Fairness and undergraduate admission: a qualitative exploration of admissions choices at the University of Oxford

Abstract

The article investigates unequal admissions patterns at the University of Oxford. Statistical work shows differences in admission rates by social class, ethnicity, gender, qualification status and secondary schooling. In‐depth interviews with admissions tutors, college and university officials and observations of eight admissions meetings provide insights into the processes behind those admissions patterns. Results illustrate that adjustment of applicants’ performance in the light of their schooling can account for the statistical observation of a lower chance of admission for private school applicants compared with their state school counterparts. The advantage in admissions chances enjoyed by post‐qualification applicants appears to be related to the lower risk they pose with regards to their attainment. Little direct evidence is found, however, to explain inequalities in admission rates by social class, ethnicity and gender. A tentative suggestion is made that homophilic tendencies among selectors might account for these unequal admissions rates. It is recommended that future research should use different methods, such as experiments or detailed observations, to gain a better viewpoint on these proposed mechanisms. The article concludes by considering implications for policy and the social stratification literature.

Introduction

The University of Oxford is the oldest university in the English speaking world. Graduates from the Universities of Oxford and Cambridge (collectively known as Oxbridge) continue to enjoy advantageous labour market outcomes. Oxford graduates in particular continue to secure leading positions in the public sector—including notably prime ministerships—the judiciary and the media (Boyd, 1973; Soares, 1999, p. 5; Oxford University Careers Service, 2006). In the national consciousness, Oxford represents the ideal of English education and the ‘cultural fact of its superior image is a closed one’ (Halsey, 1997, p. 577).

In addition to the kudos of an Oxbridge education, the fiercely competitive and distinct admissions processes of the two ancient universities mean that public, political media (Ryle et al., 2000) and policy (Schwartz Commission Report, 2004) interest in the profile of admitted undergraduates goes far beyond the level of interest bestowed on admission to any other UK universities. Enrolment at Oxbridge is scrutinised with regards to school type (private or state funded) and social class (The Sutton Trust, 2005). Those interests are related to concerns that the school education system does not level the playing field for students to compete on an equal footing for the most desirable forms of higher education, regardless of social origin (Marshall & Swift, 1993; Jackson et al., 2007; Scherer et al., 2007). For example, while fewer than 10% of all British secondary school children participate in private education, about half of the new undergraduates at Oxbridge have enjoyed this privileged kind of schooling. There is concern that the admissions process at Oxbridge further and unjustly increases the cumulative advantages children from high social class origin and privately educated students have accumulated earlier in life.

The question of admission to selective and prestigious universities also provides ample opportunity for scholars in the sociology of education to undertake empirical sociological inquiry pertinent to questions of social equality. We want to know whether and if so, to what extent, education mediates links between social origin, educational attainment and later life outcomes (Halsey et al., 1980; Goldthorpe, 1996). This inquiry can be conducted by techniques like statistical modelling. Such models can show the relative impact of individual student characteristics such as parental education, social class, gender and ethnicity, cultural capital and parental interaction on educational outcomes such as elite university admission (Mare, 1995; Shavit et al., 2007). Qualitative studies offer rich descriptions of the experience of higher education applicants (Gambetta, 1987; Archer et al., 2003; Reay et al., 2005; Reay, 2006).

It is insufficient, however, to focus merely on applicants’ aspirations and social background characteristics. Admission is a multi‐player game where individual chances of success are dependent on the decisions of powerful institutional gatekeepers. The role of educational institutions and the objectives they pursue in shaping selective admissions outcomes has thus enjoyed welcome attention in recent years, notably in the USA (Karabel, 2005; Douglass, 2007; Grotsky, 2007). Nonetheless, empirical investigations of the role of individual selectors and their value framework and preference are still in their infancy—except for two notable US studies (Lane, 2002; Stevens, 2007). But to date, there is no British study that I am aware of that has linked statistical patterns of university entry to the way individual admissions tutors ¹ describe the admissions process within a mixed methods research design. This is a shame as methodological individualism is currently the dominant epistemological paradigm at least in large parts of the quantitative research community. In other words, admissions patterns to highly selective universities have to be understood as having been generated through the individual action at the micro‐level (Weber, 1968; Coleman, 1990; Abell, 2003). This article then takes some first steps towards developing such a mixed methods framework by exploring the generative mechanisms of statistically observed admissions patterns through the accounts of individual admissions tutors, and with observations of admissions meetings. Limitations to this research design and improvements for future research are discussed in the concluding section of the article.

First, however, the article describes the University of Oxford in the context of contemporary British higher education. This discussion is followed by a description of the Oxford selection process for undergraduate admission before presenting results from interviews and observations and the concluding discussion.

Admission to Oxford in context

The higher education system in Britain is stratified with a group of large research‐intensive universities—the Russell Group—at the top of the prestige hierarchy. Within the Russell Group, the two Oxbridge Universities are outliers on measures such as wealth, age, staff to undergraduate ratio and ranking in university league tables. Further distinctive features of the two universities are teaching in very small groups or one‐on‐one tutorials and the collegiate systems.

Research supports the view that status hierarchies of higher education institutions partly overlap with class, schooling and ethnic cleavages in society. Specifically, research has shown that the higher social classes, private school attendees and white students are disproportionately represented among the Russell Group universities (Boliver, 2005) and among postgraduate students (Sastry, 2004; Wakeling, 2005; Zimdars, 2007). Such findings have not always enjoyed scrutiny by policy makers—in the case of ethnicity, this may be due to the disproportionate enrolment of ethnic minorities in higher education overall. Gender is perhaps even less obviously a cause for monitoring and concern as women are also over‐represented in higher education at large and they are well represented within the Russell Group. However, with regards to elite admission, the intersectionality between different cleavages is of interest. Specifically, there is an argument that as the most prestigious private schools continue to be largely single‐sex male schools, private education also contributes to the reproduction of male advantage in the highest echelons of educational institutions (Walford, 2005).

The Oxbridge admissions process

In addition to the advantageous position that the Oxbridge Universities enjoy within British higher education, the two ancient universities also operate an exceptional admissions process that involves face‐to‐face interviews with members of the academic faculty. But like all other higher education applications in the UK, Oxbridge applications are first filed through the central University and Colleges Admissions Service (UCAS). For school‐leavers, higher education applications usually occur in the year preceding their final secondary school examinations, during the A2 levels. The highest grade in the A2 examinations at the time of the research was grade A. Enrolment in A2 courses is subject to satisfactory performance in the GCSE (General Certificate of Education) examinations taken at the end of compulsory schooling, at age 16—the highest grade here is an A* (A star). Most Oxbridge applicants have achieved at least three AS at A2 level and a string of As and A*s in their GCSEs. It is also possible for applicants to file post‐qualification applications after having completed all their secondary school education. This is frequently the case with mature students but also happens among applicants to highly competitive courses or universities who were not successful in their first bid to join their preferred university. The analysis will show that, at least in the Oxford case, these applicants rightly believe they have a higher chance of admission the second time round as post‐qualification applicants.

The Universities of Oxford and Cambridge both operate a deadline of 15 October preceding enrolment, ² that is three months prior to the encouraged deadline of 15 January operated by other British universities. This structural feature of the admissions process may in itself give an advantage to pupils from organised (and disproportionately private) schools and parents with experience or knowledge of sending applicants to Oxbridge (Bourdieu, 1977). All application dossiers consist of a detailed attainment record, an academic reference and a personal statement. Information on ethnicity and parental social class is also collected on the UCAS form but this information is not available to admissions tutors when they decide on admissions. ³ In addition, Oxbridge frequently require evidence of written material and/or a written test. Depending on the intended subject of study, between one‐third and 95% of applicants from Britain and other European countries are then short‐listed to attend a selection interview at Oxbridge. During the annual interview weeks in early December, some subjects administer additional tests to help differentiate between applicants with equally stellar secondary school attainment. Applicants are usually interviewed at least twice in separate interviews, both usually conducted by two academics. These academics—referred to as admissions tutors in this article—then decide either within one of the 37 colleges/halls or as a subject group across the universities which applicants they wish to admit. In the case of pre‐qualification applicants, offers are conditional on satisfactory (usually AAA) performance in the A2 examinations. Post‐qualification applicants receive unconditional offers.

Many leading British Universities, including Oxford and its colleges, run access schemes designed to increase applications from ethnic minorities (e.g. National Black Boys Can, Oxford Access Scheme) and state school students (e.g. Sutton Trust Summer School, Target Schools). Applicants are encouraged to mention their participation in such outreach activities on their application forms. Some admissions tutors then refer to those applicants as ‘Access Candidates’.

Quantitative research results

Statistical research on a representative sample of 1,929 applicants for undergraduate study at Oxford during 2002 found unequal admissions rates to Oxford linked to social background characteristics (Kotecha, 2003; Heath & Zimdars, 2005; Zimdarset al., 2009). Unequal effects persist when controlling for subject choice (Bickel et al., 1975) and attainment related factors. ⁴ Ceteris paribus, female applicants’ predicted probability of gaining an offer was 10% lower than for males; there was a 9% gap in admissions probability by school type favouring state school applicants; a gap of 14% when comparing Asian with White applicants; and applicants of professional class parents were advantaged compared to those with managerial class parents by a 12 percentage point gap in admissions probabilities (the working class disadvantage was insignificant due to their low representation in the application pool (Zimdars et al., 2009). While 35% of pre‐qualification applicants gained a place, 46% of post‐qualification applicants were admitted. The analysis also found a negative effect for all applicants who applied without British school qualifications as is the case with most applicants from continental Europe, Asia and North America. In science subjects, proxies for being an Access Candidate also increase the chances of admission.

Net differences in admissions rates then favour white, male, state school educated, British post‐qualification applicants—a pattern that, with the exception of the private school effects, ⁵ largely fits a framework of effectively maintained inequality (Lucas, 2001). The qualitative work presented in this paper aims to make sense of these findings from the perspective of the admissions tutors.

Methods

The empirical research for this article consisted of 20 interviews with admissions tutors, three interviews with university and college officials and observations of eight admissions meetings. Research participants were interviewed during the academic year 2005 to 2006. The majority of interviews were conducted during November and December 2005, a time of year when issues of student selection were high on the agenda of anyone involved in undergraduate admission. Given the exceptional collegiate system and the selection process previously described, several hundred academic staff are involved in undergraduate admission. Interview participants were chosen using a maximum variation sample of Oxford academic staff to capture the breadth of admissions experiences and to identify themes that cut across different subjects and selection procedures (Patton, 2002; Robson, 2002; Ritchie & Lewis, 2003). Interview participants were recruited to maximise the variation in college membership, academic subjects, length of service at Oxford, age and academic rank. A conscious effort was made to include as many members of ethnic minorities as possible and to represent men and women equally in different disciplines. The actual profile of tutors at Oxford in terms of gender, age and ethnicity was established from official university statistics (Mitchell, 2005).

Of the 24 individuals initially asked to participate in the research, 19 consented. The five non‐respondents were either on sabbatical and out of the country (three cases) or were first time interviewers who declined because they did not yet feel comfortable talking about admissions (two cases). Hence, an additional four tutors were interviewed on a more ad hoc basis. The two first‐time admissions interviewers who were included in the study were interviewed before and after their first selection experience while other respondents were only interviewed once. Twenty of the research participants were involved in the selection process for a particular subject at a particular college. In addition, three respondents were selected as experts. Specifically, this group consisted of one person working for the central university administration; one college administrator; and one senior tutor in charge of the admissions process for her college. Interviews were semi‐structured and lasted between 45 minutes and four hours, with the majority of interviews lasting one hour. All but two interviews were recorded and transcribed verbatim.

In a second step of the investigation, selection meetings for undergraduate admission were observed. Several chairs of admission for specific subjects were asked for permission to attend post‐interview selection meetings and they all granted access. In total, two natural sciences subjects, one humanities subject and two social sciences subjects participated during the December admissions round of 2005. For most subject observations, more than one admissions meeting was observed—i.e., a meeting within a college and then a university‐wide admissions meeting for all academic staff teaching the subject. One short‐listing meeting for interview was also observed. Detailed field notes were taken in all observations. In addition, three observations were digitally recorded with the consent of the selectors.

To protect the confidentiality of the respondents and the participating academic subjects, the subsequent analyses only identify respondents and admissions meetings by their academic discipline—arts or sciences.

Results

The results section is organised thematically to reflect the unequal transition patterns that this study seeks to explain: schooling, qualification status, gender, class and ethnicity.

Schooling and access candidates

Secondary schooling was of seminal importance for the discourse admissions tutors employed when talking about their role of admitting undergraduate students to Oxford. All but one of the participating tutors stated that they paid attention to schooling information during the admission process. A non‐British born science tutor even commented that ‘I actually learnt quite a lot about the secondary schooling in the UK as a result of this [admissions] process. Public schools and comprehensive schools. The public schools are said to be good’ ⁶ (Science tutor 2 after his first admissions round).

Arts tutor 10 explained that he viewed considerations of schooling information as part of his job as selector to ‘compensate for the failures of civil society’. This position was echoed, if not without doubts, by several other tutors. Arts tutor 1 described how comparing applicants from different schools amounted to a counterfactual thought experiment of saying ‘had this person been at another school, they would have actually done better … . It is possible, I guess it is possible to say that. But, it is a difficult one.’ Science tutor 9 also wondered whether good students in bad schools were necessarily disadvantaged—‘maybe in a bad school, the teachers are actually delighted to have one bright student and the student gets a lot of special treatment and attention’ (see also Hall, 1999). Nonetheless, only one tutor openly opposed using schooling information in undergraduate admission on the ground that it was not the role of universities ‘to compensate for social inequalities’ (Arts tutor 6).

These views tied in with the observation of admissions meetings where schooling information was considered when selecting applicants for the interview stage of the admissions process; when reaching final admissions decisions concerning marginal applicants; and—in exceptional cases—when deciding the details of the conditional offer for pre‐qualification applicants. The way schooling information entered the selection process in these different stages is now discussed further.

When deciding which applicants to short‐list for interview, poor schooling could be considered when evaluating prior academic attainment of an applicant. This concerned in particular GCSE attainment, for which only three selectors extended concessions to include A2 level predictions below three As. An applicant with slightly poorer grades than other applicants might secure an interview when applying from a bad school or when labelled ‘Access candidate’.

Reflecting on a long career at Oxford, Science tutor 1 explained that:

Tutors think they are doing a balancing act. Well, you are doing things like balancing access … I think people do it and always have in Oxford, informally. I remember years ago—in the early 70s, the … Ancient Philosophy tutor … used to say ‘Oh, yes, but for somebody from that, comprehensive, or secondary modern, or whatever it was then—this is a very good performance!’ So, I think Oxford has always looked at that sort of thing. (Science tutor 1)

Arts tutor 1’s comment illustrates that schooling was also important for contextualising attainment in non‐science subjects:

If someone was applying … who met the entrance criteria and who had three A’s predicted [at A2 level], and who came from a school less than 60% GCSE passes we would automatically interview that person.

On the flipside, taking into account educational disadvantages translated into considering advantages when applicants had attended particularly good schools. Frequently, these schools were private schools. In short‐listing decisions, a good school negatively affected applicants’ chances of gaining an interview if they had not achieved outstandingly well. Science tutor 6 commented that ‘we even de‐summon [do not invite for interview] people who have “only” got an A at GCSE in [subject they want to study] rather than an A* if they come from a very good school.’

In the observation of admissions meetings, schooling considerations were not important in the discussion of applicants who had scored outstandingly highly on all admissions criteria—that is school attainment, interview performance and subject tests. In such cases, school information was merely used to reassure tutors that the right admissions decision had been taken:

to confirm the judgment. I mean, we had a few people who didn’t—they didn’t get 10 A*s at GCSEs which some people had, but they still had several A*s at GCSE, but they came from a school where they had significantly outperformed the school. (Arts tutor 1, Arts observation)

While schooling information was irrelevant for outstanding applicants, it could swing admissions decisions for marginal applicants, in particular if they were in a pool with other applicants who had also scored inconsistently across different selection tools. For example, an applicant might have performed well in the selection interview but had not been among the highest scorers in the selection tests. When confronted with such inconsistent performance, selectors gave these applicants positive consideration if they were labelled an ‘Access’ candidate or came from otherwise poor performing schools. However, admission tutors frequently rejected those who had failed to achieve well despite having attended a good school. Arts tutors 11 and 1 recalled such instances:

And there was one girl from a very, very good school who wasn’t doing very well. And, I think that came to bear upon the discussion. Because if that student had had every chance given to them and they are still not excelling—so why is that? (Arts tutor 11)

… I then ask myself—why hasn’t this person done really well? Isn’t it that they haven’t done very well because they are not hard working? Is it that they haven’t done very well because—I don’t know. You really have to then, your mind then starts working. Why haven’t they done very well so far if they actually have the potential? And, especially if they have gone to a good school. (Arts tutor 1)

In both cases the candidates in question were not admitted because their performance had not been outstanding relative to the school’s performance as a whole. This was despite the fact that their individual attainment was very respectable in absolute terms.

Once selectors had decided which applicants to admit, schooling information could also impact on the A2 level offer conditions. The standard offer for study at Oxford is attainment of three As at A2 level. However, in exceptional cases tutors can lower this offer for applicants who have experienced significant educational disadvantages, although such discretion has come under increasing pressure with standard AAA offers and the current introduction of an A+ grade at A2 level. In such cases, the triple A could be lowered to AAB: ‘We feel free to make a lower offer to someone who has potential but who got a B in a subject from e.g. a school in South London and we feel has not been taught well’ (Arts tutor 5). In one case, schooling information was also reported to be considered when an admitted student did not achieve their three A offer condition. Furthermore, in one Arts observation, the selectors decided, based on the ranking of their ‘sure admits’, which candidates they wanted so much that they would have them even if they missed their three A offer. This last example was not necessarily due to a consideration of disadvantage but the belief that three As alone were not necessarily the best predictor of who would be the best student.

The observations of admissions meetings and the accounts of individual selectors then support the statistical observation of a negative effect for having attended a private school when comparing applicants with the same attainment who had not been scored consistently high on all selection criteria. This process of discounting and compensatory sponsorship (Grotsky, 2007) flies in the face of media coverage and public perception of preferential treatment for private school applicants in the admissions process. Nonetheless, from a social stratification perspective, it is also noteworthy that schooling does not perfectly overlap with underlying dimensions of socio‐economic disadvantage. For example, allowances for poor schooling can also favour white middle classes in urban comprehensives rather than benefiting the poor and minorities who are the intended recipients of the school discounting. There was also some evidence during the observation of one admissions meeting that there is increasing awareness among middle‐class families of the private school discount. Some parents thus chose to enrol their children in state education for their last two years of schooling after 11 years of private schooling. This way the children counted as ‘state school’ despite their social background being indistinguishable from privileged private school students (Arts observation). One college administrator also pointed out that some of her college’s working‐class students had been admitted through the private school route since they had earned scholarships, emphasising the limits of using schooling as a proxy for socio‐economic position.

Qualification status (post‐qualification applicants and overseas)

Applicants’ chances of success also differed with regards to whether they had applied before or after completing their school leaving examinations and whether they applied with British or foreign qualifications. Post‐qualification applicants enjoyed a higher propensity for gaining an offer than pre‐qualification applicants; this was particularly marked for those with overseas qualifications. Post‐qualification applicants were described by tutors as being ‘like gold’ (Science tutor 3), ‘a bird in the hand, not a bird in the bush’ (Senior Tutor). In two meetings, post‐qualification status was decisive with regards to filling the last couple of places.

It aids the understanding of these effects to have an appreciation of the concept of risk in selection decisions. As one arts tutor stated: ‘I think it—all these [admissions] choices are risky in terms of whether they’ll be able to cope.’ Another arts tutor described the admissions exercise as ‘crystal ball gazing—you are looking into crystal balls whether these people are the best people’. Tutors preferred to select students whose application dossier indicated little risk, or at least the ‘right sort of risk’ and whose behaviour over the next three years of their undergraduate degree course was as predictable as possible. ⁷ Post‐qualification status scored high among tutors as signalling low risk. The notion of risk also seems to explain why the benefit of applying post‐qualification was particularly advantageous for overseas applicants. During several admissions meetings, it was observed that there was some confusion surrounding the meaning of foreign attainment records such as the Abitur or SAT scores, and the risk of admitting a foreign student was augmented when there was uncertainty about whether they would achieve highly in those examinations and the trustworthiness of school references was unknown. This preference for avoiding risk ties in with a wider body of literature on decision making and a preference for certainty over uncertainty (Kahneman & Tversky, 1979; Kahneman & Tversky, 2000).

Gender, ethnicity and class

Notions of risk reduction, however, were not straightforwardly connected with the lower admissions chance of ethnic minorities, women and those without two professional class parents. In fact, it was challenging, if not unsurprising given the strong social norms against discrimination (Pager & Quillian, 2005) to involve any of the participating tutors in discussions of class, ethnicity and gender—‘we take them as they come’ (Arts tutor 3). Considering social class was frequently understood as synonymous with considering schooling information in admissions and there was little awareness that aggregate statistics showing gender parity or even increased female and ethnic minority participation in higher education at large might still hide differences in the relative propensity of admission of particular groups (Bickel et al., 1975). Gender and ethnicity were then ‘not part of the monitoring’ (college administrator) individual colleges undertook as part of the admissions process.

The unequal transition patterns by social background then were the most challenging to investigate and it is in the exploration that the discussion enters the most slippery ground, and perhaps the most questionable leaps are suggested from the empirical data to social theory.

Applicants’ ethnic origin or their gender were generally not discussed explicitly during admissions meetings. Applicants were referred to by their surnames—and while some surnames sounded more white British than others, it was often not clear for me as an observer whether any particular applicant was male or female or to infer their ethnic origin. In one admissions meeting, a student was admitted in a straightforward manner and the chair then stated that he was delighted that this person ‘increased the diversity’ of the group of new undergraduates. It only emerged in an informal discussion with the same chair several months later that the student in question was black and that the selecting committee had been very keen to admit him as the only new student of black ethnicity in the college—but selectors had refrained from referring to his ethnic background in the presence of a researcher. It is difficult to square this incident with the observation of admissions statistics that clearly favour whites. For example, in 2003 the 141 black applicants to Oxford had a 15% chance of gaining an offer compared with an overall success rate of 30% among all applicants. The gap in success rates in favour of white over black applicants was 12.3% in 2004 and 15% in 2005 (University of Oxford, 2003, p. 8; 2004; 2005, p. 5).

In another admissions meeting, however, there was clear evidence of attempts to limit admissions prospects by geographic origin with a proposal that non‐British applicants should not be considered for marginal places. It is difficult to say whether this proposal was motivated by laziness (working out offers for overseas applicants can be more work) or xenophobia. It is even harder to link such a statement to racial prejudice as this policy would affect applicants from regions with different ethnic profiles such as mainland Europe, Africa, Asia and the Americas.

The interviews with selectors also failed to uncover any direct explanation for the different class, ethnic and gender effects. In one interview there was perhaps some suggestion of drawing inferences about students’ motivations or aspirations based on their ascribed characteristics. When asked why statistical data might show a negative female effect for admission, Arts tutor 7 replied:

I am aware of the fact that in [my subject] the gender performance is very unequal [in final examinations]. And, of course I understand enough about incentives to have a theory that that might be because female candidates don’t have such a strong incentive to bring in a high mark, because they don’t have the same ambitions to go on.

Further unprompted reflections, however, led this selector to conclude that ‘in terms of my own experience, the data—well, the cases that I am thinking of—don’t fit that pattern.’ The same selector also reflected that a hypothesised link between valuing education more as a ‘means to an end’ than ‘an end in itself’ (instrumentality) might vary by ethnicity with minority students perhaps displaying more instrumental attitudes: ‘there might be a pattern’. If substantiated, this could explain some of the linkage between ethnicity and admissions decisions as selectors also expressed a preference for admitting students wanting to study a subject for its own sake rather than for financial rewards. But overall there was no strong mechanism in the interviews or observations that linked ethnicity or gender to admissions decisions.

With regards to social class, respondents noted that they had no access to social class information at the point of making admissions decisions. If anything, respondents exhibited a preference for giving a chance to working‐class students. At the same time, several humanities tutors offered their opinion on why statistical research might find an advantage for professional class parents and why high cultural capital was positively associated with admissions decisions in humanities subjects (Zimdars et al., 2009):

I mean, ability is not something that develops in utter independence of everything else you know. I mean, what do you do? Yes, students who have come from educated backgrounds, who have been exposed more to in breadth literature or whatever have possibly a better chance of proving themselves more able for entry into a course where that’s what they’ll be doing. Not that this is necessarily a good indication of motivation, or indeed of ability but certainly of a certain kind of acculturation of training, yes for sure. And I think that’s in a sense what is very difficult to assess in interviews, is to see through the training. (Arts tutor 4)

The edge in the admissions process for applicants from educated, high culture backgrounds is then less puzzling. A further piece of evidence emerged from an interview with another humanities tutor who, unprompted, described his own background as ‘an educated family with a lot of books in the house’ (Arts tutor 3). In fact, the respondents went on to state that he thought this might be a source of ‘his own biases’ in the selection process.

It is conceivable that this idea of personal, often unconscious, biases may hold further clues as to differential admissions rates. It links the processes that occur during the admissions interview to a body of literature on who we relate to and why. Research suggests that it is social similarity (homophily) that breeds connections (McPherson et al., 2001, p. 415). At the micro‐level similarity‐preference has been found to operate along the lines of same ethnicity, same sex, but also class, and shared values (Lazarsfeld & Merton, 1954). In employment contexts, hiring decisions have been linked to reproductions of values or social demographics (Kanter, 1977; Lane, 2002). At the macro‐level, controversial research is currently emerging that links increases in the ethnic diversity of societies to declines in trust and civic engagement (Putnam, 2007). While homophily is a poor candidate for explaining changes and shifts in attitudes, the argument is sufficiently strong to be introduced as a major source of personal biases and discrimination in equal opportunities training workshop for employers (Brough, 2008). Equal opportunities legislation, and efforts to make hiring processes accountable and transparent, are policy measures designed to counter inclinations for unreflected homophily.

Several tutors acknowledged that processes occurred whereby they related or ‘clicked’ intellectually more with some than with other applicants. Science tutor 3 described this in some detail:

The problem that you have is when you start to engage with somebody, inevitably when you are dealing with human beings you are going to have rapport with some people, and you’re going to have less rapport with other people. Okay? And that is to me, that’s irrelevant for getting into Oxford. This is a place that should be taking people on the basis of intellect and merit—right? And whether I, I have a rapport with a person or not, I think is largely irrelevant. I mean, not entirely irrelevant because it’s a tutorial system and if we hated each other it’s going to be a pretty strained tutorial. But, you know, in general you are not going to find someone you take an instant dislike to. You are going to find people that—some people that you have more rapport with than others. And it is just an inevitable part of being a human being. Right? And what I discovered in the past is that I really have to try and force myself [when there is] somebody that I seem to like, not to be biased by the fact that we are engaging with each other…. That person shouldn’t get into Oxford University just because they happen to have a manner and a personality that I have cottoned on to.

As previously discussed, likeability is linked to homophily and thus may not be neutral with regards to demographic and other characteristics. If homophilic tendencies crept into the admissions process, this preference would work to the advantage of white and male applicants. As Table 1 shows, a preference for social similars along gender lines would advantage male applicants as four in five selectors are male. Tutors are disproportionately white—and, by virtue of being an academic, all staff are classified as professional class.

Table 1. Academic staff at the University of Oxford by gender and ethnicity

	All academics
Gender
Female	20.5
Male	79.5
Ethnicity
White British	72.8
Minority
Asian	2.1
Black	0.4
Other	2.5
Not known	21.6
Total	100
N	1399
Source: Personal Communication with Liz Mitchell, Staffing figures (Mitchell, 2005).

Thus, while the interviews and observations did not uncover any direct evidence for the class, gender and ethnic effects, it is possible that homophily might form part of the explanation for these effects. Homophily in turn is positively associated with the perceived degree of risk in a decision (Kanter, 1977). The perception of risk discussed in relation to the post‐qualification applicants might thus also increase the propensity for homophily.

Discussion

The Universities of Oxford and Cambridge continue to operate exceptional admissions processes that include a selection interview with prospective undergraduate students. Tutors enjoy room for discretion which means that admissions decisions are not formulaic. This article tried to understand how the admissions narrative of individual selectors and actual admissions meetings relate to statistical findings of admissions pattern. While prior attainment, in particular at GCSE level, bears the strongest relationship with admissions decisions, schooling, qualification status, gender, ethnicity and social class also come into play in the selection process.

The research had most success explaining the effects of schooling and qualification status. Conversely, for the effects of gender, ethnicity and class it was less successful. Admissions tutors described how they were careful in giving weight to information that contextualised applications within the attainment levels of their school and access initiatives. While this information was never used in clear cut ‘admit’ cases, schooling information and a discounting of the performance of private school applicants could occur for marginal applicants. This practice may mirror wider social debates on admission to selective higher education and support increased participation of state school educated students as advocated by the Office for Fair Access.

It also emerged that selectors perceived the admissions exercise as risky, an exercise of ‘crystal ball gazing’ (Arts tutor 4) in trying to foretell which applicants would be the best undergraduates. Signals that reduced the perceived risk of the decision—such as being a post‐qualification applicant—worked to the advantage of applicants. It was then proposed that the existence of homophily may also lead selectors to view those most like themselves as the least risky admissions choices. This would advantage male, white and professional class applicants. Furthermore, research in the USA has revealed that students who are like previous students might be perceived less risky and more admissible; while more unusual students have to be truly exceptional rather than merely very good to gain admission (Stevens,2007). Again, such a mechanism would advantage white, male applicants from high social class origins.

The study design employed in this project cannot exclude other possible explanations for the remaining ethnic, class and gender effects. Such explanations may include taste discrimination and bigotry. They may also include statistical discrimination whereby group membership is used as a proxy for differences in later performance that are known to be true on average (e.g. females get fewer firsts than males) but that are difficult to observe prospectively and may not be fair for the individual applicants concerned.

It would be desirable for the production of informative and valid insights into the Oxbridge selection process if future research could move beyond the observational and one‐on‐one interview method employed in this article. Specifically, it is questionable whether the social background neutral admissions behaviour reported by selectors is an accurate proxy for their actual admissions behaviour (Wicker, 1969). An application of the classic ‘matched applications’ experimental design could be fruitful (e.g. Neumark et al., 1996). This would involve having researchers posing as applicants for admission to Oxford matched on crucial factors such as attainment, subject choice and schools but differing in their ethnicity or social class or gender. Short of an experimental study design that may not gain ethical approval, it could also be insightful to gain access to observe admissions interviews. This would allow for a careful observation of the interaction between applicants and interviewers, especially of the ways embodied aspects of social‐background characteristics, such as physical appearance and demeanour, confidence and styles of speech, play out during the selection interview. Lastly, the homophily assumption would suggest that white selectors behave differently from non‐white selectors and this hypothesis could be empirically tested.

Finally, the article showed that the focus on differentials in secondary schooling may have come at the cost of neglecting other issues of higher education enrolment, namely ethnicity, class and gender. When these issues are not on the radar screen of selectors, it is possible—simply by assuming some preference for social similars in human interactions—that a holistic and personalised admissions system may inadvertently slow down the progress of these groups in gaining admission in the numbers appropriate to their prior educational attainment. Increasing awareness of factors other than schooling among selectors might suffice in yielding increased admissions rates of these groups. This point also illustrated the double‐edged implications of tutors’ individual discretion with regards to achieving social justice in admission: it can be a driver for change in admissions patterns as seen with the school effects as well as a motor for preserving privilege through inertia (class, ethnicity and gender patterns). The article suggests that conscious and goal oriented efforts would be needed to change these latter admissions patterns. It would also be desirable to have a broader social debate on participation in higher education in general and selective higher education in particular by the equality groups not covered in this article—namely disability, religion and sexuality (Equalities Review, 2007) as well as those with caring responsibilities.

Notes on contributors

Anna Zimdars is a Research Fellow in the Institute for Social Change, University of Manchester. Her research interests include access to higher education and professional occupations (in particular law), cultural capital, social stratification, civic engagement and mixed methods research.

Acknowledgements

This project greatly benefited from the helpful advice of Anthony Heath, Geoffrey Walford, Mitchell Stevens and Walter Müller and the suggestions of two anonymous reviewers. Funding through the Economic and Social Research Council (ESRC), HEFCE and the University of Oxford is gratefully acknowledged. All errors and omissions are the author’s own.

Notes

1. The term admissions tutor within Oxford is locally reserved for the one administrative officer with responsibility for undergraduate admission in a college. However, this generic term has been the most accessible terminology for international audiences to denote all academics involved in admissions decisions.

2. This deadline is also enforced for medical degrees across England and Wales.

3. In the interview‐based admissions system to Oxbridge, of course, ethnicity and to some extent social class can actually be observed by admissions tutors in the selection process.

4. A caveat to these findings is that not all parts of an applicant’s dossier were included in the modelling—missing information includes the personal statements, school references, written work and subject specific tests other than for Medicine and Mathematics. The qualitative work with admissions tutors suggested that the personal statements and references were not generally critical pieces of information in selection decisions; however, the exclusion of written work and subject tests creates some limitation to the validity of the findings.

5. It should be noted that the statistically significant private school effect only appears after controlling for attainment in GCSEs. In other words, on the surface, the raw admissions rates for private school applicants are statistically indistinguishable from the admissions rates for state school applicants.

6. The term public schools refers to the most prestigious private schools in the UK.

7. Unrelated to post‐qualification status, risk was occasionally discussed in admissions meetings but without clear implications for the statistically observed admissions patterns. For example, in one admissions meeting, a tutor commented that a particular candidate ‘might be a high risk but might get a high first [highest degree class at the end of an undergraduate degree]—the question is whether the applicant is worth the gamble.’ This student was eventually judged as the ‘wrong sort of risk’ and not admitted. A home‐schooled applicant with six As at A level also went without a place. There were indications that he had been taught by his two academic parents, so while the fairness of such an admissions decision is perhaps debatable, this incident would not have shown up in the statistical analyses.