Serving their communities? The under-admission of children with disabilities and ‘special educational needs’ to ‘faith’ primary schools in England

ABSTRACT

Around 28 per cent of state primary school children attend ‘faith’ establishments in England, the majority in Catholic or Church of England schools. Research suggests ‘faith’ schools tend to educate proportionally fewer children from low-income families (proxied by eligibility for Free School Meals [FSM]). This paper examines whether they also under-admit children ‘disadvantaged’ according to another key dimension: having special educational needs and/or disability (SEND). Descriptive statistics and modelling use the National Pupil Database census and span 2010–2020. Across years, ‘faith’ primary schools are less likely to include children with SEND, and less likely to admit children with SEND to the first (Reception) year. Accounting for area-level factors, indications of under-admission to Catholic schools become more pronounced. Some disproportionality for Church of England schools is explained by confounders – but even after attenuation, they remain less likely to serve children with SEND than non-‘faith’ schools. Together, FSM and SEND predict a substantively meaningful lowered likelihood of children attending ‘faith’ schools, so these schools, at the national level, seem to have become hubs of relative ‘advantage’. Findings therefore demand interrogation of whose interests these institutions serve, and of their part within the current English system.

KEYWORDS:

• ‘Faith’ schools
• special educational needs
• disabilities
• school admissions
• school composition
• disadvantage

Introduction

This paper examines whether children with ‘special educational needs’ and/or disabilities (SEND) are under-admitted to ‘faith’ primary schools in England. Firstly, it lays out the history of ‘faith’ schools, and their current positioning within the wider mainstream, state school system. It then considers existing evidence on ‘faith’ schools’ composition, and reasons to suspect they may under-serve children with SEND. New empirical analyses of national data indicate consistently that ‘faith’ schools have indeed, for the past decade, educated proportionally fewer children with SEND. Finally, findings are situated within the wider context, and questions are raised about the (dis)function of England’s primary education system.

Position, policies and practices: ‘faith’ schools in England

‘Faith’ establishments have long been part of state primary schooling. Until the late 19^th Century, ‘the churches were … the main providers of education’ (West et al., 2006, p. 621). Subsequently, the 1870 Elementary Education Act aimed ‘to fill gaps in the existing provision’ (West et al., 2006), but ‘faith schools [remained] the centre of the … system’ for some time (Walford, 2008, p. 690). By the mid-20^th century, despite other reforms, ‘nearly a third of school-aged children’ continued to attend church-provided schooling (West et al., 2006; see also Judge, 2001; Chadwick, 2001; Grace, 2001, for histories).

Notwithstanding their historical role in instituting mass education, in more contemporary years: ‘Our increasingly secular and multicultural society [has questioned] the ongoing relevance of explicitly Christian schools in the state education system of the 21st century.’ (Chadwick, 2001, p. 475). Yet schools with a statutorily designated ‘faith’ character remain prevalent to this day, and still the ‘vast majority … have a Christian faith’. In 2019, 28 per cent of state primary school children in England attended ‘faith’ schools, and this proportion has been consistent since at least 2000 (Long & Danechi, 2019).

‘Faith’ schools seem to have retained their position in recent decades due to a web of factors. One is the sustenance of their existence as part of: ‘Conservative, Labour and Coalition policy that school-type diversity, following market ideology, would improve the system’ (Courtney, 2015, p. 799). This commitment to ‘Choice and Diversity’ by the Conservatives (Department for Education, 1992) was continued by Labour (Department for Education and Skills, 2001, p. 45).

But ‘faith’ schools’ maintenance within the jigsaw of the quasi-market has not simply been underpinned by political encouragement of any diversity, neutral to diversity’s component parts. Governments and ministers have championed ‘faith’ schools, specifically, as inherently valuable.

In 2007, for example, the (then) Department for Children, Schools and Families, under Labour, issued ‘a joint vision statement … representing the shared views of the Government and the providers of publicly funded schools with a religious character’ (Department for Children, Schools and Families, 2007, p. 1). This publication’s ’unequivocal purpose … [was] for other parties to appreciate the contribution of faith schools’ (p. 3) and their ‘noble tradition’ (p. 2).

Shortly afterwards, Raey (2008) described how David Cameron, then leader of the Conservative opposition: ‘… defended parents who suddenly find religion as their child reaches school age. ”I don’t blame anyone who tries to get their children into a good school.”’ (p. 647). Then Prime Minister Tony Blair’s own individual explicit stance on ‘faith’ schools concurred with this sentiment and the Labour government’s public line (Walford, 2008).

In 2010, the Office for the Schools Adjudicator (OSA), the independent body ‘responsible for ruling on objections to and referrals about state school admission arrangements’ (https://www.gov.uk/government/organisations/office-of-the-schools-adjudicator/about) published its annual report. This included critical findings that some ‘faith’ schools were ‘in breach’ (p. 35) of the mandatory School Admissions Code, whose purpose is to, ‘ensure that all school places … are allocated and offered in an open and fair way’ (Department for Education, 2021a, p. 7).

Chief Schools Adjudicator Ian Craig was called before the Education Select Committee, and challenged vigorously on his organisation’s representation of ‘faith’ schools (Education Select Committee, 2010, Qs 13–28). Member (and future Secretary of State for Education) Damien Hinds’ questioning concluded with a request for the OSA to better portray ‘faith’ schools in the future. He asked: ‘ … that in next year’s report … the office makes strenuous additional efforts to put into context … the extent to which there is not a problem …’ (Q28; my italics). Ian Craig left his position shortly afterwards (Guardian, 2011), but the OSA has continued to publish annual reports noting schools’ ‘faith-based admission arrangements which did not meet … requirements’ (Office of the Schools Adjudicator, 2018; see also 2015) – with negligible acknowledgement or response from the Department for Education.

Instead, the narrative of ‘faith’ schools’ superiority has prevailed. Theresa May, then leader of the Conservative party, declared in 2016 that: ‘Britain has a long history of faith schools delivering outstanding education’ (May, 2016). This discourse, spanning governments and decades, continues to assert that ‘faith’ schools are effective and efficient, promoting ‘better’ academic ‘standards’ and directly enhancing pupil attainment (Department for Education, 2018, p. 13; Education Select Committee, 2010, Q23; Long & Adcock, 2016, p. 22; Walford, 2008).

‘Faith’ schools have thus consistently been positioned and protected by successive recent governments and key ministers as a public ‘good’: firstly, in terms of their contribution to the jigsaw of choice, diversity, and the quasi-market; and secondly, as purportedly superior to non-‘faith’ establishments in academic provision and ‘raising attainment’.

What, in everyday practice, distinguishes ‘faith’ schools from their non-‘faith’ state counterparts? Of most relevance to the current paper are policies and practices regarding admissions. While:

It is unlawful for … schools to discriminate against a child on the grounds of the child’s religion or belief in school admissions … faith schools are exempt and are permitted to use faith-based oversubscription criteria in order to give higher priority in admissions to children who are members of, or who practise, their faith or denomination. (Long & Danechi, 2019 p. 5)

While some (‘voluntary aided’) ‘faith’ schools administer their own admissions, and others (‘voluntary controlled’) take part in a centralised process administered by the local authority, all are thus allowed by statute to apply distinct prioritisation criteria according to ‘faith’ for which there is no equivalent for non-‘faith’ schools (see also Department for Education [2021a]). So state-funded ‘faith’ schools are ostensibly part of the choice set for all parents applying to primary school for their child. However – enshrined in national policy and legislation – parents who are evaluated by a school as demonstrating appropriate ‘faith’ have a higher chance of admission to these schools (for further discussion see Allen & West, 2009; Office for the Schools Adjudicator, 2010).

Previous research on the composition of ‘faith’ schools

This allowance of discrimination in admissions according to denoted ‘faith’ and of schools’ choice in selecting children interacts with other aspects of the quasi-marketised education system. The quasi-market serves to ‘provide strong incentives for schools to … maximise their outputs as measured by test and examination results’ (West et al., 2009). Within this system, it is therefore rational (if not, of course, moral) for schools with power in determining their own admissions to choose children whom they perceive as more likely to ‘succeed’ according to the system’s remit, rewards, and sanctions.

Previous research on the composition of ‘faith’ schools has accordingly indicated that they tend to under-admit children from families ‘disadvantaged’ according to income-level (usually proxied by recorded eligibility for Free School Meals [FSM]). Proportions of children recorded as FSM-eligible are often substantially lower in ‘faith’ than non-‘faith’ schools (Allen, 2007; Allen & Parmameshwaran, 2016; Allen & West, 2011; Andrews & Johnes, 2016; Bolton & Gillie, 2009; Long & Danechi, 2019; West & Currie, 2008).

Children eligible for FSM tend, on average, to be less likely to ‘succeed’ and to ‘attain’ highly within the structures of the education system (Hutchinson et al., 2020), and thus ‘minimise’ rather than ‘maximise’ a school’s ‘production’ and ‘outputs’. The majority of the workings of the quasi-market therefore disincentivise their admission, despite interventions such as Pupil Premium funding providing some pull in the other direction (Gorard et al., 2021). Correspondingly, the narrative of ‘faith’ schools’ superior ‘standards’ is consistently belied by evidence indicating that social selection largely accounts for their ‘better’ average academic results (Allen & Parmameshwaran, 2016; Andrews & Johnes, 2016; Bolton & Gillie, 2009; Gibbons & Silva, 2011; Long & Danechi, 2019).

Faith schools and children with SEND

This paper focuses on whether children who are ‘disadvantaged’ according to another key dimension predicting ‘success’ in the primary school system – having ‘special educational needs’ and/or disability (SEND) – are also under-admitted to and under-served by ‘faith’ primary schools in England. As Daniels et al. (2019) reiterate, the system context is one where: ‘… school accountability for the progress of their students … [has] resulted in perverse incentives for schools to not meet the needs of special educational needs students.’ (p. 1–4). The Office of the Schools Adjudicator (2014, p. 35) describes how: ‘ … many faith schools … give priority to all children of the faith before giving priority to other children not of the faith who have social, medical or physical needs.’ Thus at this most overt level, ‘faith’ schools – endorsed by national policy and legislation – may deliberately displace applications from and admissions of children with SEND in favour of those without disabilities or manifest access requirements whom they deem to be ‘of the faith’.

Evidence from recent decades on the composition by children’s documented SEND of ‘faith’ and non-‘faith’ schools begins to indicate disproportionalities echoing those according to FSM-eligibility. In 2009, Bolton & Gillie reported that, within all state primary and secondary schools: ‘1.2% of pupils at mainstream … state faith schools had statemented SEN and 15.9% unstatemented. This compares to 1.7% statemented and 18.9% unstatemented [in] schools with no religious character.’ (Bolton & Gillie, 2009).

Examining secondary rather than primary school admissions, Weldon (2018) also found suggestions that ‘faith’ schools are less likely than community schools to admit children with SEND, even controlling for potentially confounding factors (p. 35; p. 49). This echoes West et al’s. (2006) investigation of secondary schools across England (see also West & Hind, 2007). Additionally, Andrews & Johnes (2016) reported that, among children in the final year of primary school, 16.8% of those in ‘faith’ establishments were recorded with any SEND, compared to 19.7% in non-‘faith’ schools.

A potential limitation of this previous evidence is that it tends to rely on disparities in proportions of children with SEND as denoted and reported within schools. There are vast inequalities in SEND attribution between establishments. Children with equivalent needs in different schools are not equivalently ascribed (Campbell, 2021; Hutchinson, 2021; Office of the Schools Adjudicator, 2015). Therefore, it is possible that previous indications of under-representation of children with SEND in ‘faith’ schools may be due to variation in ascription: children with similar needs may in fact be served proportionally by ‘faith’ schools, without necessarily being similarly recorded in the data.

After mapping patterns in state ‘faith’/non-‘faith’ primary schools’ composition according to children’s recorded SEND, across the past decade, the current paper therefore directly addresses and circumvents this possible alternative explanation.

Data and research strategy

It does so by focussing on two groups of children, both in the first (Reception) year of primary school. Firstly, children who were recorded with SEND in the pre-school years; secondly, children who already have a higher-level Education, Health and Care Plan (EHCP; pre-2015 reforms, a Statement) – funded by the Local Authority and conferring statutory provision – at January of their Reception year.

Attribution of SEND to children in these groups should have been established outside the distortions of inconsistent ascription, measurement and recording within the statutory education system: prior to school entry. Their needs would be manifest and salient at the point of application for admission to primary school. Disproportionalities in either of these sets of children attending ‘faith’ and non-‘faith’ schools in the Reception year can thus more confidently be attributed, if apparent, to differences in intakes between ‘faith’ and non-‘faith’ schools. Examining both respective groups triangulates analyses, adding weight to conclusions. This paper thus examines specifically whether children with pre-school SEND and children with an EHCP in Reception are (respectively) less likely to attend a ‘faith’ school than other children.

In order also to assess whether under-admission persists over time and other system changes, analyses span the years 2010–2020. This incorporates amendments to both the SEND Code of Practice and the School Admissions Code (in 2014), as well as a plethora of other variations to the context within which primary schools operate.

Analyses use the National Pupil Database (NPD): a census of children in state-funded education in England. The NPD contains records for all those attending ‘faith’ and non-‘faith’ state primary schools, as well as all children attending funded pre-school (every child is eligible for a funded place from age three, for at least a school year before entering primary school). Permission for NPD use was given by the Department for Education; access was provided through the Office for National Statistics’ Secure Research Service; analyses were conducted in Stata v16.

De-identified individual-level data (from the Spring Schools Census and the Early Years Census) is matched longitudinally and cross-sectionally with school-level data from the Annual Schools Census. This results, for each child, in yearly information including:

• recorded SEND, FSM-eligibility, ethnicity, home language (EAL), birth month, gender;

• school attended, religious denomination of school, governance of school;

• area-level deprivation and rurality indicators;

• local authority (LA) in which school is situated.

The main sample each year is all children attending mainstream schools, falling into the ‘standard’ birth cohort for the year-group (September–August), whose schools are not in very small LAs (City and Isles of Scilly are removed).

Children attending special schools are not included. Note importantly, however, that there are no ‘faith’ special schools, so all children in this provision are served by the non-‘faith’ sector. Ns range from 573,446 for the cohort attending Reception in 2010 to 625,989 in 2020. Minimal variations in numbers between descriptive and regression analyses are due to very small amounts of missing data on controls.

Each focal year, analyses at the national, school and child- level investigate whether and the extent to which these mainstream ‘faith’ schools under-serve children with SEND compared to their non-‘faith’ mainstream state counterparts, and whether other factors may account for disproportionalities.

Distinction is made in school-level analyses between Church of England (CoE) and Catholic schools, as these religious denominations constitute the vast majority of ‘faith’ establishments (numbers for ‘faith’ schools of other denominations are too small for reporting in this paper). Given well-established patterns of disproportionality where children recorded as FSM-eligible are less likely to attend ‘faith’ schools, the intersection between recorded SEND and recorded FSM is considered in pupil-level models, to explore whether one accounts for the other, and whether there are additive or interactive associations for children with both characteristics, who may be conceived of as ‘doubly disadvantaged’.

Approaches to each piece of analysis – national, school-level, and pupil-level – are detailed throughout the next section, alongside findings.

Analysis and findings

The national picture

Figure 1 shows the composition of Reception places in each of the 150 (149 in 2020) large English LAs according to the ‘faith’ designation of the school attended by each pupil, at January of the years 2010, 2013, 2017, and 2020. Each LA is a bar; the bottom section represents the proportion of places in the LA taken in non-‘faith’ schools. In the LA with the most ‘faith’ places, over 60% of children are in ‘faith’ schools (far left). At the other end of the distribution are LAs where fewer than 20% of places are in ‘faith’ institutions. The national proportion of places in ‘faith’ schools has remained consistent at around 28%.

Figure 1. Composition of Reception places in each English local authority according to ‘faith’ of school attended.

Notes: N 2010 = 573,446; N 2013 = 631,460; N 2017 = 658,309; N 2020 = 625,989. Each bar is an LA. Places in mainstream schools only. City and Isles of Scilly are excluded.

Figure 2 illustrates the national proportion of children in each type of school – non-’faith’, CoE, Catholic – with recorded SEND, in each year 2010–2020. The top panel echoes previous research as it shows percentages of all primary children, from Reception to Year 6, recorded with any level of SEND in the given year. The middle panel focuses on Reception-aged children recorded with SEND prior to school entry, at pre-school, showing the proportion of children within each type of school recorded as such. The bottom panel shows the proportion of Reception children within each type of school recorded at January with a higher-level EHCP (a Statement in earlier years).

Figure 2. National proportion of children in England, in each school type, who have SEND recorded in the national pupil database.

Notes: National Ns. All children Reception-Year 6: 2010 = 3,852,875; 2011 = 3,887,917; 2012 = 3,960,513; 2013 = 4,062,013; 2014 = 4,167,883; 2015 = 4,269,836; 2016 = 4,372,564; 2017 = 4,456,992; 2018 = 4,500,371; 2019 = 4,515,760; 2020 = 4,505,681. Reception children: 2010 = 573,565; 2011 = 586,221; 2012 = 607,935; 2013 = 631,662; 2014 = 629,110; 2015 = 642,877; 2016 = 657,182; 2017 = 658,349; 2018 = 640,107; 2019 = 627,307; 2020 = 626,025. Mainstream schools only.

By every definition, and despite changes in overall trends (see Campbell 2021, for discussion) non-‘faith’ schools are, on average, each year more likely to include a higher proportion of children with SEND. Catholic schools are more likely than CoE schools to admit children with pre-school SEND to Reception, while CoE schools are more likely than Catholic schools to admit children with statutory EHCPs/Statements. But consistently, compared to non-‘faith’ schools, ‘faith’ schools include fewer children with SEND.

Disproportionalities measured at the school-level

In order better to understand this high-level national picture, analyses now investigate whether disproportionalities can be interpreted as arising directly from a tendency attributable to ‘faith’ schools to under-serve children with SEND, or whether, instead, contextual factors might account for associations. Regression modelling thus controls for factors that may co-vary with, explain, or confound relationships. For parsimony, four years across the decade are considered: children in Reception in 2010, 2013, 2017, 2020.

School-level linear regression results for each respective year are reported in Tables 1 and 2. In Table 1, the outcome is the proportion of children in a school who were recorded with any SEND in the NPD prior to entry, in pre-school. In Table 2’s respective models, the outcome is the proportion of Reception children in the school with an EHCP (a Statement in earlier years).

Table 1. School-level linear regression modelling: proportion of Reception children with SEND recorded in pre-school according to ‘faith’ denomination and admissions authority of school.

	Spec 1: ‘Faith’ and admissions	Spec 2: + school-level controls	Spec 3: + area-level controls	Spec 4: + LA fixed effects
Reception 2010
School: Non-‘faith’ (N = 348)	−0.75*	−0.12	−0.04	−0.19
LA: CoE (N = 2303)	−1.59***	−0.41**	−0.34*	−0.29*
School: CoE (N = 1879)	−1.15***	−0.20	−0.23	−0.27+
School: Catholic (N = 1632)	−0.71***	−0.26+	−0.57**	−0.63***
LA: Non-‘faith’ (N = 9295)	-	-	-	-
Constant	4.26***	2.16***	1.76***	1.72**
N schools	15580	15580	15580	15580
N LAs				150
Reception 2013
School: Non-‘faith’ (N = 1186)	0.31+	0.27	0.23	0.21
LA: CoE (N = 2209)	−1.78***	−0.45**	−0.28+	−0.28+
School: CoE (N = 1973)	−1.32***	−0.26+	−0.30*	−0.39**
School: Catholic (N = 1621)	−0.30+	0.19	−0.34*	−0.47**
LA: Non-‘faith’ (N = 8449)	-	-	-	-
Constant	4.77***	3.14***	2.66***	2.73***
N schools	15589	15589	15589	15589
N LAs				150
Reception 2017
School: Non-‘faith’ (N = 3100)	0.03	−0.15	−0.31*	−0.05
LA: CoE (N = 1882)	−1.82***	−0.66***	−0.42**	−0.38**
School: CoE (N = 2319)	−1.25***	−0.43**	−0.39**	−0.40**
School: Catholic (N = 1616)	−0.36*	−0.13	−0.71***	−0.88***
LA: Non-‘faith’ (N = 6743)	-	-	-	-
Constant	4.67***	2.51***	1.51**	1.86***
N schools	15823	15823	15823	15823
N LAs				150
Reception 2020
School: Non-‘faith’ (N = 4341)	−0.13	−0.25*	−0.36**	−0.12
LA: CoE (N = 1576)	−1.69***	−0.48**	−0.29+	−0.28+
School: CoE (N = 2629)	−1.35***	−0.53***	−0.46**	−0.51***
School: Catholic (N = 1622)	−0.22	−0.26+	−0.71***	−0.82***
LA: Non-‘faith’ (N = 5543)	-	-	-	-
Constant	4.68***	4.05***	3.18***	3.42***
N schools	15882	15882	15882	15882
N LAs				149

Notes: Linear regressions; outcome = school-level proportion Reception children with SEND recorded in pre-school. + p < 0.10; *p < 0.05; **p < 0.01; ***p < 0.0001. Source: National Pupil Database. Spec 2 adds controls for proportion of Reception children recorded as FSM eligible in school, mean age of Reception children in school, proportion of children in Reception who are girls, proportion of children in Reception with English as an additional language, N children in Reception, proportion in Reception who did not attend pre-school. Spec 3 adds controls for proportion of Reception children living in urban areas, and mean area deprivation level (IDACI) of children’s residences. Spec 4 adds fixed effects for LA in which school situated. Note that in 2010 urban/rural variable is not available so for this year only mean IDACI is added in Spec 3. Note that ‘faith’ schools of ‘other’ denominations are included as a separate category in regressions, but results are not reported here. ‘School’ means admissions are administered by the school itself; ‘LA’ that they are administered centrally by the LA.

Table 2. School-level linear regression modelling: proportion of Reception children with statutory EHCP/Statemented SEND according to ‘faith’ denomination and admissions authority of school.

	Spec 1: ‘Faith’ and admissions	Spec 2: + school-level controls	Spec 3: + area-level controls	Spec 4: + LA fixed effects
Reception 2010
School: Non-‘faith’ (N = 348)	0.10	0.14	0.13	0.04
LA: CoE (N = 2303)	−0.22***	−0.23***	−0.24***	−0.17**
School: CoE (N = 1879)	−0.19**	−0.19**	−0.19**	−0.20**
School: Catholic (N = 1632)	−0.30***	−0.30***	−0.25***	−0.20**
LA: Non-‘faith’ (N = 9295)	-	-	-	-
Constant	0.84***	1.05***	1.10***	1.10***
N schools	15580	15580	15580	15580
N LAs				150
Reception 2013
School: Non-‘faith’ (N = 1186)	0.16*	0.17*	0.18*	0.07
LA: CoE (N = 2209)	−0.22***	−0.24***	−0.23***	−0.16**
School: CoE (N = 1973)	−0.21**	−0.22**	−0.22**	−0.26***
School: Catholic (N = 1621)	−0.27***	−0.29***	−0.29***	−0.25**
LA: Non-‘faith’ (N = 8449)	-	-	-	-
Constant	0.90***	1.37***	1.38***	1.34***
N schools	15589	15589	15589	15589
N LAs				150
Reception 2017
School: Non-‘faith’ (N = 3100)	−0.08+	−0.08+	−0.07	−0.04
LA: CoE (N = 1882)	−0.28***	−0.19**	−0.18**	−0.19**
School: CoE (N = 2319)	−0.13**	−0.07	−0.07	−0.12*
School: Catholic (N = 1616)	−0.30***	−0.31***	−0.30***	−0.30***
LA: Non-‘faith’ (N = 6743)	-	-	-	-
Constant	0.91***	1.13***	1.17***	1.15***
N schools	15823	15823	15823	15823
N LAs				150
Reception 2020
School: Non-‘faith’ (N = 4341)	−0.05	−0.06	−0.05	−0.06
LA: CoE (N = 1576)	−0.23**	−0.1	−0.1	−0.11
School: CoE (N = 2629)	−0.23***	−0.15*	−0.15*	−0.21**
School: Catholic (N = 1622)	−0.31***	−0.34***	−0.33***	−0.28**
LA: Non-‘faith’ (N = 5543)	-	-	-	-
Constant	1.28***	1.92***	1.96***	1.93***
N schools	15882	15882	15882	15882
N LAs				149

Notes: Linear regressions; outcome = school-level proportion Reception children with EHCP/Statement. + p < 0.10; *p < 0.05; **p < 0.01; ***p < 0.0001. Source: National Pupil Database. Spec 2 adds controls for proportion of Reception children recorded as FSM eligible in school, mean age of Reception children in school, proportion of children in Reception who are girls, proportion of children in Reception with English as an additional language, N children in Reception, proportion in Reception who did not attend pre-school. Spec 3 adds controls for proportion of Reception children living in urban areas, and mean area deprivation level (IDACI) of children’s residences. Spec 4 adds fixed effects for LA in which school situated. Note that in 2010 urban/rural variable is not available so for this year only mean IDACI is added in Spec 3. Note that ‘faith’ schools of ‘other’ denominations are included as a separate category in regressions, but results are not reported here. ‘School’ means admissions are administered by the school itself; ‘LA’ that they are administered centrally by the LA.

Specification 1 simply shows the average differences in proportions according to school-type. Schools are distinguished not only by religion but also by the admissions administrator of the school: whether they are their own admissions authority, or whether the LA is the authority. This is because research has suggested that schools with autonomy in administering their own admissions may under-admit children with SEND (Black et al., 2019). Under academisation, increasing numbers of non-‘faith’ schools have become autonomous in this respect, alongside some CoE schools (‘voluntary aided’) – demonstrated by the Ns reported in Tables 1 and 2. The LA administers the admissions of other CoE schools (‘voluntary controlled’), while all Catholic schools control their own admissions.

Thus, results from five categories are reported: School admissions/non-‘faith’; LA admissions/CoE; School/CoE; School/Catholic; LA/non-‘faith’. The reference category is non-‘faith’ schools with LA admissions, so coefficients for each other school type indicate the average difference in proportion of children with SEND in this type of school from the LA/non-‘faith’ reference.

Specification 2 accounts for other school-level factors that may co-vary with and confound associations: proportion of Reception children recorded as FSM eligible, mean age of Reception children, proportion of children in Reception who are girls, proportion with English as an additional language, N children in Reception, proportion in Reception who did not attend pre-school. This is necessary because, for example, children recorded as FSM-eligible are more likely also to be recorded with SEND (Department for Education, 2021b; Hutchinson, 2021; Strand & Lindorff, 2021). Having controlled for these factors, remaining associations between school ‘faith’/admissions authority and proportion of children with SEND can more confidently directly be attributed to schools’ character.

Specification 3 adds controls for features of the local area of residence of pupils: the mean proportion of children living in a deprived family (IDACI) in the area, and the proportion of Reception children living in urban areas. Children attending CoE schools tend to be drawn from less deprived and more rural areas, and those attending Catholic schools from more urban areas with higher levels of deprivation. For example, in the 2020 sample, the mean IDACI recorded for areas of residence of children attending CoE schools with LA admissions was 15%; for those in non-‘faith’ schools with LA-administered admissions it was 21%; and for those in Catholic schools it was 24%. Among children attending CoE schools with admissions administered by the LA, 49% lived in rural areas, compared to 12% of those in non-‘faith’ schools with LA-administered-admissions, and 4% of those in Catholic schools. Thus, it may be that the proportion of children recorded with SEND in schools with different ‘faith’ denominations reflect aspects of the area, rather than any disproportionalities in admissions on the part of the school – for example, because the probability of being attributed SEND varies with area deprivation-level (Hutchinson, 2021).

Similarly and finally, given the uneven distribution of schools of different ‘faith’ denominations across LAs (Figure 1), Specification 4 accounts for the LA in which the school is situated, using LA fixed effects. This compares schools within the same area, controlling still for the other, lower-level area characteristics and school factors, to reveal whether ‘faith’ school status appears directly to be related to under-admission of children with SEND, after attenuating with all these variables, and making comparisons within- LA.

Table 1 shows, across years, that, compared to non-‘faith’ schools with admissions administered by the LA, both CoE schools administering their own admissions and those with LA-administered admissions tend to contain proportionally fewer Reception children with SEND recorded in pre-school. This pattern becomes less pronounced once controls are added – suggesting that contextual, covarying factors provide some explanation – but direct relationships still remain. For example, in 2020, in CoE schools administering their own admissions, the estimated average proportion of children with pre-school SEND recorded in Reception is 0.3 percentage points lower than the average proportion in non-faith schools with LA-administered admissions. For CoE schools with admissions administered by the LA, the difference is 0.5 percentage points lower. Given that in 2020 the average percentage of children with pre-school SEND across all schools is 4.5%, these are substantively meaningful disproportionalities.

Catholic schools also contain proportionally fewer children with pre-school SEND – and rather than associations being attenuated after controlling for other factors, they become starker and more apparent once area-level variables are added – particularly in more recent years. This suggests that while some of the apparent tendency of CoE schools to under-admit children with pre-school SEND is explained by contextual factors – though under-admission remains – Catholic schools’ disproportionate admission of children without SEND is actually more apparent when the characteristics of the areas in which they are situated and from which they draw children are made pertinent. In 2020, accounting for all controls, Catholic schools include on average 0.8 percentage points fewer children with SEND (to reiterate, the national average is 4.5%).

Figure 3 illustrates this with mean model-predicted proportions of children with recorded pre-school SEND in each school type. In 2020, for example, both without any controls (Spec 1) and with all controls (Spec 4) non-‘faith’ schools with LA-administered admissions contain the highest proportions of children with SEND. When other factors are not accounted for (Spec 1), CoE schools with LA-controlled admissions contain the lowest proportions (1.7 percentage points lower, on average). However, once all controls are added, Catholic schools seem most prone to relative under-admission.

Figure 3. Mean model-predicted proportion of reception children with SEND recorded in pre-school, in each school type.

Notes: Interpret with reference to Table 1, which also reports Ns. Whiskers are 95% confidence intervals around estimated marginal means.

Table 2 reports results from school-level regressions where the outcome is the proportion of Reception children with statutory EHCP/statemented SEND. Compared to non-‘faith’ schools whose admissions are administered by the LA, CoE schools that control their own admissions admit proportionally fewer of these children, even accounting for all controls, and the same is true of CoE schools with LA-controlled admissions (though this tendency is not indicated as statistically significant in 2020, so it is neither a precise nor a certain estimate for that year). Across years, Catholic schools contain fewer Reception children with statutory SEND recorded, including once school and area-level controls are accounted for. Figure 4 illustrates this with mean model-predicted proportions, showing that in most years studied, Catholic schools seem to admit proportionally the fewest children with an EHCP/Statement, and that this holds with and without attenuation by controls.

Figure 4. Mean model-predicted proportion of Reception children with statutory EHCP / statemented SEND, in each school type.

Notes: Interpret with reference to Table 2, which also reports Ns. Whiskers are 95% confidence intervals around estimated marginal means.

Disproportionalities measured at the pupil-level

As analyses at the school level showed that, overall, ‘faith’ schools on average are less likely to admit children with SEND, modelling at the pupil- level is based on a binary outcome of the child attending a ‘faith’ school for Reception, or not. This is in line with much of the previous research, which considered disproportionalities according to this higher-level, dichotomous categorisation of ‘faith’/non-‘faith’. Pupil-level modelling extends the investigation here as it allows SEND as a measure of individual-level ‘disadvantage’ to be considered alongside and interacted with the measure of ‘disadvantage’ previously most studied in relation to ‘faith’ school admissions: whether a pupil’s family is low-income, as proxied by recorded eligibility for FSM.

Tables 3 and 4 thus report results from linear probability modelling, chosen in order to compare coefficients across specifications (Mood, 2010). Specification 1 in these pupil-level models reports the difference in probability of a child with SEND attending a ‘faith’ school for Reception compared to a child without SEND, and specification 2 adds FSM (vs non-FSM) to explore whether both factors are related when the other is accounted for. In specification 3, SEND and FSM are interacted, to investigate whether relationships between each variable and the probability of attending a ‘faith’ school depend on the other characteristic.

Table 3. Pupil-level linear probability modelling: chances of attending Reception in a ‘faith’ school vs a non-‘faith’ school according to whether a child had recorded SEND in pre-school and/or FSM in Reception.

	Spec 1: Pre-school SEND	Spec 2: + FSM	Spec 3: SEND *FSM	Spec 4: + Child-level controls	Spec 5: + Area-level controls	Spec 6: + LA fixed effects
Reception 2010
Pre-school SEND	−0.06***	−0.05***	−0.05***	−0.05***	−0.04***	−0.03***
No pre-school SEND	-	-	-	-	-	-
FSM		−0.09***	−0.09***	−0.09***	−0.06***	−0.07***
No FSM		-	-	-	-	-
Pre-school SEND*FSM			0.02**	0.01+	0.01	0.01
Constant	0.29***	0.31***	0.31***	0.33***	0.26***	0.26***
N children	571,331	571,331	571,331	571,331	571,331	571,331
N LAs						150
Reception 2013
Pre-school SEND	−0.05***	−0.04***	−0.05***	−0.04***	−0.03***	−0.03***
No pre-school SEND	-	-	-	-	-	-
FSM		−0.09***	−0.09***	−0.09***	−0.06***	−0.06***
No FSM		-	-	-	-	-
Pre-school SEND*FSM			0.03***	0.02***	0.02**	0.01*
Constant	0.29***	0.30***	0.30***	0.33***	0.53***	0.54***
N children	629535	629535	629535	629535	629535	629535
N LAs						150
Reception 2017
Pre-school SEND	−0.05***	−0.04***	−0.05***	−0.04***	−0.03***	−0.03***
No pre-school SEND	-	-	-	-	-	-
FSM		−0.07***	−0.07***	−0.08***	−0.05***	−0.05***
No FSM		-	-	-	-	-
Pre-school SEND*FSM			0.03***	0.02**	0.02*	0.01
Constant	0.28***	0.29***	0.29***	0.31***	0.26***	0.23***
N children	656902	656902	656902	656902	656902	656902
N LAs						150
Reception 2020
Pre-school SEND	−0.05***	−0.04***	−0.04***	−0.03***	−0.02***	−0.02***
No pre-school SEND	-	-	-	-	-	-
FSM		−0.06***	−0.06***	−0.06***	−0.03***	−0.04***
No FSM		-	-	-	-	-
Pre-school SEND*FSM			0.01*	0.01	−0.00	−0.01
Constant	0.28***	0.29***	0.29***	0.32***	0.24***	0.24***
N children	624996	624996	624996	624996	624996	624996
N LAs						149

Notes: Source: National Pupil Database. Spec 4 adds controls for children’s birth month, gender, ethnicity, home language. Spec 5 adds controls for whether area of residence is urban/rural and area deprivation level (IDACI). Spec 6 adds fixed effects for LA in which school situated. + p < 0.10.

*p < 0.05; **p < 0.01; ***p < 0.0001. Note that in 2010 urban/rural variable is not available so for this year only IDACI is added in Spec 5. Linear probability models; outcome = attending a ‘faith’ school.

Table 4. Pupil-level linear probability modelling: chances of attending Reception in a ‘faith’ school vs a non-‘faith’ school according to whether a child has statutory EHCP/Statemented SEND and/or FSM recorded in January of their Reception year.

	Spec 1: EHCP/Statement SEND	Spec 2: + FSM	Spec 3: SEND *FSM	Spec 4: + Child-level controls	Spec 5: + Area-level controls	Spec 6: + LA fixed effects
Reception 2010
EHCP/Statement	−0.06***	−0.06***	−0.06***	−0.06***	−0.06***	−0.05***
No EHCP/Statement	-	-	-	-	-	-
FSM		−0.09***	−0.09***	−0.09***	−0.06***	−0.07***
No FSM		-	-	-	-	-
EHCP/Statement*FSM			0.01	0.01	0.00	0.01
Constant	0.29***	0.31***	0.31***	0.32***	0.26***	0.26***
N children	571,331	571,331	571,331	571,331	571,331	571,331
N LAs						150
Reception 2013
EHCP/Statement	−0.05***	−0.05***	−0.05***	−0.05***	−0.05***	−0.04***
No EHCP/Statement	-	-	-	-	-	-
FSM		−0.09***	−0.09***	−0.09***	−0.06***	−0.06***
No FSM		-	-	-	-	-
EHCP/Statement*FSM			0.03+	0.03*	0.02	0.02
Constant	0.29***	0.30***	0.30***	0.32***	0.53***	0.54***
N children	629535	629535	629535	629535	629535	629535
N LAs						150
Reception 2017
EHCP/Statement	−0.05***	−0.04***	−0.05***	−0.04***	−0.04***	−0.04***
No EHCP/Statement	-	-	-	-	-	-
FSM		−0.07***	−0.07***	−0.08***	−0.05***	−0.05***
No FSM		-	-	-	-	-
EHCP/Statement*FSM			0.03+	0.02	0.01	0.01
Constant	0.28***	0.29***	0.29***	0.31***	0.26***	0.23***
N children	656902	656902	656902	656902	656902	656902
N LAs						150
Reception 2020
EHCP/Statement	−0.04***	−0.04***	−0.04***	−0.04***	−0.03***	−0.03***
No EHCP/Statement	-	-	-	-	-	-
FSM		−0.06***	−0.06***	−0.06***	−0.03***	−0.04***
No FSM		-	-	-	-	-
EHCP/Statement*FSM			0.01	0.01	−0.00	−0.00
Constant	0.28***	0.29***	0.29***	0.31***	0.24***	0.24***
N children	624996	624996	624996	624996	624996	624996
N LAs						149

Notes: Source: National Pupil Database. Spec 4 adds controls for children’s birth month, gender, ethnicity, home language. Spec 5 adds controls for whether area of residence is urban/rural and area deprivation level (IDACI). Spec 6 adds fixed effects for LA in which school situated. + p < 0.10.

*p < 0.05; **p < 0.01; ***p < 0.0001. Note that in 2010 urban/rural variable is not available so for this year only IDACI is added in Spec 5. Linear probability models; outcome = attending a ‘faith’ school.

Specification 4 adds controls for birth month, gender, ethnicity, and home language, because propensity to be recorded with SEND varies according to these factors (Campbell, 2021; Department for Education, 2021b; Hutchinson, 2021; Strand & Lindorff, 2021). Like the school-level models, Specification 5 goes on to account also for deprivation level of the child’s local area of residence (IDACI), whether it is urban or rural, and Specification 6 adds fixed effects for the LA in which the school is situated.

Table 3 and the model predicted probabilities reported in Figure 5 show that, in every year, both being recorded with SEND in the pre-school years and being reported as FSM-eligible in Reception are associated with lessened chances of attending a ‘faith’ school. Once all controls are added, there is some attenuation of these relationships, but they remain substantively and statistically significant – and, with all controls, there is little interaction between the two factors. Children with pre-school SEND are less likely to be admitted to a ‘faith’ school regardless of whether they are recorded with FSM, and vice versa: there appears to be a ‘double disadvantage’. For example, in 2020, averaging across all other controls, a child with SEND and FSM is estimated to have a 22% chance of attending a ‘faith’ school, compared to a 29% chance for a child with no FSM eligibility nor pre-school SEND recorded.

Figure 5. Mean model-predicted probability of attending Reception in a ‘faith’ school, according to whether a child is recorded with pre-school SEND and as FSM-eligible.

Notes: Interpret with reference to Table 3, which also reports Ns. Whiskers are 95% confidence intervals around estimated marginal means.

Table 4 and Figure 6 show a similar overall finding when considering whether children have an EHCP/Statement. Compared to children without this statutory SEND, these children are less likely to be admitted to ‘faith’ schools – again, whether or not they are also from low-income families, as proxied by FSM. For example, in 2020, accounting for and averaging across all other controls, children with an EHCP and recorded FSM eligibility in Reception have an estimated 21% chance of attending a ‘faith’ school, compared to a 29% chance for children without an EHCP or FSM.

Figure 6. Mean model-predicted probability of attending Reception in a ‘faith school, according to whether a child is recorded with statutory EHCP / statemented SEND and as FSM-eligible.

Notes: Interpret with reference to Table 4, which also reports Ns. Whiskers are 95% confidence intervals around estimated marginal means.

Additional analyses

Sensitivity checks using logistic rather than linear regressions yield substantively equivalent results (see Appendix A for an example). Pupil-level analyses are also repeated for children who attended the year of pre-school directly preceding Reception entry in non-school nurseries (in private, voluntary and independent sector settings). This comprises nearly half the sample in 2010, rising to almost 60% in 2020. The check ensures that ascription of SEND to these sub-samples took place outside of the statutory school system. Results echo those for the complete, main samples.

Discussion

This paper set out to explore whether children with ‘special educational needs’ and/or disabilities (SEND) are under-admitted to ‘faith’ primary schools in England, in a context where these schools remain prevalent, and where they have been positioned and defended by governments and ministers as essential to ‘choice and diversity’, and as providing a public ‘good’ and a valuable service within the state education system.

New national empirical analyses spanning the years 2010–2020 find consistently that both Church of England (CoE) and Catholic schools contain proportionally fewer children with SEND. Though regression modelling suggests that some of the stark disproportionality for CoE schools may be explained by contextual factors, significant differences in their intake remain, compared to non-‘faith’ schools. When area-level factors are accounted for, the disproportionate tendency of Catholic schools to under-admit children with SEND recorded in pre-school becomes more pronounced. Both low-income (proxied by FSM) and SEND are related to lessened chances of attending a ‘faith’ school, independently of one another. So, the known tendency of ‘faith’ schools to under-serve children from low-income families seems to be reflected in their tendency also to under-admit children ‘disadvantaged’ according to disability.

That ‘faith’ schools’ part in providing ‘choice and diversity’ within the quasi-marketised system results in their positioning as hubs of relative ‘advantage’ according to both these dimensions is unsurprising, given the weight of previous evidence. But this continues to be ignored and obscured in policy-making. In 2008, West & Currie wrote that the policy of:

… school diversity in England … increases potential choice for some groups of parents … relatively privileged when making their preferences for schools by having a wider range of school options. As a result of synergies between their advantages and those of their children and current school diversity and admissions practices, they are also more likely to have a greater chance of gaining access to particular schools. (West & Currie, 2008, p. 243)

It is important to note that this paper and its findings do not provide any commentary on whether the tendency of ‘faith’ schools to under-admit children with SEND is inherently a positive or a negative thing for individual children. Pull factors to non-‘faith’ schools that serve children with disabilities may well contribute to the patterns found here, alongside push factors from ‘faith’ schools. Nor does this paper provide information on individual ‘faith’ (or non-‘faith’) schools that buck the overall trends: it reports national averages. Rather, this paper questions the existence, function, and contribution of ‘faith’ schools at the system-level: particularly important given ‘faith’ schools’ general positioning in policy discourses as ‘raising standards’ through superior academic provision.

This assertion is called into question once more by results here, given that they highlight the selectivity of ‘faith’ primary schools in under-serving children with disabilities, as well as reiterating their known under-admittance of children recorded as eligible for FSM. Any difference in ‘attainment’ outcomes is likely to be related to this selectivity, and the very notion of ‘standards’ is called into question if these ‘standards’ result in exclusion of children ‘disadvantaged’ according to both dimensions. The finding that Catholic schools particularly appear disproportionately to under-serve children with SEND once area-level factors are accounted for calls explicitly into question the policy that it is: ‘right to encourage faith communities – especially those with a proven record of success, like the Catholics – to play their full part in building the capacity of our schools’ (May, 2016).

On finding an under-serving of FSM-eligible children, Allen and West (2009) posited a ‘distortion of mission’ of some ‘faith’ schools, ‘given that they were originally set up to educate the poor’ (p. 471). If, overtly or tacitly, the function of ‘faith’ schools is now not, in fact, to serve the whole community; nor to include those children disadvantaged by low-income; if it is to exclude some children with disabilities; if practice has mutated, in the context of the quasi-market and its incentives and disincentives – to become exclusionary – this surely is not desirable or justifiable. How can over-admission of children advantaged by family income or lack of disability be defended as in keeping with a ‘noble tradition’ (Department for Children, Schools, and Families, 2007)?

Conclusion

Findings in this paper indicate that, at the national average level, ‘faith’ primary schools appear to serve as hubs of relative ‘advantage’, disproportionately serving children from more affluent families, and children less likely to have SEND. Particularly given that these characteristics are key, substantial predictors of ‘success’ and ‘attainment’ within the education system, this demands substantial, transparent policy consideration and, potentially, reform. What part do ‘faith’ establishments play within the current school system? Whose interests are they more likely to serve? And what does the seeming overconcentration of ‘advantaged’ children within ‘faith’ schools tell us about the workings of the marketised system premised upon ‘choice and diversity’?

Acknowledgements

This work was produced using statistical data from ONS. The use of the ONS statistical data in this work does not imply the endorsement of the ONS in relation to the interpretation or analysis of the statistical data. This work uses research datasets which may not exactly reproduce National Statistics aggregates. Thanks to the Department for Education for the use of the data, and to London School of Economics IT colleagues for facilitating and supporting access. Thanks to Ruth Lupton, Kitty Stewart, Anne West, and an anonymous reviewer, for helpful comments and suggestions.

Disclosure statement

No potential conflict of interest was reported by the author.

Data availability statement

The data that support the findings of this study are available from the Department for Education. Restrictions apply to the availability of these data, which were used under licence for this study. Data are available from https://www.gov.uk/guidance/how-to-access-department-for-education-dfe-data-extracts with the permission of Department for Education.

Additional information

Funding

This work was supported by a British Academy Postdoctoral Fellowship, under grant [PF2\180019].

Notes on contributors

Tammy Campbell

Tammy Campbell completed a British Academy Postdoctoral Fellowship in 2023. She was based in LSE’s Centre for Analysis of Social Exclusion. Her PhD in Quantitative Social Science for Social Policy was awarded in 2016 by the UCL Institute of Education. Previously Tammy was a Government Social Researcher, in the (then) Department for Children, Schools and Families; before that, she worked with children and young people as a practitioner. Tammy’s main interests concern the social, structural, educational and psychological processes shaping young lives. Her research has spanned inequalities in the pre-school years, ‘ability’ grouping, bias and stereotyping in assessment, the part played by ‘faith’ schools in stratification, the workings of the system intended to serve children with special educational needs and / or disabilities, and peer effects. Tammy recenty moved to the Education Policy Institute as Director for Early Years.

Appendix A.

Table A1: Pupil-level logistic modelling: chances of attending Reception in a ‘faith’ school vs a non-‘faith’ school according to whether a child had recorded SEND in pre-school and / or FSM in Reception

	Spec 1: Pre-school SEND	Spec 2: + FSM	Spec 3: SEND*FSM	Spec 4: + Child-level controls	Spec 5: + Area-level controls	Spec 6: + LA control
Reception 2020
Pre-school SEND	0.79***	0.82***	0.81***	0.84***	0.89***	0.87***
No pre-school SEND	1	1	1	1	1	1
FSM		0.74***	0.74***	0.72***	0.83***	0.79***
No FSM		1	1	1	1	1
Pre-school SEND*FSM			1.06	1.01	0.97	0.95
N children	624996	624996	624996	624996	624996	624996
N LAs						149

Source: National Pupil Database. Spec 4 adds controls for children’s birth month, gender, ethnicity, home language. Spec 5 adds controls for whether area of residence is urban/rural, and area deprivation level (IDACI). Spec 6 adds LA in which school is situated as a control. ***p<0.0001. Logistic regressions; outcome = attending a faith school. Table reports odds ratios.

Figure A1. Mean model-predicted probability of attending Reception in a ‘faith’ school, according to whether a child is recorded with pre-school SEND and as FSM-eligible.

Notes: Interpret with reference to Table A1, which also reports Ns. Whiskers are 95% confidence intervals around estimated marginal means.