British Educational Research Journal Vol. 33, No. 2, April 2007, pp. 155–178
Secondary modern schools: are their pupils disadvantaged? Rosalind Levacˇic´*a and Alan J. Marshb a
Institute of Education, University of London, UK; bNottingham Local Education Authority, UK (Submitted 10 May 2004; resubmitted 25 January 2005; conditionally accepted 31 May 2005; accepted 4 October 2005)
There are still 10 English local educational authorities (LEAs) that are wholly selective and a further 10 with some grammar and secondary modern schools. This article examines the academic performance of pupils in secondary modern schools and the funding of these schools using national data sets matching pupils’ performance at Key Stage 2 and General Certificate of Education (GCSE) as well as data on funding from Section 52 statements. Students in secondary modern schools gained one less grade on average than equivalent students in comprehensive schools while grammar school pupils obtained five grades more. After taking account of the cost factors and grant entitlements that influence funding per pupil, secondary modern schools in the years 2000/01–2002/03 were funded around £80 less per pupil while grammar school pupils received over £100 more per pupil compared to comprehensive schools. Secondary modern schools were more likely to be in financial deficit than comprehensive and particularly grammar schools. Thus, students are academically disadvantaged by attending secondary modern schools, which in most selective LEAs do not receive sufficient additional funding to offset the depressing effects on attainment of the increased social segregation arising from a selective system.
Introduction The deleterious effect of social disadvantage on student attainment is one of the key findings from the Programme for International Student Assessment (PISA) surveys of 2000 and 2003, which report the double impact of students’ own family background and the socio-economic status of their school’s student body on attainment. The strength of this relationship differs between countries. Some countries (Canada, Finland, Hong Kong-China, Iceland, Japan, Korea and Sweden) had above average performance in literacy and below average impact of social status on student performance (Organisation for Economic Cooperation and Development *Corresponding author: Institute of Education, University of London, 20 Bedford Way, London, WC1H 0AL, UK. Email:
[email protected] ISSN 0141-1926 (print)/ISSN 1469-3518 (online)/07/020155-24 # 2007 British Educational Research Association DOI: 10.1080/01411920701208209
156 R. Levacˇic´ and A. J. Marsh [OECD], 2003). There are similar findings for mathematics (OECD, 2004). For both literacy and mathematics surveys there is also a group of countries with above average social segregation in schools (in particular Belgium, Germany and Hungary), which is associated with a stronger impact of social disadvantage on student attainment. ‘Among the ten countries with the most pronounced socio-economic segregation observed in PISA, all carry out selection procedures that channel students into different streams of secondary education before or at the age of assessment’ (OECD, 2003 p. 220). In countries with a high degree of social segregation in schools the variance in student literacy attainment between schools was around 60% of the total variance in student performance, compared to 10% or less in Scandinavian countries, where social differences between schools’ intakes are much less. The UK’s between-school variance in literacy scores was 21%, though the social background of students had an above average impact on attainment. The UK also appears in PISA as a country with a single track (i.e. non-selective) secondary system, though the country actually consists of four distinct systems, one of which, Northern Ireland, is wholly selective and another, England, still has around 20 local education authorities (LEAs) which are wholly or partially selective. In the light of the PISA findings that ‘where there is a high degree of segregation along socioeconomic lines, students from disadvantaged socio-economic backgrounds do worse’ (OECD, 2003, p. 223), this article focuses on ‘secondary modern’ schools in England. These are schools in selective English LEAs, which recruit their intake from pupils who fail a grammar school entry test. As the current debate about standards, inequality and diversity in secondary education and the comparative performance of different types of secondary school gathers pace (Department for Education and Skills [DfES], 2001, 2002a; Edwards & Tomlinson, 2002; Office for Standards in Education [Ofsted], 2002; House of Commons Education and Skills Committee, 2003), the position of secondary modern schools and their pupils deserves greater attention. The purpose of this article is to present evidence on the performance and resourcing of secondary modern schools in England and the consequent disadvantages incurred by young people who attend these schools. Most of the analysis is of data from English national data sets, but we also draw on a survey of school costs undertaken as part of research conducted on behalf of the Buckinghamshire Upper Schools Forum (Levacic et al., 2002), which is campaigning for increased local authority funding for upper (i.e. secondary modern) schools. While the article adds to existing evidence on the relative academic performance of grammar and secondary modern schools (Gallagher & Smith, 2000; Schagen & Schagen, 2003), there has been little, if any, published evidence on the funding of secondary modern schools relative to other types. Our analysis falls into two main parts. The first part is a value-added analysis of General Certificate of Secondary Education (GCSE) results from a national data set for English secondary schools, with Key Stage 2 national tests as the measure of prior attainment. We find that pupils at secondary modern schools achieved slightly lower grades at GCSE compared to those in comprehensive schools, while pupils at grammar
Are secondary modern pupils disadvantaged? 157 schools on average did considerably better. The second part of the analysis examines the financing of English secondary schools. Despite some evidence that secondary modern pupils have greater learning needs, secondary modern schools were funded less per pupil than comprehensive and grammar schools in the three years examined. This comparison allows for differences costs due to variations in schools’ size, region, age range of students, and indicators of learning need, such as the percentage of students eligible for free school meals and with special educational needs. Background: selective education in England The issue of selection for grammar or secondary modern schools has been active since the introduction in 1945 of ‘secondary education for all’ by the 1944 Education Act, and has been strongly contested by a number of lobby groups, such as the National Grammar Schools Association and the Socialist Educational Association. The steer towards comprehensive education was accelerated by issuing the non-statutory Circular 10/65 (Department for Education and Science, 1965) which requested English and Welsh LEAs to provide plans for comprehensive education. Changes of government in the early 1970s and the serious economic condition of the country slowed the drive for comprehensive education. A survey in 1975 indicated that only 20 of the English and Welsh LEAs were ‘truly comprehensive’ and a quarter of all 10 year-olds still sat the 11-plus examination (Simon, 1991, p. 439). In an attempt to accelerate the process, the Labour Government passed legislation within the 1976 Education Act to require those LEAs that had not responded to the 1965 Circular to do so. In 1979 a new Conservative government was elected and repealed the 1976 Act. Over the next 18 years comprehensive education was subjected to significant redefinition as a result of policies designed to promote ‘choice’ and ‘diversity’, with the effect that the selective system continues today to be well entrenched in 10–20 LEAs. The provision of the 1998 School Standards and Framework Act for a ballot of parents to remove grammar school selection has not so far effected any further change. In 2003 there were 248 schools in England or 7% of all secondary schools1 designated as ‘secondary modern’2 by the Autumn Package GCSE/GNVQ (General Certificate of Secondary Education/General National Vocational Qualifications) benchmark information (DfES, 2003a). Although the ‘secondary modern’ schools are distributed between 20 LEAs, only 10 of these LEAs are classed by the DfES as wholly selective, i.e. LEAs subject to a possible grammar school ballot because at least 25% of secondary pupils are in grammar schools. As Table 1 shows, in 2002 five of the wholly selective LEAs were under Conservative control, three under Labour and two under Liberal Democrat control. A major factor in predisposing a selective system to disadvantage children from less advantaged families, or who are at or below average in academic ability, is concentrating them in schools with other less able and less advantaged children. In England only 2% of pupils in grammar schools are eligible for free school meals (FSM) compared to 14% in secondary moderns. This is less than the average 18% of
158 R. Levacˇic´ and A. J. Marsh Table 1. LEAs that are defined by the DfES as wholly selective Selective LEAs
LEA type
Council control
Bexley Buckinghamshire Kent Lincolnshire Medway Slough Southend Sutton Torbay Trafford Total
London Upper Tier Upper Tier Upper Tier Unitary Unitary Unitary London Unitary Metropolitan
Lab Con Con Con Con Lab Con Lib Dem Lib Dem Lab (min)
No. of Percentage of No. of No. of pupils in secondary secondary grammar grammar schools aged pupils modern schools schools 17,696 34,207 96,672 46,236 20,486 8,158 12,024 15,398 9,024 16,074 275,975
11 20 71 48 13 7 8 9 5 11 203
4 13 33 15 6 4 4 5 3 7 94
25 44 32 27 30 45 33 32 31 30 33
Sources: DfES 2001 Annual Schools Census dataset, and DfES (2003a) Pupil Level Annual Schools Census.
pupils eligible for free school meals in comprehensive schools because almost all selective LEAs are in areas of above average socio-economic status. As Table 2 shows, the percentage of pupils eligible for FSM in secondary modern schools in the 10 selective LEAs ranges from 26% in Trafford to 11% in Buckinghamshire and Sutton and, for grammar schools, from 5% in Slough to 1% in Buckinghamshire and Sutton. Only Trafford has a higher percentage of pupils eligible for FSM than the average in non-selective LEAs. A higher proportion of students in secondary modern schools have special educational needs (SEN). As can be seen from Table 3, 3.7% of pupils in secondary Table 2. Distribution by school type of socially disadvantaged pupils as measured by eligibility for free school meals: English secondary schools (January 2001) Selective LEAs
LEA
Secondary modern
Grammar
Bexley Buckinghamshire Kent Lincolnshire Medway Slough Southend Sutton Torbay Trafford All selective LEAs Non-selective LEAs
13.2 7.1 10.2 10.1 11.7 17.2 12.2 7.6 13.1 20.3 12.3 19.8
16.5 10.9 14.2 12.6 15.4 24.2 18.0 11.1 18.6 25.7 14.3
3.5 1.2 2.2 1.9 3.1 5.1 2.4 1.2 3.8 4.0 2.4
Source: DfES 2001 Annual Schools Census dataset.
Are secondary modern pupils disadvantaged? 159 modern schools had statements of SEN compared to 0.1% in grammar schools and 2.7% in comprehensive schools, while the percentages of pupils with SEN but without statements was 24.4, 3 and 18.6% respectively. Another indicator of the additional educational needs of secondary modern pupils and the costs these impose on schools is the rate of pupil exclusions. Permanent exclusions were considerably higher for secondary modern schools (2.5 per 1000 pupils) and comprehensive schools (2.3 per 1000 pupils) than for grammar schools (0.3 per 1000 pupils). Thus, the facts demonstrate, as one would expect, that selection is associated with increased concentration of socially and educationally disadvantaged pupils in particular schools—the secondary moderns. Since the socio-economic and prior attainment characteristics of the peer group (the ‘compositional effect’) influence individual children’s academic attainment, a key issue in assessing the relative performance of selective and non-selective school systems is whether the narrower educational mission of secondary modern compared to comprehensive schools can offset the negative impact of a more disadvantaged student body. It is this issue that we focus upon in summarising the existing evidence on the relative performance of secondary modern and grammar schools. The academic performance of secondary modern and grammar school pupils: existing evidence A rational assessment of the relative advantages and disadvantages of selective school systems rests on weighing up evidence on their efficiency and equity compared to non-selective systems. Since a selective school system inherently increases social segregation in schools, its main justification has to be that of greater efficiency (Gallagher & Smith, 2000). A selective system is more efficient than a non-selective one if students on average attain more highly because they are taught in schools that focus on a narrower range of learning needs than non-selective schools, which have to serve the whole ability range. A selective system could be more efficient than a comprehensive one even if secondary modern pupils attained less than they would in the latter system. This would be the case if pupils selected for grammar schools gain sufficiently higher attainment compared to a comprehensive system to outweigh the reduced attainment for secondary modern pupils so that selective systems produce better academic results on average than comprehensive systems. However, if Table 3. Pupils with special educational needs and excluded pupils: by school type for English secondary schools (January 2001) School type
Secondary modern Grammar Comprehensive England: all schools
Percentage of pupils Percentage of pupils with SEN Permanent exclusions with statements % but without statements % per 1000 pupils 3.7 0.1 2.7 2.7
24.4 3.0 18.6 18.3
Sources: Annual Schools Census (2001) and Ofsted (2002).
2.5 0.3 2.3 2.1
160 R. Levacˇic´ and A. J. Marsh secondary modern pupils perform worse than those of equivalent ability and social background in comprehensive schools then a selective system is clearly inequitable. Policy-relevant research therefore seeks to provide evidence on the efficiency and distributional impacts of selective systems of schooling. A valid comparison of the performance of different schools must control for students’ prior attainment before entry to the schools whose relative performance is then assessed in terms of the progress pupils make between given phases of education. Two main types of data have been analysed in this research: crosssectional data of pupil cohorts in a particular year in selective and non-selective schools and systems, and longitudinal data from cohort studies of people from childhood to adulthood. A brief review of previous research on the impact of selection on pupil performance by Schagen and Schagen (2003) in this journal included an earlier review of the literature by Crook et al. (1999), three unpublished papers by Jesson (1999, 2000, 2001) and a study of GCSE performance in Northern Ireland by Shuttleworth and Daly (2000). The tentative conclusion of Crook et al.’s review was that once pupil background variables were controlled for there was little difference between selective and non-selective systems’ overall results. Jesson’s analysis used Key Stage 3 as the prior attainment measure for GCSE and reported slightly better performance by comprehensive systems. In contrast, Shuttleworth and Daly’s (2000) study, which was necessarily restricted to comparing grammar and secondary modern schools only as it analysed data from Northern Ireland, found that a student of given social background and prior attainment at a grammar school obtained 16 more points total score at GCSE compared to a pupil with the same characteristics at a secondary modern school. A DfES (2002a) study found no conclusive evidence for significant differences in pupil progress in LEAs classified into three selection types—no selection, intermediate selection and selective. The exceptions were pupils with Key Stage 2 prior attainment of level 4 or above, who made the most progress over Key Stage 3 in selective LEAs, and at Key Stage 4 where pupils in schools in non-selective LEAs gained 1–2 GCSE points more than in selective LEAs. The DfES (2002a) report did not use multilevel modelling, unlike two studies by researchers from the National Foundation for Educational Research (NFER). Schagen and Schagen (2003), using Qualifications and Curriculum Authority (QCA) national data sets for around 400,000 pupils,3 found that a pupil with an average score of 4.5 at Key Stage 2 made roughly half a level more progress to Key Stage 3 in a grammar school than in a comprehensive school. More able pupils with level 5 or higher at Key Stage 2 did little better in a grammar school. There was only a small difference in favour of grammar schools from Key Stage 3 to GCSE, and no difference for pupils of highest ability. Selective systems, measured as the percentage of pupils selected in an LEA, performed slightly better at Key Stage 3 because of the impact of grammar school attendance on the least able of the grammar school pupils. A later multilevel study by Benton et al. (2003)4 found a positive and relatively large grammar school effect from Key Stage 2 to Key Stage 3 and from Key Stage 2 to
Are secondary modern pupils disadvantaged? 161 GCSE. The lowest ability pupils in grammar schools achieved 7 GCSE total score points (grades) more than they would have done in a comprehensive school. The grammar school advantage declined with ability and was non-existent for highability pupils (those with level 5 and above at Key Stage 2). A number of studies reporting the effects of selective school attendance have analysed data from the National Child Development Study (NCDS) of children born in the same week in March 1958. Steedman (1980), controlling for the background characteristics of pupils, reported no statistically significant differences in pupils’ progress in selective and comprehensive school systems. In a later study Steedman (1983) found that the average difference in value added between grammar and comprehensive school pupils was 0.92 O level grades and that between comprehensive and secondary modern schools 0.28 grades. Later studies using NCDS data have reported on the effects of grammar school attendance on qualifications controlling for family background and cognitive attainment at the ages of 7 and/or 10. Feinstein and Symons (1999) found that attending a grammar school added 7.6 points to O level/CSE (Certificate of Secondary Education) scores while being at a secondary modern cut them by 2 (compared to comprehensive school). Dustmann et al. (2003) reported that grammar school attendance was associated with two more O level subject passes compared to being at a secondary modern school. The main purpose of GalindoRueda’s and Vignoles’ (2004) study was to test whether individuals benefited from being educated in selective areas in the 1970s, controlling for ability at age 7, a range of family variables and socio-economic indicators for area of residence. Four measures of outcome were employed—maths score at age 16, years of schooling, Alevel and highest qualification. Grammar school attendance resulted in higher outcomes for all ability groups, while low or average ability students in secondary moderns did no worse compared to comprehensives. There appeared to be no effect on average outcomes of being educated in a selective area but there was a positive effect for the highest ability quintile.5 Both the school cohort studies and the NCDS studies generally find that grammar school attendance improves academic results compared to comprehensive school attendance and secondary modern schools worsen them, though the latter result is more equivocal. The school cohort studies indicate that it is middle ability pupils who benefit most from grammar school and do relatively worse in comprehensives. The NCDS studies do not report this finding. They are, however, reporting on the effect of a well-established selective system relative to a newly created comprehensive system in the 1970s—whereas very different conditions in comprehensive schools prevailed by the late 1990s, in particular the much greater importance attached to examination performance. Comparison of pupil performance from Key Stage 2 to GCSE in secondary modern, grammar and comprehensive schools Our analysis is similar to Benton et al.’s but uses the first large-scale national data set at pupil level provided by the QCA for GCSEs sat in 2001 matched with prior
162 R. Levacˇic´ and A. J. Marsh attainment data for Key Stage 2 (KS2) in 1996. Our analysis also employs multilevel modelling. The advantage of this method is that it allows for the common effect of each school on the attainment of its pupils and thus removes the statistical bias present in a statistical model which ignores the fact that pupils are nested in schools6 (Goldstein, 1995; Kreft & De Leeuw, 1998; Schagen & Schagen, 2003, p. 580). The QCA matched data set includes the National Curriculum test and examination results, as well as age, gender and school7of over 330,000 state school pupils. A range of school context variables (listed below) was obtained from the Register of Educational Establishments for 2000 and from the Annual Schools Census for the years 1997 to 2001.8 Two measures of pupils’ GCSE/GNVQ attainment are used: (i) GCSE/GNVQ total points score (derived as usual by giving an A* grade 8 points, A57, B56, C55, D54, E53, F52 and G51 and adding up the grades the pupil obtained); and (ii) the probability that a pupil obtains 5 or more A*–C GCSE/GNVQ passes. The explanatory variables for both measures of GCSE/GNVQ attainment are prior attainment at KS2 (average marks for English, mathematics and science),9 age and gender. At school level we included the percentage of girls, pupils eligible for FSM, with SEN statements and who were white, as well as the pupil–teacher ratio to take account of differences in resourcing. All the school-level continuous variables were averaged over the five years 1996/97–2000/01 to reflect the context of the school while the pupils were progressing from Year 7 to Year 11. Descriptive statistics are given of the main continuous variables in Table A1 (Appendix). A set of dummy variables was included for highest age being 16,10 type of denominational school and whether the school was a grammar, specialist or secondary modern school (the reference school being a non-denominational comprehensive school with a sixth form). A dummy term for wholly selective LEAs was included as a test of whether being in a selective LEA had an additional effect on GCSE attainment in addition to that of the type of selective school attended. The regressions were fitted in MLwiN allowing for variance at three levels—LEA, school and pupil. The effect of attending a secondary modern or a grammar school is tested by whether, on average, the GCSE results of pupils in these two types of selective school were less or greater than those predicted for pupils with the same KS2 test results who attended comprehensive schools. For the purposes of this test we are only concerned with how well KS2 tests provide a measure of pupils’ cognitive attainment which is consistent between pupils and which is a good predictor of their later GCSE performance. The much more controversial issue of whether or not rising key stage test scores over time mean increasing educational standards (Tymms, 2004; Statistics Commission, 2005) is not relevant for a study of a single cohort of pupils. In fact, Tymms (2004) reports that the KS2 tests for mathematics and English correlate well with the PIPS attainment tests for Year 6 pupils.11 A further indication of the validity of the KS2 tests is that in our data set they correlate well with teacher assessment in the respective core subjects (for English the correlation is 0.82, mathematics 0.83 and science slightly lower at 0.76). Our main concern is with the ability of the KS2 total score in English, mathematics and science
Are secondary modern pupils disadvantaged? 163 to predict GCSE results and so provide a valid measure for comparing pupil progress in comprehensive, grammar and secondary modern schools. Fifty-two per cent of the variation in pupils’ GCSE scores is explained by their KS2 total score, which is highly significant. From fitting a multilevel model including gender, age and the school context variables listed above, we found that the average estimated effect of attending a grammar school on pupils of all abilities (as measured by their KS2 scores) was 5.5 additional GCSE/GNVQ points (i.e. grades) compared to being at a comprehensive school. The effect of attending a secondary modern was on average 1 grade less at GCSE compared to a comprehensive school. These results are given in the first column of Table 4, which also contains estimated coefficients on a selection of other independent variables in the regression equation.12 (The full regression results are reported in Table A2 in the Appendix.) The grammar–secondary–modern differential at 5 grades is considerably smaller than the 16 points reported by Shuttleworth and Daly (2000) for Northern Ireland. We tested whether the effect on attainment of the type of school attended varies according to ability by including interaction terms between the pupil’s KS2 score and attending a grammar school, and between the KS2 score and secondary modern school. The estimated coefficient on the grammar school–KS2 score interaction term was negative, indicating that the more able the pupil, the smaller the advantage of attending a grammar school compared to being at a comprehensive. A pupil with a mean KS2 score was estimated to gain 9.4 GCSE points by attending a grammar school compared to being at a comprehensive, whereas a pupil with 1 standard deviation above the KS2 mean only gained 4 additional GCSE points by being at a grammar school.13 The results indicate that the more able the pupil (at KS2), the greater the loss in GCSE points from attending a secondary modern school. Whereas an average ability student obtained one less GCSE grade at a secondary Table 4. Estimates of effect of school type on GCSE total scores Independent variable
KS2 total score Girl Percentage pupils eligible for free school meals Grammar school Secondary modern school Wholly selective LEA KS2 score interacted with grammar school KS2 score interacted with secondary modern school
Size of effect on GCSE total score of a 1 unit change in the independent variable Without interaction terms
With interactions
0.323 3.375 20.146 5.50 21.04 0.27 (not sig) omitted omitted
0.328 3.375 20.146 9.37 21.23 0.02 (not sig) 24.94 21.04
Note: all coefficients are statistically significant at .05 unless otherwise stated.
164 R. Levacˇic´ and A. J. Marsh modern, a pupil with 1 standard deviation KS2 score above the mean would achieve 2 grades less. The second measure of GCSE attainment is the probability of a pupil obtaining 5 or more grades at A*–C. This requires fitting a logistic regression equation (briefly explained by Schagen and Schagen, 2003). The results are summarised in Table 5.14 A girl or boy of average ability had a 30% higher probability of getting 5 or more grade A*–C GCSEs attending a grammar school than if attending a comprehensive and only between 3 and 4% less chance if at a secondary modern. For pupils of high ability (1 standard deviation above average) boys were 15% more likely to achieve 5 good GCSEs in a grammar school and girls 10% more likely, and both were 2–3% less likely to achieve this in a secondary modern. Our results for both measures of GCSE attainment—that middle ability children gain most by grammar school attendance—are consistent with those of Schagen and Schagen (2003) and Benton et al. (2003). The extent to which the effect of grammar schools on attainment is due to the positive influence of a socially advantaged and well-motivated peer group and to what extent due to these schools’ focused academic mission is difficult to determine (Shuttleworth & Daly, 2000). Estimates of the grammar–secondary modern attainment differential which use a dummy variable to measure the effect on attainment of the two school types underestimate the attainment disadvantage of being a secondary modern pupil because the measurable effects of the peer group are controlled for already by including the percentage of pupils eligible for FSM or with SEN. To get a better idea of the attainment disadvantage of secondary moderns we need to include an estimate of the effect on pupils’ attainment of attending schools that have a higher concentration of socially disadvantaged students than they would in most cases if the local system were comprehensive. In the 10 wholly selective LEAs the average percentage of pupils eligible for FSM was 2.3% compared to 14.8% in secondary moderns. The average FSM percentage for the LEAs as a whole was 10.7. On average, therefore, secondary modern schools Table 5. Estimates of the effect of grammar and secondary modern schools on the probability of achieving 5 or more A*–C grades at GCSE compared to a comprehensive school Increased probability of a student gaining 5+A*–C grades at GCSE If pupil of average ability in terms of KS2 scores If pupil 1 standard deviation above average ability in terms of KS 2 scores If pupil 2 standard deviation2 above average ability in terms of KS 2 scores
Boy Girl Boy Girl Boy Girl
Grammar school %
Secondary modern school %
30.0 29.5 14.8 10.0 2.8 1.7
23.4 24.0 23.0 22.2 20.7 20.4
Sources: QCA matched pupil data set 1996–2001; Annual School Census 2001. Note: 3% of grammar school pupils were at or below average ability (in terms of KS2 results); 34% were between 0 and 1 standard deviation above average ability; 3.1% were 2 standard deviations or more above average ability.
Are secondary modern pupils disadvantaged? 165 had 4% more pupils eligible for FSM due to the selective system. Using the estimate of the effect of 1 percentage FSM on GCSE total points of 20.146, the effect of the 4% higher FSM percentage is to reduce attainment in secondary moderns on average by 0.6 GCSE grades. Grammar school pupils enjoy the advantage of 8.4% fewer of their peer group entitled to FSM, which has an estimated positive impact on GCSE scores of 1.2. The advantages of a better peer group for those selected for grammar school are therefore estimated to be 1.8 additional grades. As the PISA reports point out, if the negative impact of social stratification caused by selective school systems is not to be addressed by dismantling selective systems, then policies to raise the attainment of socially disadvantaged pupils need ‘to improve the resources, policies, processes and climate’ in the schools which these students attend (OECD, 2003 p. 267). However, when we examine the funding of secondary modern schools in England we find that they are in fact less generously funded per pupil than comprehensive and grammar schools, after taking account of factors which determine school costs such as size, age range, regional cost differentials, and the concentration of socially disadvantaged pupils and those with special needs. The funding of secondary modern schools Data on the public revenues received by English secondary schools were obtained from the Section 52 outturn statements on individual school budgets that LEAs submit annually to the DfES. The Section 52 statements (DfES, 2002b) provide two measures of revenue per pupil per annum:
N N
budget share per pupil (the revenues allocated by LEA funding formula plus those which come via the Learning and Skills Council for sixth forms); current revenue per pupil, which includes in addition contingency, devolved standards fund and other central grants.
We first present some descriptive data and then a regression analysis of the determinants of school funding. The data analysed include the school-level variables used in the analysis of attainment plus those added in from Section 52 statements. The schools’ budget shares per pupil15 and current revenue per pupil for the financial years 2000/01–2002/03 for the three school types are compared in Table 6. Only schools with sixth forms are compared as these receive more revenue per pupil and secondary moderns have disproportionately more schools without sixth forms. Though there are higher concentrations of socially disadvantaged and SEN pupils in secondary modern schools, there was little difference in budget share per pupil compared with grammar schools, though this moved slightly in favour of secondary moderns over the three years 2000/01–2002/3. The funding per pupil was more favourable to secondary moderns when the amounts received via central government are included. The pattern is also different by local authority. We show this just for 2000/01 in Table 7. In seven of these LEAs in 2000/01 the budget share per secondary modern pupil was less than the grammar school budget share per pupil. Slough was the most
166 R. Levacˇic´ and A. J. Marsh Table 6. Funding of secondary schools with sixth forms by school type Budget Share per pupil £
Secondary modern schools Grammar schools Comprehensive schools
Total public revenue per pupil £
2000/01
2001/02
2002/03
2000/01
2001/02
2002/03
2507 2561 2627
2703 2694 2767
2963 2936 3028
2675 2652 2804
2934 2844 3037
3258 3151 3342
Source: Section 52 Budget Statements, 2000/01 to 2002/03 (DfES, 2002b).
generous with £330 extra per secondary modern pupil, but it had only one secondary modern school with a sixth form. Sutton was the next most generous with £196 extra per secondary modern pupil, while Kent provided £47 extra. Two authorities, Buckinghamshire and Torbay, funded each grammar school pupil around £90 more. However, the comparisons in Tables 6 and 7 do not take into account differences in the factors that determine school costs per pupil—that is, the cost drivers. The main cost drivers are the size of school (as average costs fall with school size), the age range of pupils (since costs per pupil rise through the secondary key stages as the curriculum becomes more specialised), differences in regional costs (proxied by the Area Cost Adjustment Factor16 and additional learning needs, which are proxied by indicators such as the percentage of pupils eligible for FSM and with special needs. Table 7. Budget share per pupil in secondary modern and grammar schools in selective LEAs 2000/01 Schools with sixth forms Secondary moderns with sixth forms Bexley Buckinghamshire Kent Lincolnshire Medway Slough Southend Sutton Torbay Trafford
£2500 £2253 £2529 £2448 £2507 £2978 £2663 £2610 £2438 £2421
Grammar Difference between Schools secondary modern and grammar funding £2530 £2341 £2482 £2498 £2538 £2648 £2713 £2414 £2532 £2464
2£30 2£88 £47 2£50 2£31 £330 2£50 £196 2£94 2£43
All schools Secondary Difference between secondary modern moderns and grammar funding 2524 2253 2539 2594 2496 2864 2759 2594 2433 2671
2£6 2£88 £57 £96 2£42 £216 £46 £180 2£99 £207
Source: Section 52 Budget Statements 2000/01 (DfES, 2002b). Note: in Medway one secondary modern with only 128 students is omitted since it had extremely high costs due to its size.
Are secondary modern pupils disadvantaged? 167 In addition to the LEA formula-determined budget, schools received differential amounts of Standards Fund via the DfES, some of which was due to participating in particular programmes such as Excellence in Cities, and beacon, fresh start and specialist schools. We need to allow for these differences in cost drivers and in eligibility for additional Standards Fund before we can judge whether secondary modern and grammar schools are funded more or less than comparable comprehensive schools. This is done by regressing current revenue per secondary pupil17 for the years 2000/01–2002/03 on the cost drivers listed above and on a set of dummy variables for grammar and secondary modern school and for the different categories of school receiving standards funding.18 The results are given in Table 8. Of particular interest is the size and statistical significance of the coefficients on the dummies for grammar school and secondary modern school. These show the estimate of how much extra or less a school got per pupil if it was a grammar or a secondary modern compared to a comprehensive school after controlling for other factors that determine school revenues. These coefficients are consistently positive for grammar schools and statistically significant in all three years. Grammar schools were funded more per pupil compared to comprehensive schools (£131 in 2000/01, £115 in 2001/02 and £204 in 2002/3) after allowing for differences in school-level cost drivers, Standards Fund eligibility, differences in LEAs’ estimated need to spend on secondary education (measured by Standard Spending Assessment or SSA) and for the fact that wholly selective LEAs fund less per pupil. On average wholly selective LEAs spent less on secondary schools per pupil, varying from an estimated £56 in 2000/01 to £83 in 2002/03. Even allowing for this, secondary modern schools were funded less per pupil compared to comprehensive schools. The estimated coefficients of 2£85 and 2£73 are statistically significant for 2000/01 and 2001/02 but the amount for 2002/03 though still negative is not statistically significant. Grammar schools, however, are estimated to have obtained £200 more per pupil in that year than a comprehensive school with equivalent cost drivers. Examining three years of school revenue data therefore shows that secondary modern schools have been at a financial disadvantage compared to comprehensive schools in two of the years while grammar schools have received relatively more generous funding all three years.19 Budget deficits and surpluses A further indication of the relative underfunding of secondary modern schools is that they are more prone to run budget deficits than comprehensive schools, while grammar schools are in much less likely to be in the red. Section 52 statements include the balance carried forward to the next financial year. From this the schools with negative carry forwards (i.e. deficits) in each year from 2000/01 to 2002/03 were identified and the number of years from zero to three each school was in deficit was calculated and cross-tabulated by school type. The distribution of schools by type according to number of years that deficits were run is shown in Table 9. Fiftyfour per cent of secondary moderns were in surplus for all three years compared to
168 R. Levacˇic´ and A. J. Marsh Table 8. Variables explaining patterns in revenue per pupil for secondary schools 2000/01 to 2002/03 Independent variable Constant No. of pupils enrolled Inverse of no. of pupils enrolled Percentage of pupils eligible for free school meals Percentage of pupils with statements of SEN Percentage of pupils with SEN but without statements School has a sixth form Grammar school Secondary modern school School in wholly selective LEA Specialist school School in special measures School in education action zone Beacon school Excellence in Cities school Fresh start school Training school Leading edge school Leadership incentive grant school No of schools Adjusted R squared
2000/01 1838** 0.080** 306483** 12.05**
2001/02 1895** 0.107* 383038** 14.16**
2002/03 1987** 0.145** 422703** 17.21**
23.37**
28.65**
38.69**
1.09
1.55**
2.87**
133.23** 115.19** 273.00** 267.02** 43.38** 70.33* 259.81** 17.95 39.81 393.74** 47.34** 67.14** 89.48** 2532 0.72
154.47** 204.39** 243.46 283.46** 69.19** 144.0** 238.52 34.06** 94.34** 407.54** 67.12** 48.99** 62.43** 2750 0.74
128.73** 131** 285.23** 255.50* 50.3** 40.57 234.72 5.2 50.05* 531.22** 27.27 31.36* 53.96** 2452 0.69
Note 1: ** indicates statistical significance at .05 and * at 0.1. Note 2: the effect of outliers and clustering of schools in LEAs on standard errors corrected for by using robust and cluster commands in Stata.
64% of comprehensives. Fifteen per cent were in deficit for all three years compared to 9.4% of comprehensives. In contrast, 88% of grammar schools were in surplus in all three years. The chi square statistic for differences between the distribution of the three school types by number of years in deficit was significant at 1%.20 The deficit position of secondary modern schools was particularly severe in Buckinghamshire where 9 out of 20 schools had run deficits for three years and only four had been in surplus in every year. Only 1 out of the 13 grammar schools had run a deficit for three years while 11 had been in surplus throughout the period. Budget deficits have a number of deleterious implications for schools and their students. A school with a budget deficit cannot apply for specialist school status, which government policy advocates as a means for improving the quality of secondary education and which provides a school with additional funding. A school with a deficit has no contingency fund to use in the event of unexpected expenditures. A deficit must be eventually corrected and, unless additional funding is provided, this can only be achieved by spending less than current revenue, thus
Are secondary modern pupils disadvantaged? 169 Table 9. Percentage of secondary schools by type with budget deficits 2000/01–20002/03 Number of years in budget deficit in period 2000/01–2002/03 Zero years
1 year
2 years
3 years
Secondary modern Number of schools %
129 54.4
41 17.3
32 13.5
35 14.8
Grammar Number of schools %
131 87.9
11 7.4
3 2.0
4 2.7
Comprehensive Number of schools %
1586 64.3
362 14.7
285 11.6
232 9.4
England total/average Number of schools %
1846 64.7
414 14.5
320 11.2
271 9.5
Source: Section 52 Statement 2000/01, 2001/02, 2002/03 (DfES, 2001b, 2002b, 2003b). Note: no financial data for Trafford for 2001/02 and 2002/03.
impoverishing the education of the current cohort of students in order to pay for what a previous cohort received. The foregoing evidence on funding and the financial health of schools points quite clearly to secondary modern pupils being on average disadvantaged by lower funding per pupil than pupils in comprehensive schools, and particularly relative to grammar schools. The additional costs of secondary modern schools Budget deficit problems spurred the formation of the Buckinghamshire Upper Schools Forum to campaign for what they regard as fairer funding of upper (secondary modern) schools in the county. In order to estimate the additional costs that secondary modern schools incur compared to grammar schools because of the greater needs of their student communities, the Forum commissioned a questionnaire to be administered to its member schools.21 From initial discussion with head teachers additional cost factors were identified and questions asked about these in relation to staff time involved and other expenses. Thirteen of the 21 upper schools completed the questionnaire on the costs of different aspects of provision, and the estimated additional costs are an average of these. Given the purpose of the questionnaire as part of a campaign for additional resources, one would expect head teachers to have an incentive to bias upwards the costs reported. We attempted to counteract this by making conservative assumptions when deriving the cost estimates and by not including expenditures that were funded from earmarked revenues. Nevertheless, we cannot be certain that an upward bias in the estimated costs is not present, though the argument that secondary schools have
170 R. Levacˇic´ and A. J. Marsh higher costs than grammar schools due to the greater learning needs of their students is highly plausible. The specific need-related activities that were included as additional costs are explained briefly below and the estimated additional cost per student are given in Table 10. Vocational courses These are offered to a higher proportion of students in years 10–12 in upper schools compared to grammar schools and are usually more costly to run. Special educational needs The percentage of SEN pupils with and without a statement in secondary modern schools far outweighs that in grammar schools. Therefore SEN expenditure in grammar schools can be regarded as very low, so that expenditure in secondary modern schools on SEN is treated as additional costs. The number of teachers and support staff reported to be deployed in supporting students with SEN was costed for Buckinghamshire upper schools and the revenue from statements of SEN subtracted to arrive at the additional cost. Pastoral care Secondary modern schools, because of higher levels of social deprivation, have a greater need to provide pastoral care services for their students. Eleven per cent of upper school pupils are eligible for FSM compared to 1.2% in grammar schools. Data from the special needs service in Buckinghamshire showed that upper schools, which enrol 56% of secondary students, had twice as many permanent exclusions as grammar schools and five times as many fixed term exclusions. Quantifying this additional expense was problematic given the absence of data on staff deployment in grammar schools. Therefore a conservative estimate was made that upper schools have to spend twice as much as grammar schools on pastoral care and so half the cost of the additional spine points for teachers undertaking specific pastoral care roles
Table 10. Estimates of additional costs of secondary modern school educational provision in Buckinghamshire Area of activity incurring additional cost Vocational courses Special needs Pastoral care Recruitment and retention KS3 Strategy: additional costs Additional costs Source: questionnaire to Buckinghamshire upper schools 2002.
Cost per pupil £4.15 £147.00 £5.92 £13.61 £3.84 £175.00
Are secondary modern pupils disadvantaged? 171 was included. The revenue from external funding of pastoral care was subtracted from total costs to arrive at an estimate of upper schools’ additional costs. Recruitment and retention Schools reported on how much they spent per year on recruitment and retention points for teachers and on advertising costs. This amounted to £340,000. From other responses it was calculated that vacancies were approximately 6% of the teaching establishment. Almost 12% of teachers were on temporary contracts and 11% were working in positions for which their qualifications or experience were inappropriate. It is reasonable to assume these figures would be considerably lower in grammar schools and that a conservative estimate of the additional costs of recruiting and retaining teachers is that it is twice as high in upper schools. This gave a total additional cost of £170,000 for recruitment and retention in the 13 upper schools. Additional contributions from the school budget for implementing the Key Stage 3 strategy22 Schools reported the costs of implementing the Key Stage 3 Strategy from which the amount financed from Standards Fund was subtracted. It was argued by the head teachers that as grammar schools recruited the ablest students they did not incur additional costs for the Key Stage 3 Strategy. The additional cost of all the activities listed above was estimated to be £175 per pupil. This compares to £79 per upper school pupil received from Buckinghamshire’s formula for the funding of SEN without statements, so there was an estimated shortfall of almost £100 per pupil. The greater occurrence of deficit budgets among secondary modern schools and its particularly high incidence in Buckinghamshire lend support to the evidence from the questionnaire on costs that these schools do face additional costs for which they were not adequately funded. Transport cost incurred by selective systems Although selective LEAs tend to fund schools less generously than non-selective LEAs, they tend to spend more on home-to-school transport. A selective system costs more to operate than a non-selective system because students are less likely to attend the school closest to their home and therefore require public subsidies for transporting them to school. This is compounded in rural areas by the greater distances involved. An additional cost and inequity is perpetrated when grammar school pupils are provided with free transport to their school if it is more than three miles from their home, regardless of whether or not it is the nearest grammar school, while secondary modern pupils only get subsidised transport if the nearest school is over three miles from their home. This is the policy with respect to school transport in Buckinghamshire and it perpetuates an academic pecking order in the county’s grammar schools. A regression of home-to-school transport expenditure per pupil at LEA level in 2000/01 shows that the proportion of pupils in the LEA attending grammar schools
172 R. Levacˇic´ and A. J. Marsh contributes to higher expenditure. Other factors are low population density (sparsity), the LEA’s additional educational needs indicator (reflecting the costs of transporting more pupils with SEN) and average school size (which is inversely related to transport costs per pupil). The association of these factors with home-toschool expenditure per pupil is shown in Table 11. Low population density has the largest effect followed by the proportion of pupils attending grammar schools.23 Every 10% of LEA pupils in grammar schools is associated with about £10 per annum extra per pupil in transport costs. Conclusions It is clear from the evidence presented that secondary modern schools are underfunded in relation to the needs of their pupils and that pupils are in general disadvantaged academically by attending them. There is a greater intensity of need in secondary modern schools on a number of measures. The costing exercise provided some indicative evidence from one of the selective LEAs, Buckinghamshire, that the additional costs of educational provision in secondary modern schools were in the region of £175 per pupil in 2000/01. Notwithstanding the greater needs of pupils in secondary modern schools, the budget share per pupil of secondary modern and grammar schools was virtually the same in the years 2000/ 01 to 2002/03. However, when central government direct funding is included, the difference in total public revenue had moved in favour of secondary modern schools to the tune of £100 per pupil by 2002/03. Additional analysis of the 10 selective LEAs shows variability. In Slough and Sutton secondary modern schools were more generously resourced than grammar schools. The reverse was the case in Buckinghamshire, Medway, and Southend, where secondary modern schools, regardless of whether they have sixth forms, received less budget share per pupil than grammar schools. A more accurate assessment of whether secondary modern schools are underfunded is obtained from regression analysis as this enables us to control for Table 11. Factors associated with home-to-school transport costs: dependent variable: cost per pupil per annum 2000/01 Variable Constant Low population density (measured in terms of the sparsity index for the LEA in the SSA formula) Proportion of pupils in grammar schools Average size of secondary school in LEA Additional educational needs (LEA indicator used in SSA formula)
Estimated coefficient (£) 99.65 248.95 103.14 20.146 16.55
Note: all estimated coefficients are significant at .05. Sources of data: Annual Schools Census 2001 and Section 52 Budget Outturn statements 2000/01 (Df ES, 2002b).
Are secondary modern pupils disadvantaged? 173 differences in cost drivers. After accounting for size, sixth form, region cost differences, percentage of pupils with additional learning needs at school and LEA level, wholly selective LEAs’ tendency to spend less per student, and Standards Fund grants for different school categories, secondary modern schools received around £80 less per pupil than comprehensive schools in two out of three of the years examined while grammar schools attracted over £100 more per pupil in each of the three years. Selective systems cost more in terms of home-to-school transport. Resources that could have been allocated to teaching and learning are spent on transporting students to more distant schools. A rough measure is that every additional 10% of pupils attending grammar schools in an LEA is associated with an increase in transport expenditure of £10 per student in the authority. One method for reducing or at least stabilising the home-to-school transport budget would be to require parents of grammar school pupils to pay for transport to schools not in the reserved area, as is the case with parents of secondary modern pupils. Secondary modern schools’ comparative underfunding was also reflected in the difficulties they have experienced in keeping their budgets balanced. A higher proportion of secondary modern compared to comprehensive schools operated with deficit budgets in the three years 2000/01–2002/03. In contrast, a much lower proportion of grammar schools was in financial difficulties. The value-added analysis shows that students in secondary modern schools are disadvantaged in terms of their GCSE/GNVQ attainment. The penalty of attending a secondary modern school is on average 1 grade less than the student would have obtained at a comprehensive school and about a 3% reduced probability of getting 5 or more A*–C GCSE/GNVQ grades. When we take account of the additional social segregation created by selective systems, students assigned to secondary modern schools are further disadvantaged relative to those advantaged by grammar school selection by about 1.8 grades due to peer group effects. The most able pupils are less advantaged in terms of additional GCSE attainment by grammar school attendance than those of average ability—finding consistent with studies by Schagen and Schagen (2003) and Benton et al. (2003). The evidence presented in this article shows clearly that the selective system in 10– 20 local authorities works to disadvantage pupils who end up in secondary modern schools. In a few authorities the resourcing policies actually discriminate against secondary modern pupils compared to grammar school pupils. This is the downside of local political discretion to choose selective education. From an equity point of view the case against a selective system is strong but it appears understandably popular with parents whose children gain privileged access to grammar schools. If the selective system is to remain because of local political support, then there is a clear case for ensuring that secondary modern schools are better resourced than grammar schools so as to offset the compounding effects of concentrating lower income and lower ability students in schools reserved for them. The introduction by the DfES of dedicated schools budgets in 2006/724 makes such a redistribution even less likely to occur through LEA action than before, since LEAs
174 R. Levacˇic´ and A. J. Marsh are required to pass on DfES determined real per pupil increases in funding to all schools. They therefore cannot reallocate funding from grammar schools to secondary modern schools without raising their own spending on education, which they are unlikely to do once education funding is no longer their responsibility. In the longer term it is probably less expensive and more effective to raise the achievement of pupils who would be consigned to secondary modern schools by dismantling selection, though given current performance this would incur costs in terms of lower attainment for those no longer privileged by grammar school provision.
Acknowledgments We wish to acknowledge the work of David Newson and Mike Carslaw who were involved in the Buckinghamshire research on school costs, and Andrew Jenkins for assisting with the statistical analysis.
Notes 1. 2.
3.
4.
5. 6.
7.
8.
Middle deemed secondary schools (i.e. those with the highest age of pupils 14) are excluded. These are schools that classify themselves as secondary modern plus any nominally comprehensive schools in wholly selective LEAs. In the past some schools reporting themselves as ‘secondary modern’ in the performance tables may have quite legitimately been recorded as ‘comprehensive’ or otherwise within the Annual Schools Census. In order to ensure valid benchmarking comparisons, the DfES has now included ‘secondary modern’ as one of six options in the new field ‘intake type’ to be included in the Pupil Level Annual Schools Census (PLASC) (DfES, 2003a). The intake types are Comp5Comprehensive, SEL15Selective (Grammar), SEL25Selective (Secondary Modern), SEL35Selective (Technical), SEL45Selective (Religion), SPEC5Special. Due to lack of alternative data, Schagen and Schagen (2003) used Key Stage 2 in 1997 as prior attainment for Key Stage 3 test results in 2000, and Key Stage 3 in 1998 as prior attainment for GCSE results in 2000. The problem with these measures of value added is that higher value added at Key Stage 4 may be due to catching up on lower value added at Key Stage 3. Hence a better measure of secondary school performance is value added from KS2 to KS4. We use this measure in our analysis. Benton et al. (2003) used the preferable measure of prior attainment at Key Stage 2 for assessing value added at Key Stage 3 and at GCSE in 2002, as well as using prior attainment at Key Stage 3 2000 linked to GCSE in 2002. The results were weaker when an instrumental variable model was fitted in order to correct for bias due to living in a selective area being related to family background. The standard error is biased downwards in an ordinary least squares regression using nested data (i.e. hierarchically ordered data such as pupils in schools). Hence estimates may be reported as statistically significant when they are not. The data set used is taken from 398,532 matched records from a representative sample of pupils at Key Stage 2 (1996) to GCSE/GNVQ 2001 supplied by the QCA. Independent school pupils, pupils in special schools and pupils with missing data or whose schools had missing data were dropped, leaving 334,574 cases. All these are reputable national data sets widely used in UK educational research.
Are secondary modern pupils disadvantaged? 175 9. 10. 11.
12.
13.
14.
15.
16.
17.
18. 19.
Key Stage 2 attainment was entered with linear, squared and cubed terms as all three are statistically significant. Dummy variables for later intake years than Year 7 were dropped as they were not statistically significant. The correlation between KS mathematics and the Durham PIPS maths test in any one year is 0.85 and for English 0.83 (Tymms, 2004, p. 480). Stobart (2001) notes that the ‘test development process is exemplary’ but that the validity of the national tests as measures of year to year changes in standards is questionable. However, this issue is not relevant for our analysis which uses data for a single pupil cohort. In the regressions the continuous variables were standardised but in the tables the coefficients have been converted to their natural units (e.g. GCSE total points) to make their interpretation easier to understand. Hence a coefficient of 0.3 on the Key Stage 2 average score means that one extra mark at Key Stage 2 predicts 0.3 extra points at GCSE. As the regression equation was estimated in standardised units the coefficient on the Key Stage 2–grammar school interaction term is 24.94 for pupils whose Key Stage 2 scores are 1 standard deviation above the mean. These pupils would also lose a further estimated 1.04 GCSE points by attending a secondary modern school, making their total GCSE points 2.2 less than if they had attended a comprehensive school. Because these estimates are derived from a logistic regression they are non-linear with respect to the explanatory variables. This means that the probability of a pupil with a given Key Stage 2 score getting 5 or more A*–C grade GCSEs varies also with the values of all the other independent variables, so we need to assume values for these when presenting illustrative results. In the examples given in Table 5 we compare pupils who attend non-denominational, non-specialist schools, with sixth forms and which have average values for all the continuous independent variables. We then take a pupil of given ability and compare the probability that he/she obtains 5 or more A*–C GCSEs in grammar and secondary modern schools compared to comprehensive schools. The school budget share consisted of revenues delegated by the LEA plus standards grant from the Treasury and standards funds which come via the DfES. Data on pupils were taken from the Annual Schools Census. The pupil roll used to calculate the budget share per pupil for 2000/2001 was five-twelfths of the January 2000 roll plus seven-twelfths of the January 2001 roll. The Area Cost Adjustment Factor is used by the Government in calculating LEAs’ Education Standard Spending Assessment (Education Formula Share since 2003/ 04). It is an index based on differences in regional gross average earnings and varies from 1 to 1.25. To take account of differences in regional costs current revenue per pupil was divided by the Area Cost Adjustment factor since there are too few values of ACA to use it as an independent variable. The different categories are given in the Register of Educational Establishments. These are public revenues. Grammar schools with higher income pupils are likely to generate more own revenue per pupil than secondary modern schools as in Buckinghamshire (see Levacˇic´ et al., 2002).
20. Chi-square tests
Pearson chi-square n of valid cases
Value
df
Asymptotic. sig. (2-sided)
49.68 2851
6
.000
176 R. Levacˇic´ and A. J. Marsh 21. The questionnaire was not administered to grammar schools as it was anticipated they might be reluctant to respond. The evidence would have been stronger had a direct comparison with grammar schools been possible. 22. The Key Stage 3 Strategy was a DfES initiative, backed by funding, to improve teaching and learning and therefore student attainment from entry to secondary school to the KS3 tests three years later. 23. This can be seen from the sizes of the standardised regression coefficients. 24. In December 2004 the DfES announced that LEAs must pass on a mandated percentage increase per pupil to all schools and that from 2006/07 LEAs will be given a dedicated schools grant to pass onto schools.
References Benton, T., Hutchinson, D., Schagen, I. & Scott, E. (2003) Study of the performance of maintained secondary schools in England. Report for National Audit Office (Slough, National Foundation for Educational Research). Crook, D., Power, S. & Whitty, G. (1999) The grammar school question: a review of research on comprehensive and selective education (London, Institute of Education, University of London). Department for Education and Skills (Df ES) (2001a) Schools: achieving success (London, Df ES). Department for Education and Skills (Df ES) (2001b) The education (outturn statements, England). Regulations 2001 and accompanying guidance (relating to the 2000-1 financial year) (London, Df ES). Department for Education and Skills (Df ES) (2002a) Statistics of education: pupil progress in secondary schools by school type in England: 2001. National Statistics Bulletin, Issue 05/02. Department for Education and Skills (Df ES) (2002b) The education (outturn statements) (England). Regulations 2002 and accompanying guidance (relating to the 2001–2 financial year) (London, Df ES). Department for Education and Skills (Df ES) (2003a) Pupil level annual schools census (PL ASC) January 2003. Completion notes for secondary schools. Available online at: www.dfes.gove.uk/ datacollection/asc/ASC2003Home.asp. Department for Education and Skills (Df ES) (2003b) The education (outturn statements, England). Regulations 2002 and accompanying guidance (relating to the 2002–3 financial year) (London, Df ES). Department of Education and Science (1965) The organisation of secondary education. Circular/65 (London, HMSO). Dustmann, C., Rajah, N. & Van Soest, A. (2003) Class size, education and wages, The Economic Journal, 113(485), 99–120. Edwards, T. & Tomlinson, S. (Eds) (2002) Selection isn’t working. Diversity, standards and inequality in secondary education (London, The Catalyst Forum). Available online at: www.catalystformum.org.uk. Feinstein, L. & Symons, J. (1999) Attainment in secondary school, Oxford Economic Papers, 51, 300–321. Galindo-Rueda, F. & Vignoles, A. (2004) The heterogeneous effect of selection in secondary schools: understanding the changing role of ability, Institute for the Study of Labor (IZA), Discussion paper 1245, Bonn, Germany. Gallagher, T. & Smith, A. (2000) The effects of the selective system of secondary education in Northern Ireland. Main Report (Bangor, Education Department Northern Ireland). Goldstein, H. (1995) Multilevel statistical models (London, Arnold). House of Commons Education and Skills Committee (2003) Secondary education: diversity of provision. HC 94 (London, HMSO). Jesson, D. (1999) Evaluating performance of LEAs and schools of differing types (York, Centre for Performance Evaluation and Resource Management, Department of Economics, University of York).
Are secondary modern pupils disadvantaged? 177 Jesson, D. (2000) Further evidence on comparative GCSE performance between selective and non-selective schools and LEAs, paper presented at NUT Secondary Education Conference, Harrogate, 1 March. Jesson, D. (2001) Selective systems of education—blueprint for lower standards? Discussion paper. Available online at: http://www.york.ac.uk/depts/econ/rc/cperm.htm (accessed 14 September 2001). Kreft, I. & De Leeuw, J. (1998) Introducing multilevel modelling (London, Sage). Levacˇic´, R., Marsh, A. & Newson, D. (2002) The penalty costs of upper school funding: towards greater fairness in the secondary sector. Available online at: htpp://www.missingbucks.org. Office for Standards in Education (Ofsted) (2002) National summary data report for secondary schools 2001. Data version 1.2 (London, Ofsted). Available online at: www.ofsted.gov.uk/ public/docs01. Organisation for Economic Cooperation and Development (OECD) (2003) Literacy skills for the world of tomorrow—further results from PISA 2000 (Paris, OECD/UNESCO). Available online at: www.pisa.oecd.org. Organisation for Economic Cooperation and Development (OECD) (2004) Learning for tomorrow’s world: first results from PISA 2003 (Paris, OECD). Available online at: www.pisa.oecd.org. Schagen, I. & Schagen, S. (2003) Analysis of national value added data sets to assess the impact of selection on pupil performance, British Educational Research Journal, 29(4), 561–582. Shuttleworth, I. & Daly, P. (2000) The pattern of performance at GCSE. Research Paper SEL3.1 (Bangor, Department of Education, Northern Ireland). Simon, B. (1991) Education and the social order 1940–1990 (London, Lawrence & Wishart). Statistics Commission (2005) Measuring standards in English primary schools. Statistics Commission Report No. 23, 10. Available online at: http://www.statscom.org.uk/C_402.aspx/. Steedman, H. (1980) Progress in secondary schools (London, National Children’s Bureau). Steedman, H. (1983) Examination results in selective and non-selective schools: findings for the National Child Development Study (London, National Children’s Bureau). Stobart, G. (2001) The validity of national curriculum assessment, British Journal of Educational Studies, 49(1), 26–39. Tymms, P. (2004) Are standards rising in English primary schools? British Educational Research Journal, 30(4), 477–494.
Appendix
Table A1. Descriptive statistics of the main continuous variables Mean K2 total marks for English, maths and science GCSE total points score 2001 FTE pupils: average 1997–2001 FSM: average percentage 1997–2001 SEN with statements: average percentage 1997–2001 White: average percentage 1997–2001 PTR average 1997–2001 Number of pupils in school in QCA matched data set 1996–2001 (excludes independent, middle and special schools)
154.5 40.4 1056 16.4 2.5 88.4 16.9 340124
178 R. Levacˇic´ and A. J. Marsh
Table A2. Estimated regression equations for GCSE/GNVQ total score and probability of 5 or more A*–C GCSE grades Independent variables
GCSE/GNVQ total score: (no interactions) natural units
GCSE/GNVQ total score: (with interactions) natural units
Probability of 5+A*– C (logit regression) normalised units
Estimate t statistic Estimate t statistic Estimate t statistic Constant KS2 total for English, maths and science Age Girl Percentage of girl pupils: 2000 FTE pupils: average 1997–2001 FSM average % 1997–2001 FSM average (1997–2001) squared FSM average (1997–2001) cubed KS2 English maths and science squared SEN with statements: average % 1997–2001 White: average % 1997–2001 PTR average 1997–2001 Highest age 16 (51) Specialist school Church of England Roman Catholic Other religion Grammar Secondary modern Grammar*KS2 total Secondary modern*KS2 total Wholly selective LEA
23.34 0.323
219.22 323
23.32 0.328
219.11 328
20.799 2.066
228.32 158.92
0.009 52 3.375 87.5 0.014 4.01 .0002 0.8 20.146 229.2 — —
0.009 3.375 0.015 .000 20.146 —
51 87.5 4.33 0.8 229.2 —
0.135 0.519 0.046 0.001 20.589 0.174
27.00 47.18 4.18 0.07 228.05 9.16
— 39
20.024 20.024
24.80 23.00
— 0.018
— 36
— 0.019
20.20
24.25
20.017
24.25
20.038
22.71
20.08 20.27 0.89 1.47 0.89 0.96 2.08 5.50 21.04 — — —
216.2 23.75 4.6 7.6 2.56 4.17 1.96 10.6 22.16 — — —
20.08 20.27 0.89 1.48 0.89 0.94 2.51 9.37 21.23 24.94 21.04 0.02
216.2 23.75 4.6 7.70 2.56 4.08 2.36 16.76 22.56 218.29 26.75 0.03
20.204 20.021 0.155 0.131 0.039 0.088 0.568 1.252 20.164 — — —
214.57 21.62 5.17 4.23 0.74 2.51 2.87 14.73 22.78 — — —
Note. The coefficient on KS2 attainment is random. In Table A2 we present the estimated coefficients from the logit equation in standardised units, whereas in Table 5 we give the probability of the dependent variable (5 or more A*–C GCSE/ GNVQ) for given values of the explanatory variables. The probability of a pupil getting 5 or more A*–C GCSE/GNVQ varies with the values of the explanatory variables and is reported for given assumptions about their values. Given that p stands for the probability that a pupil obtains 5 or more A*–C GCSE/GNVQ, the dependent variable in the logit regression equation is the natural logarithm of p/(1–p), i.e. ln[p/(1–p)]5ebx where e is exponential, x is the explanatory variable and b the coefficient on x. The probability, p, of 5 or more A*–C GCSE/GNVQ is then calculated by working out 1/[1+e2bx].