of the country (i.e. not all schools in a MAT are necessarily within the same geographic area).

This diversity in management structures leads to some challenges in defining what is meant by school leadership. While all schools typically have a headteacher who is supported by a subordinate team (e.g. deputy headteachers), there may also be an additional level of leadership in some. For instance, MATs often have executive headteachers (a strategic leader with responsibility for more than one school) or Chief Executive Officers and their executive board. In contrast, as primary schools are less likely to be part of a MAT and are typically smaller in size, headteachers may be more autonomous and may more actively participate in the day-to-day running of their school.

Within this paper, we focus on the leadership of the specific school in which staff work, with this most directly associated with the headteacher and their immediate leadership team (deputy headteacher(s) and assistant headteacher(s)). We recognise however that some of the decisions, rules and actions of these leaders may in part stem from directives coming from more central senior figures within their broader organisational structure.

Theoretical background. Leadership and the job demands–resources model

The job demands–resources model (Demerouti et al., 2001) provides a theoretical lens to consider the link between school leadership, staff job satisfaction and retention. Broadly speaking, this theory postulates that employees face two competing factors. The first is job demands—the physical, emotional and social aspects of jobs that place a burden on staff in their working lives (Schaufeli & Taris, 2014). Within education settings, such demands include high levels of workload (particularly on non-teaching tasks) and managing the behaviour of disruptive pupils (Admiraal & Røberg, 2023). Offsetting these, however, are the workplace resources available. This includes not only access to physical resources (e.g. appropriate learning materials and space) but also social and emotional resources, such as the ability to seek support and guidance from colleagues (Ostermeier et al., 2023). It is then the balance of these demands and resources that determines satisfaction of staff in their role and—ultimately—how they perform in their job (Lee & Jo, 2023).

Tummers and Bakker (2021) provide a systematic review of 139 studies that have considered the role of leadership within the job demands–resources model. They argue that there are three main channels via which leadership operates within this framework. The first is that leaders hold direct responsibility for ensuring that staff have an appropriate balance between job demands and resources (Schaufeli, 2017). Within the school sector, leaders may not be fully in control of all such relevant issues, given the requirements of central government, and funding constraints, over which they have little influence. Others, however, such as ensuring positive, collegiate relations amongst staff, are key ‘resources’ that good school leaders will provide (Petro & Gega, 2023). Second, leaders may influence how job demands and resources impact employee wellbeing. For instance, Tummers and Bakker (2021) note how even when staff face unavoidably high workloads (as is often the case in school settings during term time), leaders can provide employees with greater autonomy, so they are better equipped to manage these pressures. Alternatively, they may provide inspiring vision so that staff are willing to put in extra hours to achieve such goals (Schaufeli, 2017). Finally, leaders can influence ‘job crafting’—where staff are proactive in changing their job demands and resources so that they remain in balance (Wang et al., 2016). For instance, if an individual feels that pupil behaviour is creating an unacceptable level of demand in their job, they will seek out ways that this can be addressed (e.g. approaching the leadership about the school's behaviour policies; asking colleagues for support—see Huang et al. (2022) for further discussion). Good leaders will encourage and facilitate such self-reinforcing ‘gain spirals’ by providing staff with the scope they need to keep their job demands and resources in check (Tummers & Bakker, 2021).

While the actions that leaders take are obviously important, so too are staff perceptions of the leadership team (Odhiambo & Hii, 2012). Indeed, the actions that staff take—e.g. whether they start to look for another job—are likely to be linked to how they personally view the leadership team (Scallon et al., 2023). This is because staff often view leaders as the individuals responsible for their working conditions; those who have it within their gift to improve their experiences at work. Thus, if a member of staff loses faith in the leadership team, they are likely to believe that the demands of their job are excessive relative to the resources they have available, with this unlikely to improve in the future (Castro Silva et al., 2023). This will quickly lead to the individual becoming dissatisfied in the workplace, leading them to seek employment elsewhere.

Prior literature

Given the importance of school leadership in models of job satisfaction, it is perhaps not surprising that previous studies have empirically investigated how these constructs are linked. For instance, based on a cross-sectional analysis of 226 secondary mathematics teachers in Germany, Jentsch et al. (2023) investigated the link between teacher perceptions of whether the leadership is supportive, communicative and efficient and their satisfaction in their jobs. They report a positive—albeit relatively modest—relationship. Surveying a cross-section of 2569 elementary and middle school Norwegian teachers, Skaalvik and Skaalvik (2011) use structural equation modelling to investigate the link between school working conditions and job satisfaction. Leadership was found to be indirectly related to job satisfaction through the mediating channel of increasing staff's feeling of belonging at their school. Conducting a cross-national analysis of the TALIS 2013 data for lower secondary school teachers, Sims (2017) found that a standard deviation improvement in staff views on school leadership was associated with half a standard deviation increase in job satisfaction. In a recent systematic review including both primary and secondary schools, Cansoy et al. (2019) investigated how different forms of school leadership are linked to the satisfaction of teachers in their jobs. They found that a transformational leadership approach had the strongest association with teacher job satisfaction, while servant and ethical leadership were also important predictors. This is consistent with the work of Liu et al. (2021), who found both distributed and instructional leadership of schools to both be positively and directly linked to how satisfied lower secondary teachers felt in their jobs. Gouëdard et al. (2023) found that ‘growing leadership’—providing teachers with opportunities to be involved in decision making processes within schools—was a key factor promoting job satisfaction amongst secondary school teachers. Further evidence on the link between school leadership and job satisfaction amongst staff can be found in Nguyen et al.'s (2023) recent evidence review.

Fewer studies have focused on the link between staff views of the leadership team and employee retention. Sims (2020) investigated the relationship between a set of working conditions and the desire of secondary school teachers to continue working in their current school, using international data drawn from the TALIS study. He found that ‘leadership shows by far the strongest relationship with desire to stay’ (Sims, 2020, p. 315), with a 2 standard deviation increase in staff perceptions of school leadership associated with a 30 percentage point increase in the desire of teachers to remain with their current school. Kraft et al. (2016) conducted one of the few studies examining the link between middle school teachers’ views on leadership and staff retention utilising panel data from New York. They report that each standard deviation increase in staff views of school leadership is associated with around a 2 percentage point decrease in teacher turnover. A similar association was reported by Ladd (2011), whose analysis included both primary and secondary teachers. She found that each standard deviation increase in perceptions of school leadership to be associated with a 1.8 percentage point decrease in teacher attrition from the school. The only analogous study we are aware of in England is Sims and Jerrim (2020), who analysed data from the TALIS secondary school sample linked to national administrative records. They report that—for an experienced teacher—each standard deviation improvement in staff views of school leadership is associated with a reduction in the probability of leaving the school by the next academic year from 4.1 to 2.3%.

School staff

Almost all the prior literature discussed above has focused on teachers. In contrast, the analysis presented in this paper includes all school staff, encompassing teachers, teaching assistants, administrators, facilities workers and those in pastoral support roles, amongst others. A key rationale for having such a broad focus is that teachers only form half the staff working in England's schools, with the other half of employees working in other roles (Department for Education, 2023). It is widely recognised that these members of staff play a pivotal role in the functioning of schools yet remain under investigated within quantitative education research (Webster, 2021).

Like teachers, schools face challenges with retaining the best members of auxiliary staff (Lucas et al., 2023). For instance, in England, it has been reported that almost three-quarters of teaching assistants had thought about changing careers over the last 12 months (NCFE, 2023). Some of the key drivers of the retention challenges will differ across staff groups. Jerrim (2024), for instance, illustrates that while workload is the primary concern of teachers, pay is the primary issue raised by teaching assistants. Yet other drivers—such as school leadership—will impact all staff groups. In other words, our discussion of the job demands–resources model—and the role of leadership within it—is not just applicable to teachers, but also to a wide array of school staff.

At the same time, the demands of their jobs and the resources they have available will also vary across staff working in different roles. Thus, while some aspects of leadership are likely to have a common effect (e.g. their general approach to communication), others may primarily affect certain groups. On the job demands side, take the approach of the leadership team to pupil behaviour as an example. The decisions leaders make here will have a major impact on the job demands of those working in pupil-facing roles (e.g. teachers, teaching assistants, pastoral support), but maybe less so for administrative staff. Alternatively, leaders may differ in the extent they allow teachers to delegate certain tasks to teaching assistants, pastoral workers and administrators. Such leadership differences in workload expectations may then lead to differences across staff groups in their job satisfaction and the rate at which they are retained.

The same holds true with respect to job resources as well. Some school leaders may prioritise providing resources—including building relationships—amongst staff working in certain roles. They may, for instance, focus professional development opportunities on teachers, with fewer options available to those in non-pupil-facing jobs. Other possibilities include leaders making different choices on staffing levels and allocations across different parts of the school workforce, which may lead to certain groups feeling particularly overworked. This may then have a greater impact on the job satisfaction of some groups of staff than others, with their retention then particularly affected.

Together, the above points towards a need to better understand the role that perceptions of school leadership play in the job satisfaction and retention of staff across the entire workforce, rather than just teachers alone.

The present study

Although the prior research has provided many important insights into the relationship between staff views on school leadership, job satisfaction and retention, there are also some notable limitations. Two key issues stand out that this paper will seek to address.

First, much of the existing literature is based on cross-sectional data, focusing on the correlation between staff views on school leadership and their job satisfaction at a single point in time. When longitudinal information has been collected, few have used panel data methodologies designed to minimise the risk of unobserved confounding driving the results. This limitation has been summarised by Nguyen et al. (2023) in their recent evidence review, noting how most existing studies ‘are correlational in nature and based on a small sample’ (p. 23), with this limitation ‘exacerbated by an over-reliance on cross-sectional data’. In contrast, this study utilises three waves of large-scale survey data to implement a staff fixed-effects approach. This allows us to control for all factors that remain consistent for staff members over time that could be driving both their views of the school leadership and how satisfied they feel in their job. Thus, while we recognise that our estimates may still not capture cause and effect, they are likely to move a step closer to achieving this goal. Our first research question is therefore:

• Research question 1: if staff hold more/less favourable views of the school leadership, does their job satisfaction improve/decline?

Second, while job satisfaction is important in its own right, the literature linking views on school leadership to ‘hard’ outcomes—such as the actual retention of staff—remains relatively scant. For instance, it is common for research to focus on teacher self-reports of their future employment intentions (e.g. their plan/desire to leave in the future) rather than whether they actually leave for employment elsewhere (Grant & Brantlinger, 2023; Nguyen et al., 2024). This is also noted by Nguyen et al. (2023) as a significant limitation, stating how most studies in their review ‘used data on teacher intent to stay/leave, rather than actual departure’. Indeed, we are only aware of one large-scale quantitative analysis linking staff views of school leadership to actual retention decisions in England (Sims & Jerrim, 2020). It is thus important that evidence linking staff views on school leadership with subsequent employment decisions is strengthened. We address this issue in our second research question by asking:

• Research question 2: how do staff views on school leadership near the start of the academic year predict staff retention by the end of the academic year?

DATA

The data we analyse are drawn from The Engagement Platform (TEP), part of the ImpactEd group. This organisation works with schools to collect survey data from staff to inform their plans for staff retention, wellbeing and school improvement. It was first piloted during the 2022/2023 academic year, before being rolled out more broadly during 2023/2024. Three waves of data have been collected at the time of writing (November 2023, March 2024 and June 2024), with future tri-annual surveys being planned for subsequent academic years. While TEP has been designed primarily to provide data and actionable insights back to schools, the information collected also has the potential to inform education research. The author of the paper is part-time research director at ImpactEd group, and thus had immediate access to these data to conduct this research. The process for applying for anonymised extracts of TEP is being finalised, with further details due to be published on the TEP website during 2025 (https://www.tep.uk/research-events#research).

The sample of participating schools is self-selecting, with school leadership teams (either headteachers of individual schools or senior leaders in academy chains) choosing to subscribe to the service that TEP provides. Once a school has chosen to participate, all staff within the school are asked to complete the survey within a 2 week window, with response rates being high within the context of school staff surveys (Jerrim, 2023). The sample of participating schools is broadly comparable with the wider population of schools in terms of observable characteristics (other than an overrepresentation of secondary schools). Jerrim (2024) provides further details. Within this analysis, we draw on data provided by 79 schools (33 primary and 46 secondary).

Within participating schools, response rates are high, standing at approximately 75%.¹ The total number of participating staff was 6499 in November 2023 (wave 1), 5420 in March 2024 (wave 2) and 5048 in June 2024 (wave 3). Of the 6499 staff members who completed the November 2023 survey, 503 had left the school by June 2024 and 4497 were still working at the school and completed the June survey. A further 908 members of staff were still working at the school but did not respond in June 2024. The remaining 591 staff were working in schools that left TEP between the first and third survey waves. The attrition rate between waves 1 and 3—after excluding staff who had left their job and schools that had ceased participating in TEP—is hence 17%.²

Staff views on school leadership

As part of TEP, school employees answer a set of 39 questions using a 0–10 scale. Our primary covariate of interest is staff views on school leadership, measured using the following five questions:

• I think that the senior leadership team is doing a good job.
• We have a strategy that is taking this school in the right direction.
• I feel supported by the leaders and managers in this school.
• The leaders and managers in this school communicate effectively about what is happening.
• When there is a challenging situation with a parent/carer I feel supported by my school.

The average response to these five questions within each survey wave is used to form our ‘staff views of school leadership’ scale. Jerrim (2024) has investigated the psychometric properties of this measure, reporting a Cronbach's α of 0.95, a 4 month test–retest correlation of 0.73 and a school-level intra-cluster correlation of around 0.2.

Job satisfaction

The outcome measure we use to address our first research question is also formed from responses to the TEP survey. Specifically, respondents were asked four questions capturing their satisfaction with working at their current school. These were:

• How likely is it you would recommend this school to others as a place to work?
• If you had local friends looking for a school for their children, how likely is it you would recommend this school to them?
• If you were offered the same job at another school, how likely is it that you would stay at this school?
• Overall, how satisfied are you working at this school?

The average response to these four questions is used to measure staff satisfaction with working at their current school. Internal consistency of this measure is high (Cronbach α = 0.95), as is the 4 month test–retest correlation (0.74). The distribution of this measure is reported in Online Supporting information.

Staff retention

For our second research question, we derive a measure of staff retention by bringing together two sources of information. First, participating schools unenroll staff from the TEP platform when they have left the school. In addition, the June 2024 (wave 3) survey included the following question, designed to capture those existing staff members that would soon be leaving the school:

‘What job will you be doing come September?’

• Working in this school in the same position/role that I have now.
• Working in this school in a new role (promotion).
• Working in this school but on parental leave/carers leave.
• Working in a different school in the same position/role that I have now.
• Working in a different school (promotion).
• No longer working as a teacher/teaching assistant (leaving the education sector).
• No longer working as a teacher/teaching assistant (taking on a non-pupil facing role within the education sector).
• I will be retired.
• Other.

Using these two pieces of information, we derive a binary measure of staff attrition. This variable is coded 1 if either (a) the staff member completed the November 2023 survey but had been unenrolled from the platform by June 2024 or (b) the staff member completed the November 2023 and June 2024 surveys but reported that they would be leaving the education sector or moving to work in another school come the next academic year. For all other staff members, this binary measure is coded 0.³ The school level distribution of this staff attrition measure is presented in Supporting information. In total, we find that 16% of school staff (and 11% of teachers) left their job between November and the end of the academic year. This is in line with national figures for school staff attrition reported by Hodge et al. (2024). The school-level distribution of the attrition rate is also similar to the distribution presented within Sims and Jerrim (2020).

Additional covariates

We also draw on a range of other survey response measures collected within TEP as statistical controls. This includes staff views on their pay, resources available, self-efficacy, enjoyment of teaching, relationships with their colleagues and manager, views on workload and their wellbeing, amongst others. Further details are presented in Jerrim (2024).

METHODOLOGY

Research question 1

To address our first research question, we restrict the sample to wave 3 respondents and estimate a set of Ordinal Least Squares (OLS) regression models. The full model specification is set out in Supporting information, with a summary provided here. These models are designed to capture the association between staff views of leadership at wave 3 (June 2024) and their job satisfaction, conditional on the other variables included in the model.

We estimate a set of models sequentially, each adding an additional set of controls. Initially, model M0 estimates the bivariate association (i.e. a model without any controls). A set of demographic background controls are added in specification M1, including characteristics such as age, gender, job role and length of employment at the school. This specification also includes school fixed effects (i.e. a dummy variable for each school), which effectively control for all between-school differences—including in their respective leadership teams. Estimates from this model thus capture the association between staff members perceptions of school leadership and their job satisfaction, net of differences across schools in their actual leadership. In model specification M2, we additionally control for prior levels of job satisfaction reported in previous survey waves. After doing so, we are effectively estimating whether staff who held more positive views of the leadership team in June 2024 were also more satisfied in their job at this point in time, over and above how satisfied they were in their job previously.

In the next two model specifications, we add additional controls based on survey responses to some of the other topics covered in the wave 3 survey. First, in model M3, we add controls for factors we believe are unlikely to be strongly influenced by staff views on school leadership. This includes confidence in their teaching skills, whether they find teaching pupils rewarding, their views on pay and the resources they have available in their role.⁴ This model is our preferred specification. Then, in model M4, we add a set of further controls for staff views on other aspects of their job, which are themselves potentially influenced by (perceived or actual) differences in school leadership. This includes staff views on their workload, relationships with their colleagues, wellbeing, professional development opportunities and whether everyone in the school has equal opportunities to succeed. While we recognise that these measures could be ‘bad controls’ (Cinelli et al., 2022)—and thus possibly be on the causal pathway linking views of school leadership to job satisfaction—our primary interest is whether there continues to be a relationship even after these potential mediators have been controlled. Finally, in model M5, we also control for staff views on the school leadership in prior survey waves. In other words, do we continue to observe a relationship between staff views on leadership and job satisfaction now (June 2024), over and above how they felt on these matters previously (November 2023/March 2024)?

One potential issue with these models is that our covariate of interest (leadership) and outcome (job satisfaction) are reported by the same member of staff at the same point in time, making common source bias a potential concern (Sims, 2020).⁵ Also, an argument could be made that the collective view of staff is likely to provide a better measure of the actual quality of the leadership, rather than the perceptions of single individuals. We thus also re-estimate our OLS models, but now replacing individual views on school leadership with the average view across all members of staff.⁶ Our anticipation is that this will lead to weaker and less precise estimates. This is due to there being less variation in our covariate of interest, and that this school-wide measure will not pick up the fact that (a) leaders may manage some parts of their school better than others and (b) individuals can hold legitimately different views on what constitutes good leadership and/or how they experience it. We nevertheless believe this to be a useful additional analysis that will help contextualise our results.

Perhaps the most obvious challenge with the OLS models presented in  (equation) is that they rely on a ‘selection on observables’ assumption. In other words, that we have controlled for all factors that cause staff to hold better/worse perceptions on school leadership that are also correlated with their level of job satisfaction. To take a further step towards addressing this issue, we utilise a staff fixed effects approach. This effectively explores how change in staff job satisfaction over time (across the three survey waves) is related to change in staff views on leadership. In doing so, these models effectively control for all factors that are constant within staff members over the course of an academic year. It hence takes our analysis one step closer towards establishing a causal relationship than the OLS approach. The formal specification of these staff fixed effects models is set out in Supporting information.

Estimates from these models now capture how changes in staff views of the leadership are linked to changes in job satisfaction. Three specifications of this model are estimated, following a similar intuition to our OLS approach outlined above. Initially, we do not include any controls (M0). Then, in model M1, we control for changes in staff views on aspects of their job that are—arguably—not likely to be strongly influenced by their views on school leadership (whether they find teaching rewarding, confidence when teaching, views on their pay and resources). Finally, in model M2 we additionally include time-varying controls that are themselves likely to have been impacted by the school leadership (and are thus potentially on the causal pathway between perceptions of school leadership and job satisfaction), including changing views on workload, wellbeing, relationships in the workplace and professional development opportunities. Together, these models will help reveal whether results from our OLS approach are likely to be driven by background characteristics of staff that we were unable to directly control.

For both our OLS and fixed effect models, we begin by including all members of staff in the sample. Each model is then re-estimated restricting the sample to only those staff working in specific roles (e.g. teachers, senior leaders, teaching assistants).

Throughout the modelling process, we standardise the leadership and job satisfaction scales to mean 0 and standard deviation 1, meaning all the estimates can be interpreted in terms of effect sizes (i.e. the standard deviation increase in staff job satisfaction per each standard deviation increase in staff views of the school leadership team). However, we also attempt to explain our results in less technical terms, so that our findings are accessible to a broad audience.

Research question 2

When addressing our second research question, we restrict the sample of staff (and set of covariates used) to those collected in the first (November 2023) survey wave. A series of logistic regression models are then estimated, with the formal specification set out in Supporting information. While we estimate multiple specifications of this model (and present these results in Supporting information) the results presented in the main body of the paper focus on predicted outcomes from our preferred model specification. These demonstrate how the probability of retention changes as staff perceptions of the school leadership move from low to high scores along TEP's 0–10 response scale.

In presenting these estimates, we recognise the limitations of our logistic regression modelling approach. While it is a strength that we can examine a ‘hard’ outcome such as staff retention, this information is available at just a single point in time (the end of the academic year). Estimates from our model are hence likely to capture conditional associations only and will not capture cause and effect. We are also only able to examine staff retention over a relatively short period of time (around 8 months), meaning staff would have had only a limited amount of time to find another job. Our measure also does not allow us to establish ‘positive’ (e.g. promotion) from ‘negative’ (e.g. thinking that employment conditions will be better elsewhere) reasons for why staff have decided to leave. Yet, despite these limitations, exploring whether staff views on school leadership correlate with hard outcomes such as their retention adds another important dimension to our results.

RESULTS

Job satisfaction

The results from our OLS regression models are presented in Table 1. Each column refers to the results from a different model specification. The control variables each specification includes is reported in the lower half of the table (indicated by a ‘Y’). For instance, model M0 does not include any background controls, while model M1 includes controls for staff demographics, the department in which they work and school fixed effects (a dummy variable for each school). The ‘effect size’ column includes the estimate of interest. This captures the standardised change in staff job satisfaction associated with a 1 standard deviation change in staff perceptions of school leadership. The uncertainty in these estimates is presented in the ‘SE’ column (referring to the standard error). However, we also describe these changes in terms of TEP's original 0–10 response scale in the written text below.

TABLE 1. Ordinary least squares regression estimates of the association between staff views on school leadership and their job satisfaction.

	M0		M1		M2		M3		M4		M5
	Effect size	SE	Effect size	SE	Effect size	SE	Effect size	SE	Effect size	SE	Effect size	SE
One standard deviation increase in the leadership scale	0.79*	0.02	0.75*	0.01	0.51*	0.02	0.43*	0.02	0.31*	0.02	0.41*	0.02
N	4661		4661		4661		4661		4661		4661
Controls
School fixed effects	–		Y		Y		Y		Y		Y
Department	–		Y		Y		Y		Y		Y
Demographics	–		Y		Y		Y		Y		Y
Prior job satisfaction	–		–		Y		Y		Y		Y
Views on pay	–		–		–		Y		Y		Y
Views on resources	–		–		–		Y		Y		Y
Self-efficacy	–		–		–		Y		Y		Y
Enjoyment of teaching	–		–		–		Y		Y		Y
Relationship with colleagues	–		–		–		–		Y		Y
Professional development	–		–		–		–		Y		Y
Relationship with manager	–		–		–		–		Y		Y
Views on workload	–		–		–		–		Y		Y
Teacher wellbeing	–		–		–		–		Y		Y
Behaviour	–		–		–		–		Y		Y
Equality of opportunity	–		–		–		–		Y		Y
Prior views on leadership	–		–		–		–		–		Y

• Note: Figures refer to the standard deviation change in staff job satisfaction per each standard deviation increase in the views of leadership scale. Model M3 has been highlighted as our preferred specification.
• * Indicates that the estimate is statistically significant at the 5% level.

The results from model M0—the association without any controls—suggest that each standard deviation improvement in staff views of school leadership is linked to a 0.79 standard deviation increase in job satisfaction. This implies that when a staff member rates the school leadership 2 points higher on TEP's 0–10 response scale, their job satisfaction increases by around 1.6 points on TEP's 0–10 response scale. The results from this initial model indicate a positive response to research question 1, that job satisfaction amongst school staff is indeed higher when they hold more favourable perceptions of the school leadership team. Interestingly, the addition of background demographic controls and school fixed effects in model M1 lead to almost no change in the strength of this association; the parameter estimate is virtually unchanged. Hence, even among staff working in the same school (and with the same leadership team) views on the quality of the leadership remain very strongly associated with their satisfaction at work.

Model M2 then adds a control for how satisfied staff reported they were in prior survey waves. This cuts the strength of the association, although it remains substantial. The addition of further controls for staff's current views about largely exogenous aspects of their job in M3 leads to only a relatively modest further reduction in the parameter estimate. Consequently, although prior job satisfaction and views on other exogenous aspects of their employment can partially explain the very strong bivariate association observed in M0 (and in M1), it is also clear that an important link between staff views of leadership and their job satisfaction remains.

As model M3 is our preferred OLS specification, in Figure 1 panel (a) we present a set of predicted margins, illustrating the association between job satisfaction and staff views of leadership in these variables’ original 0–10 metric. Note that, when doing so, scores from the leadership scale have been entered as a set of 10 dummy variables—allowing us to additionally explore whether the association may be non-linear.

FIGURE 1

Predicted margins of the relationship between staff views on leadership and satisfaction with their job. (a) Ordinary least squares (OLS) regression. (b) Staff fixed effects. Figures in the left-hand panel are based on OLS model specification M3 from Table 1, with the leadership scale entered as a set of dummy variable. Figures in the right-hand panel are based on staff fixed effects model specification M1 from Table 3. The graph illustrates the predicted score on the job satisfaction scale for each point along the views on school leadership scale. The thin line running through each point captures the 95% confidence interval.

The results presented in Figure 1a help illustrate the magnitude of the association that we observe. A member of staff who rates the quality of the school leadership as 9/10 scores 1 point higher on the job satisfaction scale than their colleague who rates the leadership as 6/10 (over and above the controls included in the model). This association is linear, with similar changes in job satisfaction occurring across different parts of the leadership distribution. There is thus little evidence to suggest that—for instance—viewing a school's leadership as at least competent (rather than poor) is any more or less important for staff job satisfaction than if the staff views the leadership as exceptional (rather than just doing an adequate job).

The final two columns of Table 1 illustrate how our estimates change when including additional (and potentially endogenous) controls. Doing so leads to only relatively minor changes to our results. For instance, after including staff views on a range of other aspects of their job (e.g. relationships with colleagues, workload, wellbeing, professional development opportunities) the fall in the parameter estimate is relatively minor and moves back towards its previous value in model M5. Together, these final two sets of results demonstrate how a reasonably substantial link between views on school leadership and staff job satisfaction remains, over and above a wide array of other workplace factors included in the model. This provides further positive evidence with respect to research question 1.

In additional analyses, we have replicated the results presented in Table 1, but with the OLS models now estimated separately for different staff groups (e.g. teachers, leaders, teaching assistants). These estimates can be found in Supporting information. Overall, very similar substantive results are obtained for each. We consistently find there to be strong bivariate associations, which are generally reduced across the model specifications (particularly with the addition of prior job satisfaction between models M1 and M2). Yet even once we reach the final model specification (M5)—which includes an extensive set of controls—we continue to consistently observe an association of approximately 0.4 standard deviations across all four school staff groups. In other words, for each staff group, those who rate the school leadership 2 points higher on TEP's 0–10 response scale report their job satisfaction to be around 0.8 points higher on TEP's 0–10 response scale.

Thus far, we have focused on the link between individuals own views of the school leadership with their own satisfaction in their job. Now, in Table 2, we turn to the link between the school-average view of school leadership (i.e. we treat this as a school-level rather than individual-level construct) and the job satisfaction of individual staff. Otherwise, the results presented in this table can be interpreted in a similar manner to those presented in Table 1.

TABLE 2. Ordinary least squares regression estimates of the association between the average member of school staff's views on the leadership and individual job satisfaction.

	M0		M1		M2		M3		M4		M5
	Effect size	SE	Effect size	SE	Effect size	SE	Effect size	SE	Effect size	SE	Effect size	SE
One standard deviation increase in the leadership scale	0.35*	0.03	0.35*	0.03	0.12*	0.01	0.10*	0.01	0.08*	0.01	0.13*	0.02
N	4695		4695		4695		4695		4695		4695
Controls
Demographics	–		Y		Y		Y		Y		Y
Prior job satisfaction	–		–		Y		Y		Y		Y
Views on pay	–		–		–		Y		Y		Y
Views on resources	–		–		–		Y		Y		Y
Self-efficacy	–		–		–		Y		Y		Y
Enjoyment of teaching	–		–		–		Y		Y		Y
Relationship with colleagues	–		–		–		–		Y		Y
Professional development	–		–		–		–		Y		Y
Relationship with manager	–		–		–		–		Y		Y
Views on workload	–		–		–		–		Y		Y
Teacher wellbeing	–		–		–		–		Y		Y
Behaviour	–		–		–		–		Y		Y
Equality of opportunity	–		–		–		–		Y		Y
Prior views on leadership	–		–		–		–		–		Y

• Note: Estimates refer to the association between the school-average views of the leadership and the job satisfaction of individual members of staff. The estimates capture the standard deviation change in staff job satisfaction per each standard deviation increase in the staff-average views of school leadership. Model M3 has been highlighted as our preferred specification.
• * Indicates that the estimate is statistically significant at the 5% level.

As anticipated, while we continue to observe an association, the strength of the relationship becomes noticeably weaker. In the bivariate model (M0), each standard deviation increase in school-average views of the leadership is linked to a 0.35 standard deviation increase in staff job satisfaction (this is around half the size of the analogous individual level association reported in Table 1). This then falls to 0.12 standard deviations in model M2—once prior job satisfaction has been controlled—and remains at a broadly similar level across all remaining model specifications. Thus, while clearly much reduced, the association between staff's collective views of the leadership and individual job satisfaction remains non-trivial. Given the fact that school leaders may indeed manage some parts of the school (or some groups of staff) better than others, and that individual staff may rate different styles of leadership in different ways, these results continue to point towards the important role that school leaders play in enhancing job satisfaction amongst staff.

Table 3 now builds on these OLS estimates by reporting results from our staff fixed effects models. These estimates fully exploit the panel nature of our data, controlling for all between-individual differences that do not vary over time that could be impacting our OLS results.

TABLE 3. Staff fixed effects estimates of the association between staff views on school leadership and their job satisfaction.

	M0		M1		M2
	Effect size	SE	Effect size	SE	Effect size	SE
One standard deviation increase in the leadership scale	0.59*	0.01	0.50*	0.01	0.35*	0.01
Controls
Views on pay	–		Y		Y
Views on resources	–		Y		Y
Self-efficacy	–		Y		Y
Enjoyment of teaching	–		Y		Y
Relationship with colleagues	–		–		Y
Professional development	–		–		Y
Relationship with manager	–		–		Y
Views on workload	–		–		Y
Teacher wellbeing	–		–		Y
Behaviour	–		–		Y
Equality of opportunity	–		–		Y

• Note: Figures refer to the standard deviation change in staff job satisfaction per each standard deviation increase in the views of leadership scale. The top row presents the estimates when the model is estimated including all members of staff in the sample. The bottom four rows present the estimates when the model is estimated separately by staff group. ‘TA’ refers to teaching assistants. Note that the ‘TA and pastoral support’ group is a sub-set of ‘staff other than teachers’. Model M1 has been highlighted as our preferred specification.
• * Indicates the estimate is statistically significant at the 5% level.

The model without any controls presented in specification M0 indicates that each standard deviation change in staff views of school leadership is linked with around a 0.59 standard deviation change in their satisfaction in their job. This implies that when a staff member's view of their school leadership team improves by 2 points on TEP's 0–10 response scale, their job satisfaction increases by around 1.2 points on the scale. In our preferred model specification—where we add time-varying controls for staff views on their pay, resources, self-efficacy and enjoyment of teaching—the magnitude of this association is only marginally lower. Moreover, even when we include a wide set of potentially endogenous further controls in model M3—including factors such as views on workload, wellbeing, relationships with colleagues and equality of opportunity—the parameter estimate remains a sizeable 0.35 standard deviations. In other words, a 2 point improvement in staff views of their school leadership team continues to be linked to around a 0.7 point improvement in their satisfaction at work. In additional analysis, we have found that similar results hold across staff working in different roles (e.g. teachers, teaching assistants, office staff). The magnitude of these staff fixed effects results is comparable to those from our OLS models. This suggests that the link we observe between staff views on leadership and satisfaction in their job is not being driven by background staff factors that our OLS models were unable to control. Table 3 thus further strengthens the evidence that staff views of the school leadership have a meaningful impact on their satisfaction in their job.

To conclude our analysis of research question 1, Figure 1(b) presents predicted margins from our preferred staff fixed effects model (specification M1). The pattern is very similar to those from the OLS estimates presented in Figure 1(a). The association between views of school leadership and staff job satisfaction is linear, with a 3 point improvement in the former linked with approximately a 1 point increase in the latter.

Staff retention

Figure 2 illustrates our answer to research question 2 by plotting predicted probabilities of staff leaving their job by the end of the academic year, according to the scores they awarded to the quality of the leadership team the previous November. These estimates are based on our preferred model specification (M2), with staff views on leadership quality entering the model as a set of dummy variables (with the model estimates presented in Supporting information).

FIGURE 2

Predicted probability of school staff leaving their job by the end of the academic year across the school leadership scale. Estimates based on logistic regression model specification M2 reported in Supporting information. This model includes controls for gender, age, length of employment, school fixed effects, department, job role, self-efficacy, enjoyment of teaching, views on pay and views on resources. Values of 0.1 on the vertical axis indicate that the staff member has a 10% chance of leaving their job by the end of the academic year. Scores on the views of leadership scale between 0 and 2 have been combined owing to the small number of observations. The thin line running through each point captures the 95% confidence interval.

The results presented in Figure 2 point towards a non-linear relationship. When staff have little faith in the school leadership (awarding ratings of between 0 and 2), they have a 30% chance of leaving the school at some point over the next 8 months. Such low leadership scores are, however, only reported by a small minority of staff (3%). The chances of staff leaving are notably lower (around a 12% probability) when they rate the quality of the leadership between 3 and 5 on the 10 point scale. Approximately 1 in 7 (14%) of the sample awarded the school leadership such scores. However, once staff rate the leadership team as at least 6 out of 10, then the chance of them leaving by the end of the academic year is small (only around 5%). After this point—for leadership scores between 6 and 10 out of 10—the relationship with short-term retention becomes relatively weak. Together, Figure 2 suggests that—when staff views of the leadership team fall below certain points—they are compelled to take relatively quick action and leave their job. However, holding somewhat ambivalent views about the school's leadership—as supposed to being enthusiastic supporters—seems to be of only marginal importance for staff retention (at least in the short term).

CONCLUSIONS

In many countries—including England—schools are struggling to recruit and retain enough high-quality staff. Indeed, it has been widely documented how England has lower levels of job satisfaction amongst teachers than most other countries (Jerrim & Sims, 2019), with this probably leading to many choosing to leave the profession. The working conditions staff face are a key determinant of how satisfied school staff feel in their job, as set out in prominent managerial theories such as the Job Demands–Resources model (Demerouti et al., 2001). School leaders play a central role within such frameworks, holding responsibility for ensuring the demands of the job and resources available to staff remain in-check (Tummers & Bakker, 2021). Prior research has highlighted the pivotal role that leaders play in school staff satisfaction and retention (Kraft et al., 2016; Ladd, 2011; Sims & Jerrim, 2020), with some suggesting that the quality of the leadership is one of the most important working conditions that employees experience within schools (Ladd, 2011; Sims, 2020).

This paper has contributed to this literature by presenting new evidence on the link between perceptions of leadership quality, staff job satisfaction and retention amongst school employees in England. We have found robust evidence that staff perceptions of the quality of the leadership team are linked to higher levels of job satisfaction in England. Specifically, staff who rate their school leadership as 2 points higher along on a 0–10 scale score almost 1 point higher on a 0–10 job satisfaction scale. This relationship holds across our different methodological approaches. We also find that staff views of leadership quality reported in November are predictive of whether they decide to leave their job by the end of the academic year. This association, though, appears to be sharper at certain points along the school leadership scale. For instance, whereas staff who rated the school leadership between 0 and 2 out of 10 had a 30% chance of leaving their role, this fell to around a 12% chance for staff rating the leadership between 3 and 5, compared with just a 5% chance for those awarding a score of 6 and above. It therefore seems that it is mainly when staff hold very negative views of the leadership that they act quickly to leave their job.

These findings are broadly consistent with evidence from the existing literature. For instance, our results demonstrating a strong link between staff views on leadership and their job satisfaction replicate findings from prior work, which is largely based on surveys conducted at a single point in time. We have, however, extended this evidence by illustrating how the association appears to be broadly linear (an issue not explored in most previous studies) and by moving a step closer to establishing causality by controlling for all factors that remain stable within individual members of staff over time via our staff fixed effects approach. This is consistent with predictions made from Job Demands–Resources theory, which postulates that employee wellbeing and performance are optimised when the demands of their job and the resources they have available are kept in check. For instance, school leaders have levers available that can help staff to feel sufficiently equipped to succeed in their role, that they have the autonomy needed to respond to any pressures and ensure staff have the capacity to craft their role to ensure their demands and resources are balanced.

With respect to staff retention, most existing research in this area focusing on ‘hard’ outcomes—i.e. whether staff leave their job rather than just self-reported intentions—comes from the United States. Both Ladd (2011) and Kraft et al. (2016) found each standard deviation increase in staff perceptions of leadership to be associated with a decrease in 1 year attrition rates of around 2 percentage points. The only previous equivalent evidence for England puts the association at a similar magnitude (1.8 percentage points, as reported by Sims & Jerrim, 2020). Results from our analysis points in a similar direction—results from our headline specification imply that each standard deviation increase in perceptions of school leadership (equivalent to around a 2 point increase along the 0–10 leadership scale) is associated with a 3.8 percentage point decline in staff exits.⁷ Together, this body of accumulated evidence provides evidence from both America and England that staff members’ current views of the leadership team are predictive of whether the individual will still be working at the school come the end of the academic year.

It is also notable how our results have a high degree of consistency across different staff groups (as reported in Supporting information). This illustrates how it is not only the job satisfaction and retention of teachers that are impacted by leadership, but other members of staff as well. Returning to the Job Demands–Resources model, leaders play a key role in balancing the demands and resources available to staff, while also ensuring that they have sufficient autonomy to deal with their workplace pressures. These factors are likely to be just as important to administrators, pastoral support workers and teaching assistants in their jobs as they are for teachers—hence the similarity of the relationship between perceptions of leadership, job satisfaction and retention we have found. It thus seems that results reported previously in the literature for teachers (e.g. Kraft et al., 2016; Ladd, 2011) generalise to school staff working in other roles as well.

There are, however, also some important limitations to our work. First, while our methodological approach controls for all factors that remain stable within staff members over time, there could still be other features that vary for individual staff over time that we have not been able to control. Thus, while we believe that our analysis has included a richer set of controls than previous studies, our estimates may still not establish cause and effect. Second, relatedly, our analysis of staff retention is based on a logistic regression model and may thus also not be capturing a causal relationship. Third, we have only been able to examine change in staff job satisfaction and retention over a relatively short time horizon. Indeed, one may argue that many staff will not have had chance to find alternative employment over an 8 month period. Future work should hence seek to explore these relationships over longer periods of time. Fourth, the number of schools included in our sample (n = 79) would ideally be higher. Finally, while our analysis can point towards the importance of school leadership for staff job satisfaction and retention in general, it cannot provide insight into some of the detailed nuances regarding leadership within the Job Demand Resources model. For instance, we are unable to establish which of the three channels discussed by Tummers and Bakker (2021)—directly balancing demands and resources, influencing the impact via staff wellbeing or influencing job crafting—drives the associations observed.

While addressing these limitations in future research would undoubtedly further strengthen the evidence base, our findings continue to have important implications. They demonstrate to school leaders the need to gather and monitor honest feedback from staff about their views of the leadership team. When there are staff who are starting to lose confidence in them, leaders should find ways to open direct (and ideally anonymous) channels of communication. It is of course unlikely that all the challenges that staff and school leadership face will be remedied by more open communication alone. Schools having insufficient resources and staff being required to respond to frequent policy changes mean that managing schools has become an increasingly difficult task. Yet our results may point towards methods of communication and leadership styles used in the current climate meriting further review. In doing so, this will hopefully help to limit potential declines in the job satisfaction of school staff before they start to look for employment elsewhere.

FUNDING INFORMATION

This work was completed without additional funding.

CONFLICT OF INTEREST STATEMENT

The author is Research Director at ImpactEd Group which owns TEP.

ETHICS STATEMENT

Ethical approval for secondary analysis of the TEP data was granted by the UCL Institute of Education's ethics research committee: REC 1922.

Endnotes

¹ This figure is based upon the data collection in June 2024.

² This is calculated as 4497/(6499 − 591 − 503) = 17%. If teachers within schools that left TEP are also included, this figure for attrition increases to 25%.

³ Individuals who reported that they would be retiring, other or moving to a non-pupil facing role within the education sector have been coded as missing data, and so excluded from the analysis.

⁴ One could argue that leaders may have some control or influence over some of these matters. However, on balance—and compared with other factors—we believe that the link between staff views on leadership and such matters is likely to be relatively weak.

⁵ This is where a spurious correlation between variables appears in the data owing to the same individual reporting information on both (potentially at the same time).

⁶ When doing so, we remove the school dummy variables from the model, as these will be perfectly colinear with this school-average measure.

⁷ This is based upon a linear probably model version of specification M2 from Supporting information.

Open Research

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from The Engagement Platform. Restrictions apply to the availability of these data, which were used under licence for this study. Data are available from The Engagement Platform with the permission of the ImpactEd Group.