Recent Trends in Formal School Exclusions in Wales

ABSTRACT

Historically Wales has been regarded as a country with relatively low levels of school exclusion, particularly in comparison with England. This has been used as an indicator of Wales’ commitment to the UN Convention on the Rights of the Child, which foregrounds a rights-based agenda that would argue school exclusion is a consequence of broader socio-economic structures than individual actions. However, simple analyses may mask a different picture of school exclusions in Wales. In this article, we study more detailed information on formal school exclusions by looking at ‘instances’ of school exclusion, which considers the frequency and length of formal school exclusions in Wales, and alongside recorded school absence. The article also demonstrates that the pupils ‘at risk’ of formal school exclusion varies by when they experience school exclusion during their educational lifecourse. The article concludes by highlighting that patterns of school exclusion are complex, and that interpretation of these patterns depends on what kind of measure of school exclusion is used. By publishing these results, the article hopes to persuade policymakers and practitioners to re-examine their processes of exclusion to ensure they are commensurate with the law and to prevent unnecessary school exclusions.

Keywords:

• exclusion
• excluded
• special education needs
• absence
• free school meals

1. Introduction

School exclusions are probably one of the most controversial educational policies in the UK and frequently polarise audiences. On the one hand, there are those who advocate for zero exclusions, arguing that inclusive education and the rights of a child should surpass any requirement to exclude pupils. On the other hand, there are those who believe exclusions are an important, and possibly necessary, way of maintaining school and classroom environments that are safe and conducive to learning.

During the 1990s, rates of formal school exclusion¹ across the UK were very high compared to recent levels. Concerns about the consequences of this and the wider relationship with social exclusion led the newly created Social Exclusion Unit inside the then Labour UK Government to require a ‘step-change’ in reducing levels of exclusion in schools (Gordon, 2001). Around the same time, administrative devolution gave Scotland and Wales powers to manage their own education sectors.

In Scotland and Wales, a children’s rights agenda came to the fore when determining new education policy and guidance. In Wales this began with the Learning Country, stating that ‘the interests of learners override all others’ and that ‘inequalities in achievement […] must be narrowed in the interests of all’ (National Assembly for Wales, 2001, p. 10). This paving document provided the framework for new ideas for the Welsh education sector (Daugherty and Jones, 2002) and made explicit reference to reducing school exclusions, with a target to reduce exclusions by a third.

A few years later, the National Assembly for Wales adopted the United Nations Convention on the Rights of the Child (UNCRC) in 2004. This provided further encouragement to ensure the interests of the child were considered in all policy decisions. This was extended further into law with the introduction of the Rights of Children and Young Persons (Wales) Measure 2011, placing a duty on Ministers to have due regard to the UNCRC. Subsequently, the National Assembly for Wales (now the Welsh Government) revised its guidance on school exclusions in 2012 to reflect these new laws and then commissioned an evaluation of the education provision of children and young people educated outside the classroom (McCluskey et al., 2013).

During this time, rates of formal exclusion in England, Scotland, and Wales did appear to decline from the heights of the 1990s, but to varying degrees in each country. However, it took several years before reliable data on school exclusions were being collected and published in each jurisdiction.

These declining levels of formal school exclusion can be seen in Figures 1 and 2. Figure 1 illustrates the overall rate of permanent exclusions – occurring when a child is expelled from school – in primary, secondary, and special schools for England, Wales, and Scotland from 2006/07 to 2020/21 (figures for Northern Ireland are too small to be published). Figure 2 shows the equivalent rate of temporary exclusions – each time a child is temporarily removed from school for a period up to 45 school days (figures for Northern Ireland were not available until 2011/12). Both Figures show a significant decline in the rate of both kinds of exclusions between 2006/07 and 2013/14 in England, Wales, and Scotland, albeit to varying degrees.

Figure 1. Rates of permanent exclusions in England, Wales, and Scotland 2006/07 to 2020/21¹

¹Figures for Northern Ireland are not available due to data suppression (low numbers)

[Data sources: Wales: https://gov.wales/permanent-and-fixed-term-exclusions-schools; England: https://www.gov.uk/government/collections/statistics-exclusions; Scotland: https://www.gov.scot/publications/school-exclusion-statistics/]

Figure 2. Rates of temporary exclusions in England, Wales, Scotland,¹ and Northern Ireland,² 2006/07 to 2020/21

¹Data only collected every two years in Scotland from 2010/11

²Data not available for Northern Ireland until 2011/12

[Data sources: Wales: https://gov.wales/permanent-and-fixed-term-exclusions-schools; England: https://www.gov.uk/government/collections/statistics-exclusions; Scotland: https://www.gov.scot/publications/school-exclusion-statistics/; Northern Ireland: https://www.education-ni.gov.uk/articles/pupil-suspensions-and-expulsions]

Several commentators have since made the connection between the greater focus on children’s rights and declining levels of school exclusion. For example, Lewis et al. (2017) noted declining levels of school exclusion in Wales from 2004 and suggested that this provides evidence of the impact of the UNCRC and the rights agenda.

The Children’s Commissioner for Wales (2020) also made this link between exclusions and the adoption of the UNCRC but remained critical of the number of school exclusions in Wales, particularly amongst young children under 8 years old. The Commissioner drew a comparison with Scotland to illustrate this point, highlighting their lower levels of exclusion due to more prescriptive guidance and ‘a strong focus on implementing children’s rights, a nurturing approach and an emphasis upon inclusion and positive relationships’ (2020, p. 9). A detailed comparison of school exclusion policies in 2019 by McCluskey et al. found that the success of Scotland was due to ‘the effectiveness of a strategic emphasis on prevention, of national/local co-design and planning, and of maintaining focus on the complexity of some young people’s lives and the often deep levels of disadvantage’ (2019, p. 1155).

As shown in Figure 2, official rates of temporary exclusion in Northern Ireland, which has had a more chequered experience of devolution since 1999 than the other devolved nations, are also notably lower than in Wales and England (see also Barr et al., 2000). Although there remains some concern about the consistency of data gathering in Northern Ireland, Duffy et al. (2021) suggest that these low rates are due to a strong ethos of caring, reflecting, again, the influence of the children’s rights agenda.

Other explanations for variation in the levels of formal school exclusions across the UK include the influence of marketisation, cultural restoration, and the ‘hollowing out’ of local government, all relatively strong in England (Cole et al., 2019; Power and Taylor, 2021). For example, increased accountability and autonomy of schools may encourage some schools to exclude pupils who they deem problematic or resource-hungry. Similarly, a return to traditionalism and tightening of rules within schools may increase the level of disengagement amongst pupils. The declining influence of local government administration in some parts of the UK has meant that there is less external control and support in mitigating school exclusions.

However, the problem with these policy explanations, what Power and Taylor (2021) refer to as the context of influence, is that (a) they rely on measures of official/formal exclusions, (b) they do not explain significant variations in formal exclusions within each country, and (c) they do not adequately explain changes in the rates of school exclusions in more recent years. For example, Power and Taylor (2020) argue that official statistics do not include large amounts of ‘hidden’ exclusion, such as managed moves and internal exclusions.

Similar claims have been made by others over the years, including children being sent home from school during the school day (Children’s Commissioner, 2020), elective home education and illegal schools (Daniels et al., 2022), parents being encouraged by schools to keep their children at home (Gordon, 2001), and processes of ‘off-rolling’ whereby schools illegally remove pupils from their school roll (Done and Knowler, 2020). But few have been able to evidence these claims or, more importantly, demonstrate the scale of informal exclusions. Done (2022) provides a detailed analysis of the methodological and policy challenges of studying unlawful or illegal school exclusionary processes. A recent exception to this is Power and Taylor (2022) who attempted to record the level of classroom exclusion (a form of internal exclusion) in Wales. They found that nearly a third (31.2%) of secondary school age pupils surveyed said they had been asked to leave a classroom at least once in the previous 12 months, 15.4% reporting that they had been asked to leave a classroom ‘several times’ in the 12 months, and 5.4% saying they had been excluded from the classroom ‘about once a week’. Although these examples are not official, or formal, exclusions, they nevertheless contribute to the same outcome – being excluded from the classroom learning environment. Furthermore, one could argue that these informal examples of exclusion are potentially more damaging to pupils than formal exclusions because they are unregulated, invisible, and less likely to provide any kind of educational support or additional learning activities to affected pupils; something that is meant to be provided if a pupil is formally excluded.

A second problem with these policy explanations is that there remains significant variation in levels of school exclusion within countries. Given the relatively low number of exclusions in each school, it is not straightforward to compare the exclusion rates of individual schools. However, studying exclusion rates at the local authority is possible, particularly when pooling numbers of exclusion over time. Figure 3 shows recent variations in the rates of school exclusion by local authority in Wales. This shows that in some parts of Wales, exclusion rates can be 10 times higher than in other areas.

Figure 3. Average rate of school exclusions by local authority in Wales

[Source: Tseliou (2021)]

As Tseliou (2021) demonstrates, much of this geographical variation can be accounted for by differences in levels of additional learning needs and socio-economic disadvantage in each area. The impact of school exclusions on particular groups of children has received considerable attention over the years (McCluskey et al., 2016), particularly for the most disadvantaged (cognitively, emotionally, and socio-economically).

The third problem is that whilst official government guidance on school exclusions has been strengthened over time, and policy- and decision-makers are increasingly aware of the UNCRC and the importance of the rights of the learner, this has not prevented school exclusions in Wales from rising again in more recent years. The direction of school exclusions over the last two years is not entirely clear because all schools were closed for significant periods during the COVID-19 pandemic, thereby affecting the official rates of formal exclusion.²

As a result of these issues, the relationship between official school exclusions and the policy discourse is far more complex than it might first appear. Other explanations for variations in formal exclusions must also exist. Crucially, these unacknowledged conditions (Power and Taylor, 2021) can often be in direct conflict with other policy discourses when it comes to school exclusions. For example, school exclusions can occur when the actions of one child are at the detriment of another child or adult practitioner. They can also be employed when the school cannot provide the resources required to keep the child in a mainstream setting. In turn, these pressures or tensions give rise to unintended consequences such as other forms of informal or even unlawful exclusion (Power and Taylor, 2021). As Lewis et al. (2017) have previously discussed, the tension between the discourse of poverty and inequalities and the discourse of rights remains unaddressed. The argument here is that policies giving attention to the participation and agency of young people do very little to address the socio-economic structures and context that frame their participation. Conversely, policies to address child poverty are usually framed in terms of their future economic prospects, not the young people’s current ability to participate in democratic change.

In this article, we develop these arguments further by studying more detailed information on formal school exclusions. We examine ‘instances’ of school exclusion, which considers the frequency and length of formal school exclusions in Wales. We also consider school exclusion alongside recorded school absence. This is designed to consider whether the official trends presented above are indeed an accurate reflection of the progress, albeit partial, made by the dominant policy discourses to reduce school exclusions and enact the UNCRC in Wales.

This analysis shows a different trend in the occurrence of school exclusion in Wales and a higher incidence of pupil absence from maintained education than perhaps the dominant policy discourse about education policy and children’s rights in Wales would suggest (Lewis et al., 2017). The article also demonstrates that the pupils ‘at risk’ of formal school exclusion vary during their educational lifecourse – pupils with different characteristics appear to be at higher risk of exclusion at different stages of their schooling. The article concludes by highlighting that patterns of school exclusion are complex, and that interpretation of these patterns depends on what kind of measure of school exclusion is used. By publishing these results, the article hopes to persuade policymakers and practitioners to re-examine their processes of exclusion to ensure they are commensurate with the law and how to prevent unnecessary school exclusions.

2. Data and Methods

As discussed earlier, the prevalence of ‘hidden’ forms of school exclusion limits any analysis of trends or the impact of policies to reduce exclusions. Since we have almost no measure of the scale of ‘hidden’ exclusions, we cannot assess how accurate official records of exclusion really are. However, the recording of formal exclusions has improved considerably over the last 25 years, and the work of charities, such as SNAP Cymru https://www.snapcymru.org/, and local authorities to increase the transparency around school exclusions has certainly helped. Despite their limitations, administrative data on school exclusions are still valuable.

In this article, we use anonymised pupil-level administrative records. These are collected annually by schools for the Welsh Government. These include pupil-level demographic data and are linked to records on every formal exclusion instance a pupil experiences. These data are then made available to researchers through a secure research environment in SAIL and with strict protocols in their use (Ford et al., 2009; Lyons et al., 2009).

These contain several linked datasets, including the Welsh National Pupil Dataset (NPD), that allow us to study all pupils in mainstream education (excluding Pupil Referral Units) and any instances of formal exclusion they receive over time. These unique linked administrative datasets also allow us to consider school characteristics as well as information on their educational progress, attendance, and attainment.

We first explored summary statistics of pupil numbers and exclusion instances over time between 2011/12 and 2016/17 (N = 476,775). Given the time lag in making administrative data available for research purposes, these were the most recent years for which we had access to a complete dataset of all pupils in Wales. We then focused on a study cohort that included pupils who were 15 (born 2000–2001) at the beginning of the 2016/17 school year (N = 27,085). For this cohort, herein referred to as the 2016/17 cohort, personal characteristics included gender and ethnicity. Although Wales is relatively homogenous, pupils were grouped into White British, White, Asian, Black, and Mixed/Other. Free school meal (FSM) eligibility, rather than FSM receipt, was also included as a measure of disadvantage as it has been previously linked to poor educational outcomes (Taylor, 2018). To capture the dynamic changes of FSM eligibility status, we monitored annual recording of eligibility (yes/no) across academic years, grouping pupils into four categories: no FSM, FSM eligible one year, FSM eligible multiple years, and FSM eligible all academic years.

The 2016/17 cohort occurred before the Additional Learning Needs and Education Tribunal (Wales) Bill 2018 was introduced, a new framework on additional learning needs (ALNs), which replaced the legislation on special educational needs (Welsh Government, 2018). Therefore, we use and refer to special educational needs (SENs) throughout this article. We investigated different aspects of SEN, including:

1. Level of provision: Pupils were grouped into no provision; School Action (or Early Years Action); School Action Plus (or Early Years Action Plus) referring to a request for help from external services; and Statutory assessment, a separate category of pupils who have personalised provisions and have been issued with an official SEN statement (and does not necessarily correspond with the presence of more severe needs).

2. Type of need: Pupils were grouped into wider categories including children with no SEN, behavioural or mental health (including emotional and social difficulties), cognitive or learning (including special learning difficulties), communication and interaction (including speech language difficulties), and sensory or physical (including hearing and visual impairment) needs.

3. Complexity of need: Several primary and secondary needs can be identified at different stages of education. We grouped separately those pupils with a single type of need identified and those with multiple needs to investigate how the number of SEN relates to the initial identification of need and how this might relate to the severity of exclusion experiences.

In our 2016/17 cohort, data on school exclusion instances included information on type and length of exclusion as well as reason for exclusion. School exclusions were grouped into fixed-term (during which a child is temporarily removed from school for a period of up to 45 school days in one academic year) and permanent (occurring when a child is expelled from school). The number of half-day sessions lost by each exclusion was also accounted for to estimate the total days of school lost for each academic year. School absence for reasons other than exclusion was also included, focusing on pupils who were persistently absent, losing more than 61 sessions (or 30.5 days) per year. As we were interested in the cumulative role of absence across academic years, pupils were grouped into not persistently absent, persistently absent for 1–2 years, and persistently absent for 3 years or more.

School-level characteristics were assessed to allow for the comparison of variations in school exclusion rates across different schools. They included: (a) the primary language used in each school (grouped as English medium, Welsh medium type, and dual/other medium type) and (b) school religion (grouped into no faith or faith schools).

In addition, information on the 22 local education authorities (LEAs) of Wales was available across years. LEAs vary in size of pupils and schools within each council, resulting in small numbers that could be disclosive, so they are grouped into the four regional educational consortia in Wales that work closely with schools. These included North (Anglesey, Gwynedd, Conwy, Denbighshire, Flintshire, and Wrexham), West (Powys, Ceredigion, Carmarthenshire, Pembrokeshire, Swansea, and Neath Port Talbot), Central South (Rhondda Cynon Taff, Bridgend, Vale of Glamorgan, Cardiff, and Merthyr Tydfil), and Southeast Wales (Newport, Monmouth, Caerphilly, Torfaen, and Blaenau Gwent).

Of course, during their educational lifecourse, many pupils move schools and local authorities. Similarly, schools and local authorities also change over time. Tracking these changes during a pupil’s educational lifecourse is complex, so for simplicity in this analysis all school-level and LEA-level variables are based on their situation and the location of pupils in 2016/17, when they were in Year 11.

Our analysis is presented in two parts. First, we present new and detailed analysis on the rates of exclusion in Wales. In this analysis, we compare the number of excluded pupils with instances of exclusion and days lost to exclusion. The second part of the analysis examines the characteristics of excluded pupils in Wales using descriptive statistics and logistic regression models. We consider whether particular pupil characteristics are associated with being formally excluded at different stages of the educational lifecourse. We do this by categorising pupils according to when they were temporarily excluded from school at three different points of their schooling:

1. Key Stage 2 (Year 6)

2. Key Stage 3 (Years 7–9)

3. Key Stage 4 (Years 10 and 11)

We do not use exclusion data before Year 6 because of the very low number of exclusions recorded in the Foundation Phase and the first few years of Key Stage 2. As a result, we are only using data for children from Year 6 (end of KS2, aged 10–11 years) to Year 11 (end of KS4, aged 15–16 years).

3. Rates of Exclusion in Wales

Excluded Pupils

The official rates of school exclusions are based on calculating the number of exclusion instances per 100 pupils. As we will see later, an individual pupil could be excluded multiple times, contributing more than one exclusion instance to the official rates. Therefore, the number of pupils who experience at least one exclusion each year may look very different to the rates of exclusion.

Figure 4 presents the number of pupils in Wales who were excluded in each academic year between 2011/12 and 2016/17. This shows that the highest number of excluded pupils during this period was in 2016/17. Although pupil numbers vary year-on-year, this rise is quite significant and can be seen in Figure 5, which shows the proportion of pupils being excluded each year. Crucially, this rise is not as evident in the rates of exclusion presented in Figures 1 and 2, which suggests a pattern of a larger number of pupils experiencing a smaller number of exclusions during the year.

Figure 4. Number of excluded pupils in each academic year over time

Figure 5. Percentage of pupils excluded each year

Days Lost to Exclusion

Another way of examining trends in school exclusion is by looking at the number of days lost due to formal exclusions. Figure 6 presents the total and average number of days lost to exclusion over our selected time period.

Figure 6. Total and average number of days lost due to exclusion in each academic year

In 2016/17, the total number of school days lost due to temporary school exclusions was 3,953 days, an increase from 3,469 days in 2013/14. Although the total number of days lost to temporary school exclusions declined and then rose over time (just like the number of pupils who were excluded at least once in each year), the average number of days lost to formal exclusion remained relatively constant. For example, in 2011/12, the average number of days lost due to school exclusion was 2.7 per excluded pupil in Wales. This has declined slightly over time, perhaps suggesting that attempts to return pupils back to school or reduce repeat exclusions may be working.

Instances of Exclusion

Finally, we examine each instance of school exclusion (Figure 7). An instance of temporary school exclusion can last for one half day up to a series of temporary exclusions not exceeding 45 days per school year (Welsh Government, 2019).

Figure 7. Total and average number of exclusion instances in each academic year

The number of exclusion instances went from 15,013 in 2011/12 to 16,100 in 2015/16. Each instance will have been recorded separately by schools. Importantly, the trend over time is not as pronounced as the other measures presented above and is certainly very different from the official rates of formal exclusions in Figure 2. It is likely that the lived experience of schools and teachers is likely to reflect the number of exclusion instances rather than the rate of exclusions or the number of days lost to school exclusions. This could have important consequences on how professionals are evaluating the effectiveness of their policies and enactment of Welsh Government guidelines.

Figure 7 also shows the average number of exclusion instances per pupil. This rises more markedly in 2014/15 and 2015/16 than the average number of days lost to exclusion in Figure 6. This suggests that over time pupils are being excluded for shorter periods of time but more frequently.

We can prove this by looking at the average number of exclusion instances per pupil over time (Figure 8). As expected, the average number of exclusion instances indeed rises each year from 2011/12 to 2015/16. However, we observe a drop in the average number of exclusion instances in 2016/17, possibly reflecting the rise in the number of pupils who were excluded in that year.

Figure 8. Average number of exclusion instances per cohort pupil by age

What we have seen from looking at pupil-level exclusion data in more detail are quite different trends over time – quite different from the officially published trends in Figures 1 and 2. Not only do they provide a more complex picture of school admissions they can also reveal trends in different processes and practices of school exclusion being used. If this analysis were repeated for other countries in the UK, then perhaps our judgement about how well we are managing school exclusions and how ‘inclusive’ the education system in Wales has become may be very different (positively or negatively). But it is certainly clear from these results that the overall levels of school exclusion are perhaps higher in Wales than first anticipated.

We now go on to look in detail at the characteristics of pupils who have been excluded from school in Wales.

4. Characteristics of Excluded Pupils in Wales

We begin this section of the article by conducting a fairly standard analysis of the characteristics of excluded pupils in Wales. In these results, we seek to identify the characteristics associated with pupils who have ever been excluded. As noted earlier, there are two groups of pupils who are most likely to be excluded: children with additional learning needs and children from disadvantaged socio-economic backgrounds. It is, however, important to note the overlap of these two groups. So, to disentangle the relative importance of these different characteristics, we employ multivariate analyse, allowing us to simultaneously observe the association of more than one characteristic or variable on our chosen outcome: having ever been excluded.

However, as we have seen previously, the occurrence of exclusions is complex and can occur at any time during a pupil’s educational lifecourse. So, in the second part of this analysis, we look to see whether the characteristics of pupils excluded at different stages (or ages) are different.

Throughout this section, we use a single cohort of pupils who were in Year 6 in 2011/12 and look to see if and when they were formerly excluded from school between Year 6 and Year 11 (the cohort were eventually in Year 11 in 2016/17). This 2016/17 cohort, therefore, consisted of 27,085 pupils with 9.9% having experienced school exclusion (N = 2,676) at any time-point between Years 6 and 11. We used a combination of longitudinal characteristics that change over the lifecourse, including free school meal (FSM) eligibility status and school attendance, along with other more stable characteristics such as ethnicity, school type, and regional consortia between Year 6 or Year 11. We did not explore changes in special education needs (SEN) over the lifecourse, as pupil needs might change in terms of type and/or complexity but cannot be calculated as an entry/exit characteristic for pupils across education stages.

Characteristics of Pupils Ever Been Excluded

We begin by looking at the characteristics of those pupils who were ever excluded between Year 6 and 11 (Table 1). We can see very clearly that excluded pupils were more likely to be male (70.2% of all excluded pupils), to have been eligible for free school meals for more than one year (54.5% of all excluded pupils),³ to have behaviour and mental health needs (8.9% of all excluded pupils), to have cognition and learning needs (32.1% of all excluded pupils), to have had more than one special educational need (50.2% of all excluded pupils), and to have been persistently absent across years (45.2% of all excluded pupils). These results are generally consistent with other analyses of excluded pupils in the rest of the UK. However, we also see that White British pupils were slightly more likely to be excluded in Wales than pupils of other ethnicities, and pupils in Welsh Medium schools were less likely to be excluded than pupils in English Medium schools, results which may be more surprising given the attention given to the exclusion of minority ethnic pupils in England.

Table 1. Number (and percentages) of pupils ever excluded between year 6 and year 11 by characteristics

		Total cohort	Excluded pupils
Gender	Female	13,002 (48.0)	797 (29.8)
	Male	14,083 (52.0)	1,879 (70.2)
Ethnicity	White British	23,785 (87.8)	2,460 (91.9)
	White Other	1,156 (4.3)	91 (3.4)
	Asian	708 (2.6)	37 (1.4)
	Black	269 (1.0)	26 (1.0)
	Mixed/Other	1,167 (4.3)	62 (2.3)
FSM status	No FSM	18,826 (69.5)	1,062 (39.7)
	FSM once	1,456 (5.4)	155 (5.8)
	FSM multiple	4,595 (17.0)	882 (33.0)
	FSM constant	2,208 (8.2)	577 (21.5)
SEN type	No SEN	19,212 (70.9)	1,332 (49.8)
	Behaviour & mental health	677 (2.5)	238 (8.9)
	Cognition & learning	5,308 (19.6)	858 (32.1)
	Communication	1,186 (4.4)	173 (6.4)
	Sensory & physical	702 (2.6)	75 (2.8)
Comorbid SEN	No SEN	19,212 (70.9)	1,332 (49.8)
	1 SEN	5,823 (21.5)	813 (30.4)
	Multiple SEN	2,050 (7.6)	531 (19.8)
SEN provision	No provision	14,564 (53.8)	607 (22.7)
	School Action	5,974 (22.1)	679 (25.4)
	School Action Plus	5,458 (20.1)	1,192 (44.5)
	SEN Statement	1,089 (4.0)	198 (7.4)
Persistent absence	No	21,678 (80.0)	1,466 (54.8)
	1–2 years	3,808 (14.1)	798 (29.8)
	3+ years	1,599 (5.9)	412 (15.4)
Medium type	English	20,330 (75.0)	2,195 (82.0)
	Welsh	5,325 (19.7)	429 (16.1)
	Other	1,430 (5.3)	52 (1.9)
School religion	No faith	21,736 (80.3)	2,166 (80.9)
	Faith school	5,349 (19.7)	510 (19.1)
Regional consortia	North Wales	6,377 (23.5)	647 (24.2)
	West Wales	7,713 (28.5)	621 (23.2)
	Central South Wales	8,120 (30.0)	884 (33.0)
	Southeast Wales	4,875 (18.0)	524 (19.6)

However, there are considerable overlaps in many of these categories, so Table 2 presents the results of multivariate logistical regression that helps control for the other characteristics. This table presents the odds ratio, or probability, that a pupil with a particular characteristic will have ever been excluded between Year 6 and Year 11. The comparison group always has the odds ratio of 1.00. There are two sets of results in Table 2, the univariate odds ratio (without being controlled for by other characteristics) and the multivariate odds ratio (controlled for by other characteristics). Although we also present the 95% confidence intervals for each variable it should be noted that we are using a population dataset, so the results (the odds ratios) are not dependent on any significance testing.

Table 2. Logistic regression models using ever excluded as outcome (odds ratios with 95%CI)

		Univariate	Multivariate
Gender	Female	1.00	1.00
	Male	2.36 (2.16–2.57)	2.15 (1.96–2.37)
Ethnicity	White British	1.00	1.00
	White Other	0.74 (0.60–0.92)	0.72 (0.57–0.91)
	Asian	0.48 (0.34–0.67)	0.56 (0.39–0.79)
	Black	0.93 (0.62–1.39)	0.89 (0.57–1.38)
	Mixed/Other	0.49 (0.38–0.63)	0.56 (0.42–0.73)
FSM status	No FSM	1.00	1.00
	FSM once	1.99 (1.67–2.38)	1.62 (1.34–1.95)
	FSM multiple	3.97 (3.61–4.37)	2.32 (2.08–2.58)
	FSM constant	5.92 (5.28–6.63)	2.82 (2.47–3.21)
SEN type	No SEN	1.00	1.00
	Behaviour & mental health	7.28 (6.16–8.60)	4.86 (3.82–6.18)
	Cognition & learning	2.59 (2.36–2.84)	1.84 (1.57–2.16)
	Communication	2.29 (1.93–2.72)	1.02 (0.83–1.24)
	Sensory & physical	1.61 (1.26–2.05)	0.67 (0.51–0.88)
Comorbid SEN	No SEN	1.00	1.00
	1 SEN	2.18 (2.00–2.39)	0.45 (0.38–6.18)
	Multiple SEN	4.69 (4.19–5.26)	Suppressed
SEN provision	No provision	1.00	1.00
	School Action	2.95 (2.63–3.30)	2.19 (1.91–2.51)
	School Action Plus	6.42 (5.79–7.13)	3.54 (3.11–4.04)
	SEN Statement	5.11 (4.29–6.08)	2.33 (1.90–2.87)
Persistent absence	No	1.00	1.00
	1–2 years	3.66 (3.32–4.01)	2.22 (2.00–2.47)
	3+ years	4.79 (4.23–5.42)	2.22 (1.92–2.55)
Medium type	English	1.00	1.00
	Welsh	0.72 (0.65–0.81)	0.81 (0.72–0.92)
	Other	0.31 (0.24–0.41)	0.59 (0.44–0.79)
School religion	No faith	1.00	1.00
	Faith school	0.95 (0.86–1.05)	0.97 (0.87–1.09)
Regional consortia	North Wales	1.00	1.00
	West Wales	0.78 (0.69–0.87)	0.63 (0.56–0.72)
	Central South Wales	1.08 (0.97–1.20)	0.90 (0.80–1.01)
	Southeast Wales	1.07 (0.94–1.20)	0.84 (0.73–0.96)

Generally, the univariate results support the observations made of the descriptive statistics in Table 1. However, these results are useful in demonstrating the gradient effects of being eligible for free school meals – the longer a pupil is eligible for free school meals, the greater the likelihood that they were ever excluded. Being eligible for free school meals just for one year nearly doubles the likelihood that a pupil has ever been excluded compared to a pupil who was never eligible for free school meals (OR = 1.99). The odds, or probability, of this increases to nearly four times more likely (OR = 3.97) if a pupil is eligible for more than one year and increased to nearly six times more likely to ever been excluded (OR = 5.92) of they were persistently eligible for free school meals over this period compared to a pupil who was never eligible for free school meals.

The univariate results also highlight the greater likelihood of ever being excluded if a pupil went to an English Medium school. The odds ratio of being excluded in a Welsh Medium school was only 0.72 compared to being in an English Medium school; i.e., 28% less likely to have been excluded. We also see that pupils in West Wales were 22% less likely (OR = 0.78) to be excluded than pupils in North Wales.

However, these last results highlight the limitation of univariate analyses since the composition of pupils in Welsh Medium schools or pupils in West Wales may be very different to other parts of the Wales. Hence, the results of the multivariate analysis in Table 2 are very important.

When comparing the univariate with the multivariate models, a significant amelioration was observed in the effect of FSM status, particularly for those pupils who were FSM eligible across all years (OR = 5.92 and OR = 2.82, respectively). This demonstrates that some of the association with socio-economic disadvantage (measured here by persistent eligibility for free school meals) can be accounted for by other characteristics, such as their special educational needs.

We see a similar effect on many other characteristics associated with ever being excluded. Nevertheless, there are some important results that can be observed. First, the association between being male and being excluded remains relatively unchanged by the inclusion of other variables into the model. This means that their greater likelihood of being excluded compared to females is not due to other characteristics they may share or be associated with. We also see that the association between the ethnicity of pupils and the likelihood of being excluded remains relatively unchanged – White British pupils are the most likely group to be excluded in Wales.

Generally, pupils with special educational needs are still more likely to be excluded from school than pupils without a special educational need, but the odds of this are reduced after taking into account their other characteristics. Although pupils with sensory and physical needs appear to be less likely to be excluded from schools than pupils without any additional learning needs when we consider their other characteristics, this is most likely because they receive significant SEN provision (also included in the model). But we do observe that the impact of multiple special educational needs (comorbidity) is much reduced once we account for their specific educational needs and provision.

Although a gradient association was observed between persistently absent school years and exclusion (1–2 years OR = 3.66 and 3 years+ OR = 4.79), that effect disappeared in the fully adjusted model (1–2 years OR = 2.22 95 and 3 years+ OR = 2.22); absence from school is still associated with being excluded, but the persistent nature of school absence does not appear to worsen that association.

Finally, we see that the influence of school and region are generally reduced after controlling for other pupil characteristics, suggesting that the pupil composition in these contexts is quite different. Nevertheless, pupils in Welsh Medium schools are still slightly less likely to be excluded from school than their equivalents in English Medium schools. And interestingly we see that pupils in West Wales (and Southeast Wales) remain less likely to be excluded from school than pupils in North Wales even after controlling for other pupil characteristics. These results may suggest that differences in school and local authority exclusion practices have some bearing on the chances of being excluded.

Characteristics of Pupils Excluded by Key Stage

We extend this analysis to consider the relationship between pupil characteristics and when they are excluded from school over their educational lifecourse. Table 3 sets out the results of a multivariate multinomial analysis. This is very similar to the multivariate analysis presented in Table 2 except that instead of predicting one outcome (ever been excluded), it attempts to predict three outcomes. Those three outcomes are:

1. Been excluded in Key Stage 2 (Year 6) and Key Stage 3 (Years 7–9)

2. Only been excluded in Key Stage 4 (Years 10 and 11)

3. Been excluded in Key Stage 2 (Year 6), Key Stage 3 (Years 7–9) and Key Stage 4 (Years 10 and 11)

Table 3. Multivariate multinomial regression models for school exclusion trajectory groups (reference group for analysis pupils with no exclusion) (odds Ratios)¹

		Excluded in KS2 & KS3	Excluded in KS4	Excluded in KS2 & KS3 & KS4
Gender	Female	1.00	1.00	1.00
	Male	2.25 (1.94–2.61)	1.93 (1.67–2.20)	2.61 (2.16–3.16)
Ethnicity^2	White British	1.00	1.00	1.00
	Non-White British	0.79 (0.60–1.05)	0.43 (0.31–0.62)	0.58 (0.39–0.89)
FSM status	No FSM	1.00	1.00	1.00
	FSM once	1.78 (1.33–2.39)	1.51 (1.5–1.97)	1.88 (1.28–2.74)
	FSM multiple	2.57 (2.17–3.04)	2.03 (1.72–2.36)	2.99 (2.42–3.70)
	FSM constant	3.21 (2.63–3.790)	2.52 (2.08–3.06)	3.74 (2.93–4.75)
SEN type	No SEN	1.00	1.00	1.00
	Behaviour & MH	2.72 (2.10–3.52)	2.29 (1.90–2.77)	2.58 (1.89–3.53)
	Cognition & learning	1.00 (0.84–1.9)	0.97 (0.82–1.15)	1.15 (0.93–1.42)
	Comm & interaction	0.76 (0.57–1.02)	0.68 (0.50–0.93)	0.76 (0.54–1.07)
	Sensory & physical	0.55 (0.37–0.81)	0.58 (0.38–0.87)	Suppressed
SEN provision	No provision	1.00	1.00	1.00
	School Action	2.10 (1.68–2.64)	2.29 (1.90–2.77)	1.81 (1.33–2.45)
	School Action Plus	4.27 (3.45–5.28)	2.94 (2.43–3.57)	5.32 (4.05–6.99)
	SEN Statement	4.24 (3.17–5.77)	1.16 (0.80–1.70)	3.64 (2.50–5.32)
Persistent absence	No	1.00	1.00	1.00
	1–2 years	2.05 (1.74–241)	2.09 (1.78–2.42)	3.27 (2.69–3.97)
	3+ years	2.29 (1.81–2.74)	1.72 (1.47–2.14)	3.32 (2.60–4.23)
Medium type	English	1.00	1.00	1.00
	Welsh	0.72 (0.60–0.87)	0.89 (0.76–1.04)	0.66 (0.52–0.85)
School religion	No faith	1.00	1.00	1.00
	Faith school	0.89 (0.75–1.07)	0.92 (0.78–1.09)	1.19 (0.91–1.38)
Regional consortia	North Wales	1.00	1.00	1.00
	West Wales	0.89 (0.72–1.10)	0.48 (0.40–0.57)	0.68 (0.53–0.87)
	Central South Wales	1.47 (1.21–1.79)	0.61 (0.51–0.72)	0.85 (0.68–1.08)
	South East Wales	1.08 (0.86–1.36)	0.68 (0.56–0.82)	0.91 (0.74–1.18)

¹SEN comorbidity has been removed from this model because it was not related to exclusions in Table 2.

²Ethnicity has been aggregated into White British and non-White British to maintain anonymity.

We combined Key Stages 2 and 3 due to a small number of exclusions being recorded during Key Stage 2 and potentially leading to confidentiality issues in our cohort.

The results in Table 3 demonstrate that there are some important differences in the characteristics of pupils excluded at different stages of their educational lifecourse. Whilst many of the pupil characteristics associated with school exclusions are still associated with being excluded at different stages of their education, these are, in the main, less likely to predict pupils who are only ever excluded in Years 10 and 11 (Key Stage 4). For almost every pupil characteristic, the odds of being excluded in Key Stage 4 is less than it is for the other two categories of excluded pupils. This suggests that as a group of excluded pupils, these are somewhat different to the other groups of excluded pupils. So, distinguishing between pupils who have been excluded in Key Stage 4 and at other Key Stages from pupils who have only ever been excluded in Key Stage 4 is very important. This may suggest, for example, that there are other factors associated with being excluded in Key Stage 4, possibly relating to their educational achievement or progress given the focus on GCSEs during Key Stage 4.

In terms of pupils who are excluded repeatedly across all three Key Stages, we see that these are more likely to be male, to have been consistently eligible for free school meals and to have been persistently absent from school. These could be the very high-risk factors for being excluded. It is also worth noting that the association with special educational needs is not that different between the three different groups of excluded pupils. This would suggest that whilst special educational needs are a strong predictor of being excluded from school, they do not help distinguish between pupils with repeated exclusions across their lifecourse. Although the results are slightly mixed, there is some suggestion that pupils with SEN are slightly less likely to be excluded in Key Stage 4 if they have been excluded in Key Stages 2 and 3. It is possible that either being excluded assists in the identification of special educational needs and subsequent support or that the support these pupils receive very slightly reduces the probability of being excluded again in later years.

Finally, the results in Table 2 suggest that there may be some interesting differences in practice between schools when it comes to excluding pupils. For example, pupils in faith schools appear to be more likely to be excluded in Key Stage 4 than in earlier Key Stages (again, perhaps reflecting the high stakes of GCSEs in those years and in such schools). A similar conclusion could be drawn from the results by region of Wales. Pupils in North Wales appear to be more likely to be excluded in Key Stage 4 than in other Key Stages when compared to pupils in other regions. Earlier we observed that pupils in West Wales were less likely to be excluded than pupils in North Wales. But it now appears that this was because of the relatively high rates of exclusion in Key Stage 4 in North Wales than anything to do with the pupils in West Wales.

5. Conclusions

Despite the increasing availability of better quality and accurate data on school exclusions and pupils that experience school exclusions, there remains considerable uncertainty about what these data can tell us. The reliance on official statistics, like those in Figures 1 and 2, can often give a misleading picture. Furthermore, such statistics are likely to lead to over-simplified interpretations about the underpinning causes of exclusions. For example, the Welsh Government recently published their Anti-racist Wales Action Plan (Welsh Government, 2022) that includes an action to ‘Strengthen our guidance on Exclusion from Schools and Pupil Referral Units in relation to learners who we are aware can be disproportionately subject to permanent or temporary exclusions; this includes, but is not limited to ethnic minority learners and learners with special educational needs (SEN)’ despite analysis presented in this paper which shows that Black and other minority ethnic pupils are less likely to be officially excluded than White British pupils.

Although this article has not been able to consider hidden forms of school exclusion, we have attempted to demonstrate that analyses of formal school exclusions can provide important and valuable insights if undertaken with care. We studied recent trends in school exclusions in three novel ways: first by examining more detail about each instance of exclusion pupils in Wales experience; second, by examining the characteristics of pupils who experience exclusion at different stages of their educational lifecourse; and third, by employing multivariate statistical techniques to the analysis of pupil characteristics.

This has shown, for example, that the use of temporary school exclusions appears to be changing over time in Wales, with the use of shorter periods of exclusion but used more frequently. As a result, the number of days lost due to school exclusions appears to have remained relatively high over time, overall and for each excluded pupil, despite changes in the rates of exclusion. We also begin to demonstrate the possible relationships between school exclusions and school absenteeism, which could suggest the use of other, informal, means of removing some pupils from the classroom instead of following the formal exclusion process. For now, this remains speculative but casts doubt on the meaning and interpretation of low official exclusions.

In terms of which pupils are most likely to be formally excluded from school we have been able to demonstrate that this depends on a variety of factors and their interrelationships. This also appears to depend on which stage of the educational lifecourse a pupil is at. For example, we have been able to show that pupils who have only ever experienced a school exclusion during Key Stage 4 appear to have different characteristics to pupils who experience school exclusion earlier in their schooling. They are more likely to be White British and less likely to have the three defining characteristics of all other excluded pupils (i.e., boys, eligibility for free school meals and special educational needs). We also observe differences by geography and school type, suggesting differences in the way school exclusion guidance is used.

This all leads to a much more complex picture of school exclusion and inclusion than typically understood (Daniels et al., 2022), but also highlights the limitations of relying on the individual ‘lived experiences’ of school exclusions by pupils and practitioners, which appears to underpin much of the policy narrative currently. For example, whilst the Children’s Commissioner (2020) may be correct to highlight concerns about school exclusions during the Foundation Phase (for 3 to 7-year-olds), this ignores the perhaps more problematic use of exclusions during Key Stage 4.

We would argue that a more informed approach to determining school exclusion policies must consider a wide range of factors, including changes in pupils’ circumstances, where they live, which school they go to, their specific additional learning needs, and where they are in their educational lifecourse. Although not considered in this paper, this should also consider the different reasons for, and trajectories into, being excluded from school.

Finally, the analysis presented here suggests that the relationship between attendance and exclusion should also be explored further, since this may help identify the use of informal exclusions alongside formal exclusions. This would also include the relationship between exclusions and ‘off-rolling’, particularly towards the end of compulsory education. This would begin to help address the concerns raised by Power and Taylor (2020) about the reliability of formally recorded school exclusions.

6. Acknowledgments

This study was conducted as part of Excluded Lives, a multi-disciplinary collaboration across multiple UK universities. The project employs a range of qualitative, quantitative, and policy-oriented methods to investigate school exclusions across the UK. WISERD Education Data Lab undertakes independent analysis of administrative education data, survey data and data linkage, alongside knowledge exchange and public dissemination of findings to inform national debate on some of the most contemporary and pressing educational issues facing Wales. The statistics used in this study have been approved for publication by SAIL. This does not imply Welsh Government’s acceptance of the validity of the methods used to obtain these statistics, or of any analysis of the results. Rather, they have been deemed to be non-disclosive (i.e., individual pupils cannot be identified).

7. Data Availability Statement

No new data was created during this study.

8. Disclosure Statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the ESRC under Grant ES/S015744/1 (The political economies of school exclusion and their consequences) and ESRC under Grant ES/W012227/1 (Administrative Data Research Centre Wales (Part of ADR UK)).

Notes

1 Throughout this paper we use the term ‘formal exclusions’ when referring to exclusions that have been officially recorded by schools, local authorities, and Governments. However, it is acknowledged that this does not include ‘informal exclusions’, that is exclusionary practices that go unrecorded officially.

2 All schools were closed for significant periods during 2019/20 and 202/21, so rates of exclusion, permanent and temporary, would be expected to decline over these two years. Official publication of exclusion rates in each of the four nations have continued with the same methodology for calculating exclusion rates: dividing the number of exclusions by the number of pupils. During COVID-19 these exclusion rates could have been adjusted by the number of days schools were closed for to provide a more accurate picture.

3 In our analysis we distinguish between three groups of pupils eligible for free school meals: those who were only eligible once between Year 6 and Year 11, those who were eligible for more than one year but not for all years, and those who were eligible every year between Year 6 and Year 11 (persistently eligible). For these descriptive statistics we can combine the last two categories.