Can higher education compensate for society? Modelling the determinants of academic success at university

This paper examines the role that social characteristics play in determining the academic success of students who begin university with roughly similar entry grades. The data used were drawn from the administrative records of over 38,000 UK-domiciled undergraduate students from one British university between 1998 and 2009. Results show that the characteristics of entrants have varied only slightly over this period and intake is still largely in favour of ‘traditional’ entrants: namely those from professional occupational backgrounds, the privately educated and those of traditional age. The relationship between background characteristics and eventual academic success also reflects patterns seen at earlier education stages. However, when prior attainment was taken into account, the link between degree outcome and many social characteristics does diminish – notably for students who were privately educated and who came from professional occupational groups. This suggests that once students have overcome barriers to admission, it is entry grades rather than social characteristics that may most strongly influence eventual academic success.


Introduction
The title of this paper is taken from Basil Bernstein's widely referenced work 'Education Cannot Compensate for Society' that appeared in New Society in 1970. Bernstein's concern in this article was largely about the contexts in which children learn and the 'social assumptions underlying the organisation, distribution and evaluation of knowledge ' (1970, 347). In particular, he contested the assumption that there is just oneoften a largely middle classway of educating our children. Arguing that the compensatory programmes which support those children of 'low social class whose material circumstances are inadequate' (Bernstein 1970, 344; original emphasis) offer a deficit model of education in which schooling is expected to compensate for something that is perceived to be missing from within the family.
Although Bernstein was not explicit here in suggesting that education policy cannot act as a replacement for social policy, the title of his paper has become synonymous with the argument that social class and education are 'inextricably linked'; that academic progress is largely determined by family circumstances and that there is little that schooling can do about it (Davison 2010, 145;Coleman et al. 1966;Halsey, Heath, and Ridge 1980). While we might wish to 'reject the stance that attempts to reduce educational inequalities are inevitably futile' (Power 2008, 19), it is well established that academic success is socially stratified and that educational inequalities persist across the life course (for example, Smith 2012). Although it is important to note that the academic achievement of children and young people in schools in England continues to improve.
The expansion of the UK higher education system over the last three decades has provided numerous opportunities for examining the extent to which social inequalities are replicated in post-compulsory education. This is particularly pertinent as the higher education student population becomes more diverse as a consequence of initiatives to both widen and increase participation during this phase (for example, DfES 2003). Therefore it is important to consider the extent to which educational inequalities that are present during the compulsory phase continue into higher education. For example, it might be argued that once students have met threshold academic levels and gained admittance to university, the effect of social characteristics on academic attainment might be reduced, as entrants will now share a similar educational experience over the course of their undergraduate careers. With this in mind, the main aim of this paper is to examine the role that social characteristics play in determining eventual academic success for students who enter university with roughly similar entry grades. Essentially, it asks how sensitive the higher education system is to background characteristics such as class, ethnicity and previous schooling experience. The paper will address the following research questions: (1) What are the characteristics of full-time, home-domiciled undergraduate students studying at this university? (2) To what extent have these characteristics changed since the late 1990s? (3) Once prior attainment has been controlled for, what is the relationship between student background characteristics and the likelihood of graduating with a 'good' degree?
The impetus for the Robbins Review, as it was to become known, was the predicted shortfall of university places caused by the large numbers of young people who were born after the War and who would be eligible to enter higher education in the late 1960s (Committee on Higher Education 1963). The university system at the time of this review was arguably a system for the elite, starkly segregated according to social background. While Robbins recognised that 'the reserves of untapped ability may be greatest in the poorest sections of the community' (Committee on Higher Education 1963, para. 142), there were no specific recommendations about how participation might be widened. In fact it was not until the 1990s that widening access to university became a key focus of national education policy with the publication, towards the end of that decade, from the National Committee of Inquiry into Higher Education (NCIHE 1997). This report placed an emphasis on widening access to an expanding university system and its aims were quite explicit: that the future of higher education in the United Kingdom would require institutions to: … encourage and enable all studentswhether they demonstrate the highest intellectual potential or whether they have struggled to reach the threshold of higher educationto achieve beyond their expectations. (NCIHE 1997, para. 5) As Figure 1 shows, the number of people who apply and are accepted to study at university has increased steadily over the last 25 years. This growth in applications is mirrored by growth in the number of acceptances, as the number of university places available increased to accommodate demand. Even though higher education has been transformed over this period from an 'elite to a mass experience', the experience of students from less financially well-off families is 'impoverished' in comparison with those from more affluent groups (Furlong and Cartmel 2009, 121). While going to university has now become part of the 'normal taken for granted experience of the middle classes' (Furlong and Cartmel 2009, 145; see also Blanden and Machin 2004;Mangan et al. 2010), it is still the case that fewer than one in five young people from the most disadvantaged geographical areas enter higher education compared with more than one in two for the most advantaged areas (HEFCE 2011). In short, recent research suggests that two things have happened to university participation in the United Kingdom. First, the number of people entering higher education has increased and applicants who have the required grades have usually been able to secure a place. On the other hand, there is compelling evidence to suggest that many students from previously under-represented groups, notably older students as well as those from certain ethnic minority and occupational groups, are less likely to study at 'elite' 1 universities. As Ball (2013, 15) argues, 'if we look at participation in higher education, the continuation of class inequalities is very clear'. However, in this paper the aim is to examine inequalities in academic outcomes at university and to consider the extent to which class differences, in particular, persist at the end of an undergraduate degree The determinants of academic success in higher education The issue of which characteristics are associated with educational success at university is a topical one. The move towards a mass system of higher education, the rising cost of tuition fees and an increasingly competitive graduate labour market are all factors that may have implications for ensuring 'fair and equal' access to university as well as reinforcing higher education's purported status as 'an engine of economic growth and social justice' (Naylor and Smith 2004, 416).
Prior attainment is usually the key factor determining whether or not a student is offered a place in higher education. However, research has shown that academic success during compulsory education is predicated on attainment at the previous educational stage and that at the aggregate level young people from less affluent social groups achieve at lower levels throughout schooling (for example, Gorard et al. 2007). This means that when it comes to applying for and being successfully admitted to university, intake is likely to be socially as well as academically stratified. However, once students have met threshold academic levels and gained admittance to higher education, the effect of social characteristics on academic attainment might be less pronounced, as students could now share a relatively similar educational experience.
Previous work into the determinants of educational success in higher education has identified an association between certain student characteristics and the likelihood of achieving a 'good' 2 degree. For example, research has shown that, in general, female students are more likely than men to achieve an upper-second degree or higher, but this is contingent on the subject studied (for example, Barrow, Reilly, and Woodfield 2009;Naylor and Smith 2004;Richardson and Woodley 2003). There is also some evidence that women tend to be less likely to graduate with a first-class degree. This may be accounted for by the increased propensity for male undergraduates to study subjects that tend to award a higher proportion of first-class degrees, although this pattern is not apparent in all studies (see Farsides and Woodfield 2007;Mellanby, Martin, and O'Doherty 2000;Woodfield and Earl-Novell 2006). Recent initiatives to widen access to higher education have focused on the participation of older students as well as those belonging to ethnic minority groups. However, according to Naylor and Smith (2004, 418) there appears to be 'no general consensus' on the effect of age upon degree success with findings varying depending on the datasets used. With regard to attainment according to ethnic group, research has tended to suggest that those from minority ethnic groups are less likely than white students to achieve a 'good' degree (Leslie 2005;Richardson 2008). A number of explanations have been offered: students from ethnic minority groups are more likely to be accepted into higher education (they have a higher age participation index, possibly indicating a broader academic mix of students), they tend to be less well qualified upon entry and they also tend to study subjects that are less likely to award first-class and uppersecond-class degrees.
Perhaps unsurprisingly there is a large body of evidence that indicates there is a substantial relationship between A-level score and the likelihood of doing well at university (see, for example, Barrow, Reilly, and Woodfield 2009;McManus et al. 2003;McNabb, Sarmistha, and Sloane 2002). Smith and Naylor (2001), for example, estimate that each additional grade awarded in each of three A-level subjects (i.e. from BBB to AAA) is associated with around a 10 percentage point increase in the likelihood of graduating with a 'good' degree. The impact of previous education experiences have also been widely scrutinised, particularly with regard to the effect that school sector has on degree outcomes. Work on this so-called 'school-type effect' suggests that those who are educated in the independent sector perform less well at university than 'equivalent' students from the state sector (see, for example, HEFCE 2003; Ogg, Zimdars, and Heath 2009;Naylor and Smith 2004). According to Smith and Naylor (2001), in order to equalise the likelihood of obtaining a good degree for those who come from the state and the independent sector, applicants from the latter group would need to achieve approximately one grade higher in each of three A-levels. These findings could lend support to arguments for a process of affirmative action in university admissions whereby lower admissions offers are made for state school students than for those from the independent sector (but see Smithers 2003;Guardian 2003). This evidence is also congruent with the hypothesis that independent schools give an 'artificial' short-term boost to A-level performance which is not sustained at universitya finding that is worthy of further future consideration. In addition, as Naylor and Smith (2004) argue, the finding that a selective education may have a limited effect on eventual degree success might suggest a similar relationship for those who come from a more 'privileged' family background. They argue that a student with BBB at A-level from a professional occupational background could have less potential for successful attainment at degree level than an equivalent student from a routine occupational background who perhaps had a less privileged academic upbringing. As with those from state schools, the student from the routine occupational background would then perform at a higher level at university than their peer with similar A-level grades but from a professional occupational background. There is only limited empirical evidence to support this hypothesis, however, and as Naylor and Smith (2004, 439) report, 'academic performance at university is better the more "advantaged" is the student's home background'.
So in general, previous work suggests that for equivalent students there are differences in degree performance which are associated with characteristics such as ethnicity, age, gender, occupational background, pre-university educational attainment and degree subject. In the present study these differences are reconsidered for successive cohorts of students from one 'elite' university. Data analysis extends over nine years and provides unparalleled detail of the characteristics and academic outcomes of tens of thousands of undergraduate students.

Research approach
As described above, a key aim of this paper is to examine the determinants of degree success. The paper considers the relationship between the social, geographical and academic characteristics of entrants and their academic outcomes at the end of university. The data used in this analysis are drawn from individual student administrative records that were collected by one large, British university. Data were available for the whole of the university student population, but for the purpose of this paper were restricted to fulltime, home-domiciled undergraduate students who entered the university between 1998 and 2006. This corresponds to the majority of students graduating between 2001 and 2009. There were two reasons for selecting this group for the study: first, there are relatively few part-time students at this university; and second, considering only UK-domiciled entrants focuses on a group who will have shared a reasonably common pre-university educational experience.
The final population comprised between 3000 and 4000 students in each cohort, giving a total of 38,236 students. Despite providing a rich source of high-quality contextual data, as with all secondary (and particularly administrative) datasets, information was not available on all of the variables that could plausibly be related to academic outcomes. For example, the data can tell us nothing about the many unobserved or unrecorded factors or events that will have occurred in these students' lives, as well as their levels of motivation and interest in their subject or the quality of teaching on their course (see, for example, Mellanby, Zimdars, and Cortina-Borja 2013).
In addition to summative data on a student's degree performance, the data used in this study provides a wealth of information on each individual. The variables used in the analyses can be grouped into the following categories.

Birth characteristics
These variables include all of those characteristics respondents could be considered to have from birth, including: sex; ethnic group; parental occupational class and age. Categorising student data in this way is not without difficulty, however. The large number of ethnic group categories can result in small cell sizes when the data are analysed, for example. The distribution of students' ages reflected the structure of common post-compulsory educational trajectories but also included substantial numbers of 'mature' entrants. Additionally, allocating higher education students to an occupational group that depends on their family circumstances can be particularly problematic, especially for older entrants (Gorard 2008;Harrison and Hatt 2009). With this in mind, data preparation and analysis was informed by theoretical, practical and statistical considerations and a number of adjustments were made to the variables.

Ethnic group
This characteristic was self-declared by the applicant and coded into 10 categories as specified by the University Central Admissions Authority. In order to aid comparison, students were grouped as either being from the majority ethnic group (i.e. white) or from a minority ethnic group. The limitations of this approach are recognised and an analysis of patterns for the largest minority ethnic groups was also undertaken; however, due to lack of space they are not included in this paper.

Occupational class
Data on occupational class were collected in four categories based on the Standard Occupational Classification 2000 scale. For the purpose of multivariate analysis these were collapsed into two groups: higher/lower professional, and intermediate/routine occupations. This categorisation has been used in many previous analyses of class differences in higher education participation.

Age
Two categories were used: students aged 17-20 years on entry, and those aged 21+. This division reflects the distinction between students entering higher education shortly after leaving school or college and those 'mature' entrants who participate later in life.

Geographical characteristics
Geographical characteristics of the student population are important to consider when investigating participation and outcomes in higher education. This is particularly relevant in terms of an institution having a representative mix of students from the local area and is a key element of this university's fair access agreement (Office for Fair Access 2014). The data used in this analysis are derived from the students' home postcode at the time they made their original application to the university. They enable the distance from each student's home address to the university campus to be calculated, in turn allowing them to be categorised as 'local' to the university (notionally residing less than 20 km from campus) or 'non-local' (residing at a distance of greater than 20 km from campus). By combining these postcode data with HEFCE Participation of Local Areas maps, it was also possible to classify students according to rates of post-compulsory educational participation in their area of residence. Two of the Participation of Local Areas indicators of participation were used in the analyses: the young participation rate; and the adult higher education engagement rate. The Participation of Local Areas classifications are presented as five ordered quintile groups (HEFCE 2013).

Academic characteristics
These include the previous educational establishment that the student attended (four categories of data) and entry qualifications (in terms of A-level grades). Students were separated into two groups according to their entry qualifications: those who entered with grades ABB or above, and those whose grades were below ABB. ABB+ was chosen to differentiate students with the highest entry grades because at the time of writing this is the level the government adopts as part of its 'high grades' policy that allows universities and colleges to recruit unlimited numbers of students with the highest grades (HEFCE 2013).

Academic outcomes
The academic outcome variable used in the multivariate analysis was dichotomous, with students divided into those who graduated with an upper-secondclass or first-class degree and those who received a lower-second degree or below. 3 Across the cohort, 66% of students fell into the upper-second/first category with 34% graduating with a lower-second-class degree, third-class degree or fail, a distribution of grades that reflects the pattern nationally (HESA 2014). The reason for collapsing the data in this way is because of the premium that appears to be attached to an upper-second-class or first-class degree in today's graduate labour market. For example, research by the Association of Graduate Recruiters suggests that over three-quarters of graduate employers will only interview candidates with an upper-second-class degree or higher (BBC 2010). Similarly those who graduate with a lowersecond-class degree or below might also see many opportunities for postgraduate study closed to them: entry for many initial teacher training courses, particularly those who train for the primary sector, require at least an upper-second-class degree. Thus the 'life chances available to a student with an upper-second-class and a student with a lower-second-class degree are very different indeed' (Seaton 2011), and it is pertinent that potential determinants of success which lie outside the influence of the university itself (in terms of background, geographical and academic characteristics) are subject to greater scrutiny.

Analysis
Patterns in the data were first explored using univariate and bivariate techniques. These are described before the results of the multivariate analyses are presented. Logistic regression analysis was used to identify the social, economic and educational characteristics most strongly associated with achieving an upper-second-class degree or higher at the end of university. A number of different models were run and variables were retained in the models based on theoretical assumptions derived from the literature and the extent to which they contributed to the explanatory power of the model. At all times the aim was to balance the explanatory power of the model with parsimony and the need to control for theoretically important independent variables. As the data used in this model represent the entire population of home-domiciled undergraduate students at this university, no sampling was necessary and therefore it would not be appropriate to use inferential statistics to draw probability-based conclusions about levels of statistical significance (for example, Gorard 2003;White 2014).
Variables were entered into the models in a series of 'blocks'. These blocks correspond to the chronological order in which they occurred in a students' life in an attempt to differentiate between the relative explanatory power of the different groups of variables. Only variables that improved the predictive power of the model were retained at this stage. The first block contained information relating to the social and economic backgrounds of students and included variables relating to sex, age, occupational class and ethnicity. These variables were all dichotomous. Students were divided according to whether they were: male or female; 'traditional age' or 'mature' entrants; from a professional or other occupational background; and whether they identified themselves as 'white' or as a member of a minority ethnic group. The second block features geographical characteristics, while the third includes information relating to students' education before entering higher education. Students were grouped according to whether they had attended a fee-paying or state-funded school or college (including comprehensive and grammar schools as well as further education colleges) when they applied to the university and whether they had achieved grades of AAB or above in their A-levels. Block four contained interaction terms for a number of key variables. For reasons of space, discussion is limited to patterns across all subjects (an analysis of differences between subjects is presented elsewhere; Smith and White 2014).

Results
The resulting multivariate models are discussed later in the paper. Before examining these in detail, the results of exploratory univariate and bivariate analyses are described directly below.
The characteristics of home-domiciled undergraduate students Before considering the characteristics of students who attend this university, it is worth examining the national context for the same period. Selected background characteristics for all home-domiciled undergraduate entrants to all UK universities between 1998 and 2006 are shown in Figure 2. National patterns of participation in terms of students' age, ethnic group and occupational background remain relatively unchanged over the period. In comparison with the national profile, the university considered here recruited fewer students aged 21 and over (around 10% compared with 16% nationally) and more students from higher professional occupational backgrounds (around 24% compared with 18% nationally) (see Table 1 for institutional data). Looking at recruitment trends within the university from 1998 to 2006, there has been a slight increase in the proportion of female entrants, from 53% to 60%, as well as those from higher professional occupational groups, which has increased from 21% to 26% (Table 1). Despite a reduction in the proportion of entrants from independent schools over the same period, the university still recruited a disproportionate number of students from this sector. In 1998, 26% of entrants were privately educated, and by 2006 this had fallen to 20%. This is approximately three times the national level of 7-8% of undergraduate students coming from the fee-paying sector (UCAS 2013). In terms of geographical characteristics, there has been a slight increase in the proportion of 'local' entrants (here defined as living within 20 km of campus), with this group comprising one-quarter of all entrants in 2006. Recruitment from areas with low levels of participation in post-compulsory education and where few adults hold higher education qualifications continues to be lower than the national average. This analysis indicates not only that the demographics of the university have shown little variation over the last decade, but that entrants are disproportionately from groups that are over-represented in higher education in the United Kingdom. Over the nine years studied there has been little progress in terms of widening participation at this institution; indeed, the university appears not only to mirror national patterns in terms of changing rates of participation among underrepresented groups (for example, Smith and Gorard 2011), it appears to represent an exaggerated 'elite' version of them.

The determinants of degree performance
This section presents the findings from the multivariate analysis in two parts. First, it considers the relationship between prior attainment (here A-level success) and student characteristics. Second, a fuller model is then developed that explores the relationship between degree outcome and a number of key explanatory variables (including prior attainment), as well as accounting for any interactions between theoretically important variables.
The relationship between prior attainment and student characteristics As outlined above, prior attainment is one of the key determinants of academic success at university. This is unsurprising as it might be expected that students who were academically successful at school or college would also be successful at university. However, research has also shown that measures of prior attainment, such as the A-level, are not socially neutral; ABB+ students are likely to be different from lower attaining students in their social characteristics as well as their academic ability. Indeed, this is a key problem with using prior attainment data to predict future academic success because academic attainment can be socially stratified: girls often outperform boys in certain subject areas and students who are eligible for free school meals perform at the lowest levels on aggregate measures throughout compulsory schooling.
In this first model, to examine the interaction between prior attainment and student background characteristics, a binary logistic regression model was run with A-level grades as the dependent (or outcome) variable. Students were allocated to one of two groups: those who achieved ABB+ and those who achieved lower A-level grades. Table 2 presents the fit of the model and provides the odds ratios for students with different educational characteristics achieving grades ABB or higher at A-level for each of the nine cohorts. A number of points are worth noting. First, students who were educated in the state sector were less likely to achieve ABB+ than those from the independent sector (odds ratios are less than one); so students entering the university with the highest grades were already likely to be different from other students because they were more likely to have been educated privately. A second feature is that female students were more likely to have higher entry grades (for some years they were almost twice as likely as males) and there is a similar, although less pronounced, pattern for non-local students. Interestingly there appears to be little difference in the likelihood of achieving ABB+ for students from different professional occupational backgrounds. However, this model is relatively weak in terms of its explanatory power and, given that none of the models explain more than a minority of the variation in degree outcomes, it is important not to place too much weight on the absolute size of the odds ratios associated with individual characteristics. In general, the model suggests that being female, not local and privately educated are associated with a somewhat enhanced likelihood of attaining ABB+ at A-level. Indeed this analysis offers support to previous work that suggested academic outcomes from schooling are socially stratified, although interestingly differences according to parental occupational group and membership of the ethnic majority group do not appear to be very pronounced at this stage.
The relationship between student characteristics and degree outcome This section presents the findings for the multivariate model that considers the relationship between all of the explanatory variables (including prior A-level attainment) and degree outcome. Adding prior attainment to the model now lets us consider the impact of student characteristics on higher education attainment while taking prior attainment into account. In order to gauge the effect of these different variables on the final model, variables were entered in four blocks (birth, geographic and academic characteristics, followed by interaction terms in Model 4; see Table 3). Data are presented for the entire cohort (outputs for each individual cohort are presented in Appendix 1).
By far the strongest determinant of degree success is, unsurprisingly, A-level grades on entry. Being educated in the state sector, coming from a professional family, being white, being female and coming more than 20 km from the university are all associated with better degree performance. Students who were white appear to be twice as likely to graduate with an upper-second-class or first-class degree than non-white students (odds ratio = 2.03), although this effect appears to diminish once geographic and academic characteristics are taken into account in Models 2 and 3. For students from professional backgrounds the odds ratios are close to one, suggesting little difference in the chances of students from professional and non-professional backgrounds graduating with a 'good' degree. For those from state schools, odds ratios are around 1.2, showing a slight advantage over those from independent schools. Note how this is different to the odds ratios for school sector when A-level scores are the outcome variable. Students from the independent sector were more likely to achieve the highest grades at A-level, but with regard to outcomes at the end of university the advantage appears to lie with students who had been educated in the state sector. Finally, and perhaps unsurprisingly, students who entered the university with grades ABB or higher were 3.7 times more likely to graduate with a 'good' degree (when other characteristics are taken into account) than those with lower entry grades. Including interaction terms is useful in helping to see whether the impact of school type, for example, is moderated by students' A-level scores. The coefficients for the interaction variables, as presented in Model 4, were very close to one and suggest little difference in the odds ratios between the two groups.
Finally, the pseudo-R 2 value provides an indication of the amount of variance in degree outcome that can be accounted for by this model. At around 17% for Model 3, this number is relatively small and this does add further weight to the argument that background characteristics play a reduced role in mediating the effects of academic attainment at university, especially when compared with analysis of similar variables undertaken at the level of compulsory education (for example, Smith 2003).

Discussion
The title of this paper references an important work in the sociology of education: Basil Bernstein's (1970) renowned piece 'Education Cannot Compensate for Society'. In his paper, Bernstein draws attention to the complex social, cultural and economic factors that influence education outcomes and question how they might be overcome by the school system. As Gorard, in a paper echoing Bernstein's title, also argues: 'to a very large extent then, schools simply reflect the local population in their intakes, while being relatively ineffective in addressing the stratification of attainment that results ' (2010, 59). In the current paper we continue this consideration of the relationship between an individual's background and their academic attainment by focusing on achievement in higher education. The premise for the paper is as follows. It is well established that academic success is socially stratified and that educational inequalities persist across the life course. However, once students have met threshold academic levels and gained admittance to higher education, the effect of social characteristics on academic attainment might arguably be limited, as students will now share a common educational experience. With that in mind, this paper has examined the role that background characteristics play in determining eventual academic success for students who enter higher education with roughly similar entry grades. The data used in the analysis were drawn from individual student administrative records that were collected by one British university. They span the period 1998-2009 and correspond to a time of significant change in UK higher education: from policies aimed at widening access, to the introduction of tuition fees. The scale of the data has enabled participation in terms of the characteristics of applicants to be tracked over time and the determinants of academic success to be modelled. Recent research suggests that two things have happened to university participation in the United Kingdom over the period examined in this paper. First, the numbers of people entering higher education from under-represented groups have increased and applicants who have the required grades have usually been able to secure a place (Gorard et al. 2007). However, on the other hand, recent research suggests that many students from these previously under-represented groups are less likely to study the most prestigious subjects at the elite universities (for example, Furlong and Cartmel 2009;Guardian 2011;Mangan et al. 2010). This horizontal stratification of higher education can be seen in the data for the institution considered here. Over the decade-long period of analysis, the characteristics of students who enter this university have varied only slightly and intake is still largely skewed in favour of traditional entrants: namely those from professional occupational backgrounds, the privately educated and those of traditional age. However, the extent to which these same groups of students are successful at the end of their undergraduate careers is less certain. This is particularly noticeable in a weakening of the relationship between school sector and academic outcomes from A-level to university. We note, in common with Ogg, Zimdars, and Heath (2009), that being privately educated does not appear to confer an advantage, in terms of degree success, when compared with those who were state educated.
Indeed, when looking at prior attainment, the influence of background variables as determinants of degree success appear to be less strong in higher education, as exemplified by this institution, than they do during the compulsory phase. While it is certainly the case that pre-university attainment is not socially neutral (e.g. students from the independent sector arrived at the university with the highest A-level grades), when prior attainment is taken into account there appears to be a weaker relationship between student background and likelihood of graduating from this institution with a, upper-second-class) or first-class degree. In addition, not only does the relationship between academic attainment and school sector lessen at university, the same is true for the outcomes of students from professional occupational backgrounds. However, the models produced in this analysis are relatively weak in terms of their predictive power, suggesting that most of the variation in degree outcome might be attributed to non-standard variables of socio-economic status (i.e. variables other than parental occupational class) as well as variables that are not related to academic attainment (such as those related to engagement and motivation).
In short, this paper has contributed to the literature on inequalities and participation in post-compulsory education by suggesting that while social, economic and cultural factors are strongly associated with academic success during the compulsory phase, their influence might be less strong for students at this university. While the results can only relate to a single institution, it is reasonable to assume that they would also be relevant to similar types of 'elite' UK institutions. A number of key findings emerge, first, that the relationship between occupational background and degree outcome appears to be relatively weak. Secondly it provides no evidence that students from independent schools outperform those from the state sector, even when prior attainment is accounted for. And finally the model presented here accounts for relatively little of the variance in degree outcome at this university, suggesting that birth, geographical and academic characteristics may play a smaller role in determining degree outcomes than during the earlier stages of education. Indeed, the results of this analysis suggest that once students have overcome barriers to gain admission to this institution, then it is their entry grades and not their social characteristics that appear to most strongly influence eventual degree success; although even prior attainment is not a strong predictor of future academic success at this university.
Two possible implications from this research come to mind. First, the variables researchers use to measure inequality during the earlier phases of education may be inadequate for explaining patterns in higher education and alternative measures might need to be sought. As the findings presented here show, standard measures of inequality, such as age and occupational background, appear to be of limited use when it comes to explaining variation in degree outcomes at this institution. The availability of such data through institutional and national administrative datasets means that this is a fruitful area for analysis. However, there are limitations in terms of the breadth and scope of the data that are available and further large-scale research that gathers data on other potential determinants of academic success in higher education would be worthwhile.
Second, to return to the title of the paper and the extent to which higher education might be able to 'compensate' for inequalities that persist during compulsory education. If standard measures of social inequality are useful in explaining academic outcomes at university, then we might argue that Bernstein's contention is incorrect and that higher education may be able to compensate for some of the educational inequalities that persist at the earlier phases, notably attendance at a fee-paying school and family occupational background. However, that the characteristics of entrants to this institution have hardly varied over the period considered is, of course, a concern. So what does this mean for university admissions? One response might be to call for a programme of affirmative action that removes pre-university qualifications as the discriminating factor in determining access to university. This would then open up university access to a wider range of applicants. As Walford (2004) argues, the possession of qualifications are too often seen as an end in themselves or as a screening process to 'thin' the applicant field whereby potentially very able students might be overlooked. In this way, the democratic nature of the universities as espoused by R.H. Tawney far back in 1914 would still be relevant today: There ought to be a system of higher education which aims at … universal provision, which is accessible to all who care to use it, and which is maintained not in order to enable intellect to climb from one position to another, but to enable all to develop the faculties which, because they are the attributes of man, are not the attributes of any particular class or profession of man. (Tawney 1964, 73) Notes 1. The term 'elite universities' is used as shorthand for research-intensive universities who generally (although not exclusively) require the highest entry grades. They tend to comprise those institutions that attained university status prior to the 1992 Higher and Further Education Act. The Russell group comprises 24 of these institutions. 2. For the purpose of this paper, a 'good' degree is defined as an upper-second classification or higher.