Research Report No 3
School of Geosciences
In March 1994 the Central Community Relations Unit commissioned Dr Ian Shuttleworth of Queen's University, Belfast to conduct research on community differences in the qualifications and immediate post-school histories of school leavers in Northern Ireland as assessed by information from the Secondary Education Leavers' Survey. The following is a report of his findings.
This work is one of a series of research projects commissioned by the Central Community Relations Unit in the context of its preparations for the Review of Employment Equality in Northern Ireland. In commissioning this series, the policy of the Central Community Relations Unit has been to publish the research findings with a view to informing wider discussion of issues relating to employment equality. When the Standing Advisory Commission on Human Rights was tasked with the conduct of the Review in November 1994, it was agreed that CCRU should continue to publish this series.
As with all other work in the series, the views expressed in the
report are responsibility of the author and should not necessarily
be regarded as being endorsed by the Central Community Relations
Unit or by the Standing Advisory Commission on Human Rights.
Ireland Secondary Education Leaver's Survey
This report describes and analyses the educational and post-school
experiences of a sample of young people who left Northern Ireland
schools in the academic year 1990-91. Two main topics were investigated;
firstly the relationship between social background and educational
attainment and, secondly, the social and educational determinants
of post-school outcomes, it is well-known that certain social
characteristics, for example family size, influence educational
attainment and post-school outcomes. Therefore, the analytical
approach used in the report was to isolate how much of the difference
between Catholics and non-Catholics could be explained by the
differing social composition of the two communities and how much
could be attributed to the effects of religion,
Raw differences were identified between Catholics and non-Catholics in the subject mix of exam passes at GCSE and A Level. Differentials in educational attainment were also noted, In general, Catholics were less likely to have passed examinations in Science/Mathematics and, compared to non-Catholics, had relatively fewer examination passes.
The differences in subject mix can be explained by the proportion of Catholics who attended grammar school. Once the type of last school attended was controlled for,, few statistically significant differences remained at either GCSE or A Level.
Differential examination performance at both GCSE and A Level was explained by social background and the type of last school attended. This means that poorer Catholic examination performance can largely be explained by poorer relative socio-economic status (SES) and by the relatively lower proportion of Catholics who attend grammar school.
Three further topics for research were suggested, Firstly, the
implications of variations in examination-entry policy between
schools were considered to be worth exploring. Secondly, the possible
impacts of religious differentials in staying-on rates in post-compulsory
education on attainment at A Level and, finally, the consequences
of religious differences in transfer behaviour from primary school
for educational attainment in the state and Catholic school systems.
Six post-school states were defined; employment, training, unemployment, further education,, higher education,, and other inactivity. The influence of social/educational background on the likelihood of being in any one of these states was estimated on the one hand for fifth-form/lower-sixth form leavers and on the other for upper-sixth form leavers,
It was found that few unexplained religious differentials remained for upper-sixth form leavers once social/educational background had been taken into account, However, there were marked, and unexplained, religious differences observable for fifth-form/lower-sixth form school-leavers. In summary, Catholic males were much more likely to be on a training scheme than their other denomination counterparts, Catholic males also are, everything else being equal, more likely to continue to participate in post-compulsory education. Similar patterns were not observed for females.
Two other findings are worthy of further comment, Firstly, the
subject type of examination pass at both GCSE and A Level was
generally insignificant as a determinant of post-school destination.
In no case did it influence entry to employment. This suggests
that those arguments that suggest that a substantial amount of
the Catholic/non-Catholic differential in unemployment rates can
be explained by the differing 'human capital' endowment of the
two communities may be substantially misconceived. secondly, it
seemed that informal methods of job- search may be important for
young people in gaining employment. The use of 'family and friends'
to find work rather than more formal channels has obvious policy
Summary of findings
The author would like to acknowledge the Department of Education,
Northern Ireland (DENI) , the Training and Employment Agency (T&EA)
and the Department of Finance and Personnel (DFP) for support
in the collection of the data on which the report is based. The
author is also grateful to colleagues,, and most particularly
Anthony Murphy, who have commented on the analyses in Chapters
3 and 4,, and to the Northern Ireland Economic Research Centre
(NIERC) for its support. The author is also grateful for the advice
of PPRU. Responsibility for the content of the report, and for
any errors, rests, of course, with the author alone.
This report presents the results of an analysis of the Pilot Secondary Education Leavers' Survey (SELS) dataset. This dataset, collected between April and October 1992, contains information on the personal/family background, educational qualifications and immediate post-school destinations of 1,600 young people who left Northern Ireland schools in the academic year 1990-91. The central theme that runs through all the analyses that are later reported is the investigation of differences by religion between Catholics and members of other denominations in educational and immediate post-school experiences.
These questions are important on several counts, Firstly, research in both the Republic of Ireland and Great Britain has demonstrated the links between unemployment and disadvantage in early career stages and its likely continuation in later life (Breen 1991; Bates and Riseborough 1993), Though disadvantage in immediate post-school destination does not f ix inequality for the remainder of a person's working life, it might be reasonably hypothesised that initial inequalities between young people of different religious groups in Northern Ireland could have a substantial effect on later differentials in the labour market.
Secondly, an analysis of possible religious differentials in the contemporary experience of young people in Northern Ireland offers an extra dimension to the debate about inequalities by religion in Northern Ireland. Much of the debate has been conducted in terms of the adult male labour market (cf Compton 1991; Smith and Chambers 1991). This debate has been lively among both academics and policy makers. In this context it is possible to dismiss contemporary observed inequalities as being historical relicts that reflect features of the labour market conditions and educational system of the 1940s and 1950s. Taking up these points, an analysis of the situation of contemporary young people allows some consideration of the extent to which inequality is reproduced under present-day conditions in the Northern Ireland education and training systems, and labour market.
Thirdly, little is known about how changes in the structure of
early careers (such as the increased participation in education
and training for young people since 1980) have influenced the
balance of religious inequality in Northern Ireland. It is well-
known that there are inequalities between Catholics and Protestants
in educational attainment (Cormack and Osborne 1991; Miller et
al 1993) and similar inequalities in employment and unemployment
rates (Whyte 1990). However, the relationships between education,
employment and training, and the part they play in the reproduction
of religious inequality may also be usefully addressed by an analysis
of the conditions of young people.
The structure of the report, and the major areas of interest to be examined, are as follows. Chapter 2 sets the scene for the major analytical chapters of the report by outlining the structure of the SELS data and discussing general analytical principles that will guide the interpretation of the results that are presented in later chapters.
Chapter 3 begins the substantive section of the report with a discussion of religious differentials at GCSE and A Level. This Chapter begins by reviewing the literature on religious differentials in subject choice between Catholics and Protestants at GCSE and A Level and considers the extent to which these hypothesised differentials are replicated in the SELS data. It then proceeds to examine the nature of religious differentials in attainment in the raw SELS data at both GCSE and A Level before concluding with a multivariate analysis that attempts to investigate the degree to which these differentials can be explained once socioeconomic status (SES) and other factors have been taken into account .
Chapter 4 investigates the post-school destinations of the SELS respondents. The chapter is not narrowly directed at the transition from school to work; evidence from other areas suggests that the growth of unemployment, the expansion of post- compulsory education and the increase in training provision has made the concept of 'youth transition' more appropriate as a means of understanding the complicated (and lengthened) period of transit from school. In short, economic activity, inactivity and participation in education and training are all analysed to create a comprehensive picture of the behaviour of SELS respondents. For young people in Northern Ireland in the 1990s. The question is not only who gets work but also who participates in training and post-compulsory education, Data from the Northern Ireland Labour Force Survey (NI LFS) is used to outline the general background of the status of young people aged 16-24 resident in Northern Ireland as the local context in which the SELS-based analysis is undertaken. Following the strategy of Chapter 3, this review section is succeeded by a description of raw religious differentials in immediate post-school destination. The chapter then continues with a multivariate analysis that investigates how much these differentials can be explained when SES, the number of qualifications and personal variables such as gender and religious background are taken into account. A special feature of this analysis is that it considers the subject type of qualification at GCSE and A Level and its impact on post- school destination.
Chapter 5 draws together the main results of Chapters 3 and 4.
The central questions, for which answers are sought, concern the
extent to which religious differentials in educational attainment
and post-school destination exist and their possible implications.
Questions specifically relevant to fair employment policy are
raised with reference to the likely means by which young people
find work and the potential usefulness of measures to increase
the human capital endowment of young people in Northern Ireland.
The purpose of this chapter is to consider the nature of the SELS
database and to discuss broadly the types of analytical approaches
adopted. This is necessary as a preamble to the main part of the
report because it is useful to show the coverage of the SELS and
to examine the methodological basis for any inferences made from
the data. More detailed discussions of these issues will be introduced
where appropriate in Chapters 3 and 4.
The SELS database is the final product of a survey undertaken in mid 1992 using postal questionnaires and personal interviews. The survey was designed to contact a sample of about 10% of those young people who left Northern Ireland schools in the academic year 1990-91. Addresses for the survey were to be drawn from all schools. In practice, some schools declined to co-operate with the survey and some 2,000 names and addresses (approximately equivalent to a 10% sample) were selected from those schools (about 70% of the total) who took part in the survey.
To the first, postal, part of the survey a response rate of 63%
was achieved. Because of concern about possible non-response bias
(eg the less well-educated members of the sample being less likely
to respond to a postal enquiry) careful efforts were made to contact
poorly-responding sections of the sample by personal interviewers.
After validation the SELS data were found to be closely comparable
with information derived from other government data sources. It
was therefore concluded that the SELS data had captured most of
the features of the population of 1990-91 school-leavers and that
it was, as far as could be estimated, a representative sample.
The SELS questionnaire covered a wide range of questions .it would be tedious to list these in this report as the most important ones will be described in the analyses of educational attainment and post-school destination. The main comment to be made is that similar information was sought to that obtained in comparable surveys in the Republic of Ireland, Scotland and England. However, some remarks can be usefully addressed to the structure of the SELS database. The main point to make is that the SELS is not a year-cohort survey in that it does not look only at fifth formers or 16 year olds. Instead, it is what might be termed an event-cohort because it covers young people who left school in 1990-91, whether from fifth form, lower sixth form or from the upper sixth. Because of this it can only be used to answer certain types of question; it would, for example, be difficult to say much about fifth formers who stay-on at school (though, as will be shown in Chapter 4, some of their characteristics can be estimated). Table 2.1 gives some idea of the structure of the SELS sample by showing the numbers of respondents by sex and form.
Note: All tables, unless otherwise stated, are unweighted a exclude missing values. For example, in the above table, certain cases have been omitted because either form or sex were unknown.
Because of this data structure, care must be taken during analysis.
In particular, it would be inappropriate to consider the dataset
as a whole when looking at post-school destinations. Higher education,
for example, is largely a destination for upper-sixth leavers
whereas YTP is mainly a destination for those who left from fifth
The primary point to make is that all data from the SELS are to be treated as estimates of the characteristics of 1990-91 Northern Ireland school-leavers. As a sample, the SELS may not capture all the variability within the population from which it is drawn either because the sample is not large enough or because of other sources of bias. In this sense the sample is a model of reality.
When analysing the SELS data (in this case comparing Catholics and Protestants) the main task is to decide whether apparent differences between the two groups arise because of chance. This is usually done by means of a significance test. The usual significance level taken is 0.05; this means that there is a 1 in 20 chance of being mistaken when we say there is a statistically significant difference between Catholics and Protestants. In Chapters 3 and 4 simple two-way tables by religion are presented are shown to present raw differences in the SELS data and sometimes they are tested in this way.
Of course, these differences might not only be associated with religion. Other factors may operate in the background so that it seems like religion is linked with various types of inequality. To examine problems like these, multivariate techniques are used to control for the existence of these other variables. An example of this is educational attainment at GCSE. It might be that lower Catholic attainment is linked with factors such as parental background, family size and type of school last attended. A multivariate analysis would take account of the interplay of these factors. The use of multivariate techniques also permits a summary of complex relationships in the data.
Two notes of caution should be made about the use of multivariate
techniques in this type of analysis, Firstly, these techniques
make estimates of the relationships within the data (again with
a 1 in 20 chance of being mistaken). Secondly, they also only
control for the variables that are present in the SELS dataset
and that can be entered in the models.
The chapter has outlined the broad structure of the SELS data
and has suggested the limits within which the information can
be safely interpreted. This means that there are restrictions
on the conclusions that can be drawn. However, the SELS data are
reliable (as far as one can reasonably tell) and considerable
confidence can be placed in the database. In many cases the patterns
revealed by the data are very strong and point strongly to the
importance of certain factors.
This chapter considers the related issues of attainment at GCSE and A Level in Northern Ireland. It commences with a review of the main conclusions reached in recent years on inter-religious differentials in educational attainment. It then goes on to discuss differences in attainment between Catholics and Protestants by subject type before considering the extent these differentials are reduced once other factors, such as the type of school last attended, are controlled for in the analysis.
After this, the chapter deals with the issue of attainment at
GCSE and A Level in terms of the number of passes/points scored
in a more formal manner. The major questions concern the extent
to which raw differences in attainment between Catholics and Protestants
are explicable once SES and differential uptake of grammar school
places are taken into account.
Concern about inter-religious differences in educational standards has been a feature of much research on education in Northern Ireland since 1969. The central theme has been the extent to which school systems reinforce, with differences in curricula, ethos and attainment (Murray and Osborne 1983). divisions in the labour market and distinct communal identities.
Evidence from successive Censuses of Population (Eversley 1989; Cormack and Osborne 1991) and other government sources such as the Continuous Household Survey (CHS) (Smith and Chambers 1991) points to small but significant differences between Catholics and Protestants of all ages in educational attainment.
A similar picture is observed when looking at the output from the Northern Ireland school system. Broadly, Catholics are more likely to have lower qualifications, or to be unqualified, than Protestants. Moreover, these averages conceal a great degree of inequality in the educational attainment of young people from both communities. Though the mean level of attainment (in terms of numbers of GCSEs and A Levels) is higher in the province than the rest of the UK, this results from a polarisation of educational output (Regional Trends 1993). with a greater proportion of the highly-qualified in Northern Ireland plus a greater proportion of the unqualified or lowly qualified relative to Britain. In particular, there is evidence to suggest that the chances of leaving school with no qualifications is highest for those young people from working-class backgrounds who attend secondary schools (Wilson 1987).
Beside these quantitative differences in attainment, there is also strong and consistent evidence for qualitative differences in the educational experiences of Catholics and Protestants in the types of subject taken at school.
Murray and Osborne (1983), using evidence from the 1970s. show that pupils educated at Catholic schools were less likely than their state-educated counterparts to have achieved passes at GCE O and A Level in science, maths and craft subjects. More recent evidence cited by Murray (1993), points to the persistence of these patterns with regard to A Levels. It seems, for example, that the proportion of non-Catholics achieving A Level passes in 1989-90 is 9.4% higher than for Catholics. This greater Catholic propensity to take arts/humanities subjects is noted for both A Levels and degree subjects for recent Northern Ireland graduates (Miller et al 1993).
Catholics therefore appear to be distinguished from their Protestant counterparts in having fewer educational qualifications, and qualifications of a different type, Though the general features of these differences between the populations of Catholic and Protestant school-leavers appear to be clear, there is more uncertainty about their causes.
Two structural arguments have been advanced to explain why Catholic school-leavers, as a population, in terms of number of passes/points scored at GCSE and A Level have lower attainment than Protestant school-leavers. Firstly, differences in mean SES between the Catholic and Protestant communities have been identified as one factor that could underlie differential educational performance (Murray and Osborne 1983; Osborne et al 1991). It is well-known that Catholics are more likely to be unemployed than Protestants (Whyte 1990; Smith and Chambers 1991; Murphy and Armstrong 1993). Moreover, Catholic males are more likely to be in unskilled or semi-skilled jobs than Protestant males and the Catholic community tends, on average, to have larger families and to be more dependent on state benefits (Harbison 1989). Links between these factors and low educational attainment have been observed in British research (for example, Wiseman 1964; Paterson 1991a; Jesson et al 1992) and there is no reason to assume that similar relationships between SES and educational attainment do not exist in Northern Ireland.
Evidence about the relationship between SES and educational attainment in Northern Ireland is mixed. Boyle (1977) finds that no specific disadvantage is associated with religion but that Catholics are caught in a cycle of disadvantage in which low SES leads to low educational attainment which in turn leads to low SES. More recently, analysing data on examination performance from the 1980s. Daly (1991) also finds that no disadvantage is associated with religion once social class, gender and last school attended are included as controls in his model. On the other hand, Miller et al (1993) suggest that relatively lower Catholic attainment at A Level persists even after factors such as gender, social class and subject choice have been taken into account. As Osborne et al (1991) comment, these inter-communal differences in SES raise interesting questions that require further research.
The second structural argument advanced relates to differential uptake of grammar school places between Catholics and Protestants. The clearest exposition of this hypothesis is provided by Osborne et al (1991). Simply, it is argued that a higher proportion of Protestant children attend grammar school than Catholic children. This results in part from a lower proportion of available places in the Catholic school system and cases where Catholic children, despite being entitled to a free grammar school place, transfer to a secondary school. Since, in the selective Northern Ireland school system, grammar school pupils have on average higher attainment than children attending secondary school, it is reasonable to assume that the differences between the two communities in the proportion of children attending grammar school can go some way toward explaining lower Catholic educational attainment.
The hypotheses advanced to explain differential subject balance are not so clear-cut. Three main arguments have been suggested. Firstly, the cost of science teaching may inhibit the teaching of science in Catholic schools as they did not receive the same state support for capital funding as schools in the state sector (Osborne et al 1989). Secondly, it has been contended that the ethos of Catholic schools differs from that of state schools (Osborne et al 1991) and that this may lessen Catholic scientific orientation. Finally, difficulties in recruiting teachers for Catholic schools may reduce science teaching relative to the state sector.
Little work has been done to examine these f actors. it is probable,
however, that a simple analysis will be insufficient to explain
religious differentials in subject balance. All these potential
explanations may be important. Additionally i pressures from the
labour market, where Catholics have a lower relative concentration
in occupations requiring technical/scientific qualifications,
may also influence the choice of subject taken in Catholic schools
Charts 3.1 and 3.2 show the proportions of SELS respondents gaining examination passes by subject area and religion. Chart 3.1 shows GCSE passes at grade C or higher; Chart 3.2 at grade G or higher. These alternative definitions of a 'pass' were chosen because official data usually defines a GCSE pass as being at grade G or higher whereas there is a possibility that employers consider a GCSE pass to be at grade C or above (to match an old-style GCE O Level pass) . Chart 3.1 also gives data on A Level passes by subject group. These are defined as a grade E or higher. Since no alternative definitions are used, A Levels are omitted from Chart 3.2.
In Chart 3.1, there are significant differences between Catholics and Protestants in GCSE science and maths passes (at the 0.05 significance level). For A Level science and maths there are again highly significant differences between Catholics and Protestants (at the 0.01 significance level which means that there is a 1 in 100 chance of being wrong in the inference made). In Chart 3.2, which records GCSE passes at grade G or higher, the religious differentials are yet again highly significant (at the 0.01 level for science and maths).
These religious differentials are consistent with those reported elsewhere in the literature; the proportion of Catholics passing exams in maths and science is significantly less than that of Protestants. Moreover, when looking at GCSE passes at grades A- G, the proportion passing GCSE English is also significantly lower than for Protestants. Why might these patterns exist?
Part of the answer to this question (and to some questions about low attainment) might lie in the number of exams studied and sat on average by each of the religious groups. There are three ways in which a person might fail to gain a qualification. Firstly, they may not study any subjects for examination, secondly they may study a subject but not sit an exam in it. Thirdly, they may study for an exam, sit a paper, but then fail. In each of these three stages there are religious differences. About 6% of Protestants do not take any subjects for examination at GCSE compared with 8% of Catholics. Then about 10% of Catholics do not study for an exam, or do not sit an exam paper, as compared with about 8% of Protestants. Finally, about 2% of Catholics study and sit an exam at GCSE but then f ail to get a pass at grade G or higher compared with 3% of Protestants. it is therefore likely that part of the answer to the question of why Catholics pass fewer GCSEs and A Levels in maths and science (and why they have poorer exam results than Protestants) is that they enter and sit fewer exams rather than fail the exams they sit. These comments are not meant to be definitive; more research is needed to explore these issues and to throw light on how examination entry policy varies between state and Catholic schools.
Other parts of the answer are shown in Table 3.1, Table 3.2 and Table 3.3. These show religious differentials in passes by subject group controlling for type of school attended and sex. Two features are immediately noticeable from these tables. Firstly, the statistically significant differences between Catholics and Protestants observable in Charts 3.1 and 3.2 are diminished once the analysis controls for type of school. The second point is that controlling for gender, as well as school type f the differences between Catholics and Protestants generally remain statistically insignificant. A further interesting feature that can only be speculated about is the generally higher pass rates for Catholics at secondary schools when compared with Protestants. Without data on prior attainment, it is impossible to say why this might be so. Possible answers might range from more effective Catholic secondary schools to differences in the intakes to these schools in terms of ability. It is tempting to believe, however, that since some Catholics who are entitled to a grammar school place go instead to a secondary school (Livingstone 1987 cited by Osborne et al 1991), that this may be a result of Catholics with high transfer scores (relative to Protestants) entering secondary school.
Source: SELS Note: * indicates significance at the 0.05 significance
Source: SELS Note: * indicates significance at the 0.05 significance
Source: SELS Note * indicates significance at the 0.05 significance
These observations can be interpreted as meaning that the religious differentials shown in the charts are, at least in part the result of compositional effects, That is to say that they reflect curricula differences between secondary and grammar schools and differences in the ratio of Catholics and Protestants who attend grammar school.
Statistically significant differences between Catholics and Protestants
in subjects taken at GCSE and A Level were observed in Charts
3.1 and 3.2. Once school type and gender were controlled for in
the analysis, these differences diminished. In the SELS database
it therefore seems that religious imbalances in subjects taken
at school are largely a consequence of compositional effects;
that is to say they reflect curricula differences between grammar
and secondary schools and differing ratios of grammar school places
between the two communities rather than factors directly linked
to religion. Further research needed to confirm these findings
and to look in more detail at the hypotheses outlined in the literature
(Osborne et al 1991). One fruitful line of enquiry would be to
look in more detail at policy about examination entry and how
it varies between schools and between the state and Catholic school
Turning attention from subject imbalances to the number of passes/points scored at GCSE, this section of the chapter is used to explore the part played by personal/family background in explaining attainment at GCSE. Chart 3. 3 and Table 3 c 4 show religious differentials in attainment at GCSE using a variety of contrasting measures. No attempt is made to differentiate the cause of exam failure in terms of whether an exam was studied for or sat; the object of interest is simply attainment and the likelihood of gaining or not gaining a qualification, Table 3.4 shows the number of exam passes, measured at grade G or higher or grade C or higher by religion, whereas Chart 3.3 presents data on the distribution of points by religion scored by individuals in the SELS sample where A=7, B=6. . and G=1. Attainment at GCSE was measured at completion of schooling and so includes fifth- form and sixth-form school-leavers. Since the great majority of GCSEs are taken during fifth form, however, the analysis gives quite a firm idea of attainment at age 16.
The main feature to note, of course, is the generally lower attainment of Catholics relative to Protestants. In Table 3.4, for example, the difference between the two communities is statistically significant at the 0.05 significance level which means there is 95% probability that the differentials are not due to chance. In Chart 3.3, Catholics generally have lower GCSE scores than Protestants with a mean difference between the two groups of about two points (equivalent to a grade F). This is small but statistically significant.
Arguments about the causes of this differential have already been
rehearsed. They concentrated on the twin features of differential
access to grammar school by religion and differential SES by religion
both of which it was argued may contribute to the poorer performance
of Catholics, In the SELS data the first source of inequality
is addressed by looking at the proportion of each religious group
whose last school was a grammar school, As might have been expected
there is a marked imbalance that is statistically significant
at the 0.05 significance level; 40.3% of Protestants attended
grammar school as compared with 32.5% of Catholics.
Differences in mean attainment at GCSE in terms of personal/family background are shown in Table 3.5. The variables selected include gender, parental labour market status, type of school last attended and eligibility for free school meals. These variables were chosen partly because of the availability of data from the SELS but also in part because these, or similar, variables had been found to be important in previous research. Factors which appear to make a difference to mean GCSE performance include:
This table confirms many prior expectations, highlighting the interactions between religion and other social characteristics (Miller et al 1993). Catholic school-leavers have a greater likelihood of coming from larger families, of having both parents unemployed and of being eligible for free school meals (a proxy for Income Support) than their Protestant counterparts. In short, a greater proportion of Catholics than Protestants have characteristics which are known to be linked to lower educational attainment.
Table 3.4 and Chart 3.3 show that between Catholics and Protestants
there are significant differences in GCSE performance . Table
3.5 suggests that a number of family/personal characteristics
also have some bearing on GCSE attainment. These characteristics
vary systematically between Catholics and Protestants as demonstrated
in Table 3.6. The central question is therefore the extent to
which inter-religious differences in GCSE performance are a function
of the different SES and structural characteristics of the two
groups and how much can be attributed to religion alone. The data
used to answer these questions is discussed in the next section
which seeks to show the opportunities and the limitations of the
To set the scene for the analysis, the characteristics of the
SELS database are briefly described. The full database was not
used; listwise deletion of missing cases meant that 1,480 cases
were used in the analysis and that missing values were excluded
from all analyses. All data used in the models were unweighted.
The analysis was supplemented by data on the percentage per school
in receipt of free school meals and failing to pass GCSEs at grades
A-C published by DENI. The variables used in the analysis are
3.5.1 The outcome variable
The outcome variable was the sum of points scored per person GCSE
on completion of schooling where A=7 .. G=1. The maximum value
was 84 and the minimum value was 0 with a mean of 46.4.
These scores were standardised so the series had a mean of 0 and
a standard deviation of 1. No attempt was made to differentiate
the causes of scoring no points at GCSE in terms of not studying
for an examination or failing to sit an examination. The object
of interest was simply the social correlates of GCSE success.
3.5.2 The explanatory variables
The analysis included the Personal/family characteristics which were earlier shown to have a bearing on educational attainment at GCSE. These were augmented by school-level variables, namely the type of last school attended (eg whether secondary or grammar), or the sex status of school (eg whether co-educational or single sex), and the percentage of pupils per school in receipt of free school meals. A brief discussion of the nature of these variables, and some comments on the reasons for their use, is offered below.
Five individual-level explanatory variables all taken from individual responses to the SELS, were included in the investigation. Firstly, parental labour market status which was coded as a series of dummy variables (i.e. coded with an indicator of either unity or zero) to represent cases where both parents were employed, where one or the other was in work, or where both were unemployed/economically inactive. This variable was chosen, firstly, because it offered a useful proxy for socio-economic status especially when the distribution of employment is polarised between 'work-rich' and 'work-poor' households (Pahl 1988) but also, secondly, on the practical grounds that the SELS did not collect data on parental education and concerns about question-specific non-response bias to questions on parental occupation. Secondly, gender was included as a variable given its importance in British studies (cf Goldstein et al 1993). It was treated as a dummy and was equal to 1 when a respondent was female. The third variable used was religion, treated again as a dummy, and coded as 1 when the respondent was Catholic. Religion was used because of its similarity to ethnicity in British studies and the well-known (if not well understood) relationship between religion, labour market behaviour, educational opportunities and life-chances in Northern Ireland (Cormack and Osborne 1991). The fourth variable chosen was family size. This was simply the number of siblings given by the respondent, ranging between 0 and 12. with a mean of 2.7. The final individual-level variable chosen was personal eligibility for free school meals. This variable was selected because of its use in British analyses of deprivation (Jesson et al 1992) and the use of various free school meal-based measures in official indices. Again it was treated as a dummy and coded as 1 for any SELS respondent ineligible for free school meals.
In addition, three school-level variables, taken from the SELS and administrative sources, were used. Firstly, type of school last attended. This was chosen because of the selective nature of the Northern Ireland school system and the close relationships between social class, entry to grammar school and educational attainment (Osborne et al 1993). It was coded as a dummy variable which was set to 1 if the last school attended was a grammar school. As no measure of prior attainment was available it could also be treated (with caution) as a very crude proxy for attainment at age 11 given the relationship between success at age 11 and entry to grammar school (Sutherland 1993). Secondly, it was considered useful to incorporate a measure of whether a school was single-sex or mixed-sex because Northern Ireland has proportionately more single-sex schools than Great Britain. This variable was a dummy coded as 1 when a SELS respondent's last school was mixed sex. The final school-level variable was the percentage of all-age pupils in receipt of free school meals per school. This was taken from data published by DENI and ranged from 0% to 75% with a mean of 25% and was designed to investigate the effect of the school context of schooling on individual attainment. The full descriptions of the variables together with the way in which they are used in the analysis is fully described in Appendix 2.
The use of SELS data in combination with information from administrative
databases gives the opportunity to examine educational attainment
in a way hitherto impossible. For the first time in Northern Ireland
the effects of individual/family SES are considered simultaneously
along with a school contextual effect as measured by the percentage
of pupils per school in receipt of free school meals. Moreover,
individual data on religion are available, as distinct from the
religious affiliation of school attended used in earlier research
(Gallagher 1988; Daly 1991).
3.5.3 Limitations of the SELS data
No measures of prior attainment were available in the SELS database.
Since performance in previous assessments has been shown to be
closely related to performance at 16 years old (Daly 1991), this
means that the models that are constructed, as they omit an important
variable and may therefore lack precision. The variables that
are available from the SELS (personal/family background and social
characteristics) are less clearly related to educational processes.
On the one hand, these characteristics appear to differentiate
very effectively between different groups of school-leavers. On
the other, care should be taken in the interpretation of the models.
They can only describe the relationships within the SELS dataset
as it exists. If answers are sought to questions about school
effectiveness, or to how different scores and religious differences
in the transfer procedure translate to success at GCSE, the SELS
database is plainly insufficient. However, the SELS database
can be used to summarise the relationship between SES, personal/family
background religion and attainment at GCSE and to show how social
characteristics influence attainment for 16 year olds.
To disentangle this web of factors, and to separate out the independent effect of each variable on GCSE attainment, a multivariate technique was used known as multi-level modelling (MLM) . The objective of this was to show how knowledge of pupils' personal/family characteristics, either singly or together, add to our understanding of variations in individual GCSE scores. The strategy was simply to add explanatory variables to the model to investigate how much they explained differences in GCSE attainment and to investigate the significance of religion once other personal/family background variables were taken into account,
The full results of the models, and information on how to interpret them, are set out in Appendix 2. Table 2a in Appendix 2 shows the fixed coefficients when GCSE attainment is regressed on personal, family and school background. The so-called null model has no explanatory variables except a constant term. It represents the base from which the modelling procedure starts.
Model one introduces religion to the analysis. Looking at the
negative sign of the coefficient associated with DCATH, it suggests
that Catholics have lower GCSE attainment than Protestants. As
a rule of thumb, the coefficient is said to be significant at
the 0.05 significance level when it is twice the size of its standard
error. In this case, as -0.159 is not quite twice as large as
0.089, the religion coefficient just falls short of statistical
Model two adds further terms for personal/family characteristics to address some of the questions raised by Osborne et al (1991) about the impact of SES and family size on attainment. The size of the religion coefficient (DCATH) is markedly reduced once gender, parental labour market status, family size and eligibility for free school meals are controlled for in the model. Religion is now markedly insignificant in its effects. On the other hand there are statistically significant effects associated with parental labour market status, family size gender and free school meal eligibility. School-leavers with both parents in work, or only their father in work, perform much better at GCSE than those with unemployed/inactive parents; females perform better than males, school-leavers who were ineligible for free school meals have higher GCSE attainment than those entitled to this benefit and school-leavers from larger families tend to perform less well than those from smaller families. These effects are very much like those noted in Britain (see for example, Jesson et al 1992; Jesson and Gray 1993). Interestingly, these studies also isolated similar social factors as being of significance.
Model three drops all individual-level variables (except religion) and adds instead school-level variables. Religion remains insignificant. The largest single effect is associated with grammar school membership (about 90% of a standard deviation). This is hardly surprising given the selective nature of the Northern Ireland school system. A small, but significant effect is associated with percentage per school in receipt of free school meals; as this percentage rises so individual performances, on average, fall thus suggesting the possible importance of social context.
Model four is the fullest that can be specified. Religion remains insignificant. If anything, being a Catholic seems to have a positive effect on GCSE attainment everything else being equal. Strong effects can be observed associated with grammar school membership, gender, parental labour market status and eligibility for free school meals. To summarise these, school-leavers from grammar schools are estimated to perform better than those from secondary schools; females better than males; those with both parents employed or father employed better than those with unemployed/inactive parents, and those ineligible for free school meals better than those who were eligible.
Unlike Ordinary Least Squares Regression (OLS), MLM does not provide an R2 statistic to summarise the explanatory power of each model. Also unlike OLS, MLM, because it recognises the hierarchical structure of data (in this case school-leavers clustered within schools), can apportion variance between schools and between pupils.
However, the variance each model explains, relative to the null model, can be used to approximate R2. In Table 2b the variance reduction of the f our models is therefore summarised showing how much variance between pupils and between schools is explained.
Model one, with only a constant term and a religion term, explains only 0.7% of the variance in GCSE scores. It explains about 0. 2% of the variance in scores between pupils and about 1.5% of the inter-school variance in GCSE scores. Model two, as would be expected given its greater number of explanatory variables, has a greater variance reduction. It is more successful at school level than pupil level. Model three, with only school-level variables except for religion at the individual level, is plainly more powerful at school-level. This means that about 73% of the variance in individual GCSE performance between schools in Northern Ireland can be 'explained' if we know whether they attended a grammar school or secondary school, a co-educational or single-sex school and the social context of the pupil intake of the school attended. The dominant variable in this is grammar school membership. Model four, the fullest model, explains about a third of the variance in GCSE scores and again is much more effective at school-level.
The explanatory power of the pupil-level variables appears to be limited. This means that knowledge of whether a pupil is eligible for free school meals, is male or female, is Catholic or Protestant, and has unemployed or employed parents, on their own say relatively little about likely GCSE performance. This lack of precision matches prior expectations of the importance of these variables given the well-known importance of prior attainment as a major determinant of examination success at age 16 (Daly 1991; Jesson et al 1992). Some of the effects of personal/family background might be diminished if a measure of attainment at age 11 was available; certainly it would be interesting to see how such a model performed in comparison with those used in this chapter.
However, the relationships between social characteristics and
attainment at GCSE are effectively summarised. What the final
model says is that if a Catholic pupil is identical to a Protestant
pupil in type of last school attended, eligibility for free school
meals, family size, gender, parental labour market status and
percentage of pupils in his/her school in receipt of free school
meals, then there is no statistically significant difference in
their GCSE performance. Catholics do not appear to be disadvantaged
in any significant way at GCSE once family background and school
type have been taken into account; a finding supported at school
level by other Northern Ireland and British research (Daly 1991;
Goldstein et al 1993). This implies that religious differentials
in terms of points scored at GCSE can be largely attributed to
the operation of family background variables and access to grammar
school and are artifacts of the differential SES structures of
the two communities.
A similar type of analysis can be conducted for attainment at A Level. Table 3.7 shows the difference between Catholics and Protestants in the number of A Levels passed at grade E or higher.
Chart 3.4 shows inter-religious differences in terms of the points scored at A Level (where A=5. .E=1). There are statistically significant differences between Catholics and Protestants at the 0.05 significance level.
This information refers to SELS database as a whole. However.
there are particular problems in understanding attainment at A
Level because A Levels are a post-compulsory education qualification.
In addition to prior attainment, school effects. socioeconomic
status and school examinations policy, there is the added difficulty
of participation rates in post-compulsory education. The question
of who stays-on at school or remains in education is clearly related
to that of who passes A Levels; but in its wider implications
it is quite different and will be dealt with in greater detail
in Chapter 4. The only comment that will be offered for the moment
is that the nature of participation in post-compulsory education
in Northern Ireland is complex. Staying-on rates vary both by
gender and religion with Catholic males more likely to stay in
education past the age of 16 than their Protestant counterparts
and Protestant females more likely to remain in education than
their Catholic counterparts.
A full analysis of the determinants of participation in post- compulsory education lies outside the remit of this report but would be an interesting topic for further work. However, the factors that influence the decision to leave education probably include local labour market conditions (Raffe and Willms 1989; Gray et al 1992), individual educational attainment (Jesson et al 1992), the demand and supply of places in higher education and views of education as a consumption or investment good.
These complicating f actors need to be recognized but may be treated
as contextual in terms of the present work. The analysis of the
SELS data will therefore be based on all those who stayed on to
upper sixth, hence by-passing questions of the determinants of
participation in post-compulsory education. Note too, that nothing
is said about the contribution of other sectors (such as colleges
of further education) to the pool of qualifications of young people
and the potential impact on differential educational attainment
by religion. This is ignored in this analysis of A Level attainment
as it was for GCSES.
The same hypotheses used to explain differential attainment at 16, namely differences in SES between the two communities and inequalities in the type of school attended, can be applied to A Level attainment. Therefore, the same modelling procedure as for GCSEs is used, as are the same explanatory variables. The only additional variable used is attainment at GCSE which is used as a measure of prior achievement. The outcome variable was points scored at A Level (where A=5 . . E=1). These variables were not transformed into z-scores but were instead expressed as normal scores. The reason for this normalisation of A Level and GCSE point scores was for the practical reason that the series. in this case, were more easily modelled. There were 576 cases after individuals with missing values were deleted as the dataset for analysis was restricted to upper-sixth school-leavers.
The null model, in Table 2c. shows the amount of variance in the model. Model one begins the analysis by including the religion variable; using this dataset it is insignificant. Model two adds other personal/family characteristics; again these are all statistically significant. Model three incorporates all these variables and also adds points scored at GCSE; this seems to be highly statistically significant with high GCSE point scores being positively related to high attainment at A Level. Model four drops all individual-level variables, except religion, and only incorporates school-level variables. The only variable that is statistically significant is type of last school attended; grammar school membership is associated with higher A Level scores. The final analysis, Model five, places all the explanatory variables together to consider their joint effects. Once a measure of prior attainment is included in the model, the grammar school effect diminishes to statistical insignificance.
Religion, as is the case with all other measures of social background, appears to have little influence on A Level attainment. In each model, religion is statistically insignificant. This also applies to parental labour market status, family size, gender and free school meal eligibility which persist in statistical insignificance and in some case change signs from model to model.
If few explanatory variables appear to be important in influencing A Level performance for upper-sixth form leavers , the explanatory power of the models are somewhat higher than for the GCSE models, The variance explained by Model one (with just religion) is negligible at less than 1%. With only personal/family characteristics, the variance explained is higher but remains less than 5% (though it nears 13% for between-school variance), Adding points scored at GCSE raises the total variance explained to near 45% and to over 96% for the between- school component. Removing these individual-level variables and substituting them with school-level variables reduces the explanatory power of the pupil-level part of the model from 28.10% to 1.55% though the school part is untouched. The final model explains in total about 47% of the variance in the data which is composed of 29,79% at pupil level and over 99% of between-school variance.
These models suggest that once a pupil is in sixth form. personal/family characteristics,' including religion, are insignificant in explaining A Level performance, The most important variables, depending on how the data are analysed, are attainment at GCSE and type of last school attended. Attainment at GCSE diminishes the size of the effect associated with grammar school membership and so can be considered to be more 'powerful'.
This analysis, given the sample on which it is based, is naturally
restricted. Sixth-form pupils are to some extent self-selected.
Moreover, little can be said about the take-up of post-compulsory
education and its effects on A Level attainment. This is possibly
a fruitful area for further research.
Religion appears to be statistically insignificant as an influence on attainment at GCSE and A Level once SES and type of last school attended were controlled for in the analysis. This suggests that differences in mean SES between the communities, and in the proportion of pupils going to secondary or grammar schools, accounts for differential examination performance for 16 and 18 year olds. The findings for GCSE are more robust than those for A Levels as they refer to an outcome that is measured largely at the completion of compulsory schooling. Those for A Levels are only for upper-sixth form school-leavers and ignore the social processes that lead young people to stay-on in post-compulsory education.
These findings are similar to those reported on the performance of Catholic schools in Scotland where denominational differences in performance can be explained by lower Catholic SES (McPherson and Willms 1986; Willms 1992), Willms (1992) suggests that findings that argue f or the importance of mean school SES are particularly relevant to comparisons between Catholics and non-Catholics in Scotland. This is likely to be the case, too, in Northern Ireland as it is probable that the lower mean SES of the Catholic community will result in a lower mean SES for pupil intakes to Catholic schools. This means that a Catholic pupil might suffer from low family SES; but in any event, whether family SES is low or high, in some cases low school SES will have an effect on individual examination performances.
At both GCSE and A Level, Catholics perform as well as non- Catholics once SES and type of last school attended are accounted for in the analysis. This being so, questions are raised about why Catholics perform as well (or perhaps better at GCSE) than non-Catholics everything else being equal.
Part of the answer lies, as was suggested, in differences in SES and type of last school attended. However, there is evidence that the relationship between SES and educational outcomes at GCSE might be more complex than hitherto imagined. An analysis which is not presented in the report, found that there are interesting interactions between SES (as measured by free school meal eligibility) and religion, Catholics who were ineligible for free school meals fared significantly better at GCSE than a reference category of non-Catholics eligible for free school meals. On the other hand, non-Catholics ineligible for free school meals and Catholics eligible for free school meals, did not perform significantly differently from non-Catholics who were eligible for free school meals. This suggests that the relationship between SES and outcomes at GCSE may differ by religion. This reinforces the conclusion that SES is one of the major factors that explain differential educational outcomes at GCSE in Northern Ireland but also indicates a potentially complex relationship between SES, religion and educational outcomes in Northern Ireland.
Ways in which these complex relationships might be structured are suggested by literature on the Scottish experience. Firstly. Catholics might respond to observed opportunities in the labour market in that they perceive effort at school as the only means to overcome possible biases. Secondly, Catholics might suffer from blocked social mobility in that if two pupils had equal 'educability', one Catholic and one not, then it would be expected that the Catholics would on average have lower family SES (Paterson 1991b). Even though it seems that some progress has been made towards unravelling the links between religion, SES and educational attainment in Northern Ireland, the links are not fully understood and further research is needed.
The first potentially fruitful line of enquiry would be to control for prior attainment at age 11 in the transfer procedure. This would give the opportunity of more precise modelling that could examine the issues that have been raised by this exploratory work. The models presented in this chapter show that raw religious differentials in attainment are reduced once SES and type of last school attended are taken into account. This is not quite the same as saying that pupils with the same transfer score at age 11 perform the same regardless of religious background, controlling for type of school attended and SES.
Differences in transfer behaviour mean that Catholics with a given score are less likely to enter grammar school than a non- Catholic. It is also known that the probability of entering grammar school is also closely related to family SES (Sutherland 1993), and here the same arguments as were made about poor Catholic examination performance reappear (the result of lower relative SES or religion-specific differences?). More research about the relationships between SES, religion, transfers to grammar school and educational attainment would go some way to explore these issues. Indeed, this type of augmented analysis would enable a better estimate of variations in the response to schooling between Catholics and Protestants, controlling for SES, and so enable the two hypotheses about Catholic attainment advanced in Scotland to be tested in Northern Ireland.
A second area for research would be to investigate the impact of school-examination policy. The analyses performed in this chapter have taken attainment as given. However, it was shown that there were differences in the proportions of the two communities who sit examinations at both GCSE and A Level. In both cases we do not know the reason for this, At GCSE school- level variables of ethos and examination entry might be important (some Catholic pupils might not be entered for examinations and so not given an opportunity to achieve some passes) . At A Level. the causes of continued participation in education, and their religion-specific impacts, are simply not well understood in Northern Ireland. Lower educational attainment for Catholics may therefore be a consequence of f actors beyond the control of individual pupils, The value of explanations which refer to the school-level variables of ethos, examination policy and staying-on in education ought to be tested.
The analyses reported suggest that Catholics perform as well as
non-Catholics everything else being equal. This does not mean
that all is well in the Northern Ireland education system. There
are many parts of the relationship between religion, SES and attainment
that are not understood. It is only in academic models that everything
else can be held equal to isolate the importance of a religion
effect. Subject differences persist; in real terms religious differences
in examination attainment continue and there is a close enough
relationship between social deprivation and religion to suggest
all else is not equal. Moreover, the equality between Catholics
and Protestants in educational outcomes suggested by the controlled
environment of the models, may not transfer to equality after
school. Questions about post-school destination will therefore
be the subject of Chapter 4.
 The sum of cases for all variables do not always equal 11, 600 because of missing values being excluded for certain variables
 FSM indicates individual eligiblity for free school meals
 This mean differs from that reported in Table 3. database used in the analysis excludes all cases with missing values. This applies to all other variables whose means are reported from now on.
 This technique is explained more thoroughly in Appendix 2
 This chart was based on data that excluded some cases
with missing values