OXFORD has often been criticised publicly for its pattern of undergraduate admissions. Primary attention has focussed on the published figures which show that over 40 per cent of our places are given to candidates from independent schools. There is usually the implication that the collegiate university is 'biased' in its treatment of applicants from different types of school. Overall figures such as these can be highly misleading, since they may largely reflect patterns of application rather than any 'bias' (intentional or unintentional) in how selectors treat candidates. However, the more detailed information published by the admissions office (www.admissions.ox.ac.uk/adstats.pdf) shows that, among those who actually applied in 2004, applicants from independent schools did have a somewhat better chance of acceptance (a 31.5 per cent rate) than candidates from state schools (26.0 per cent). They also show that some ethnic minority applicants such as Indians (20.8 per cent) and Black Caribbeans (23.1 per cent) have substantially worse chances than white English applicants (29.3 per cent). So the accusation that the collegiate university is 'biased' cannot be refuted from the publicly available data.
The Educational Policy and Standard Committee's Learning and Teaching Strategy states unequivocally that: 'The University seeks to admit the best applicants irrespective of their social, regional, ethnic, or educational background, solely on the basis of academic ability and potential' (EPSC, Revised Learning and Teaching Strategy, 2002-2005, p.7). So there is clearly a challenge for the collegiate university to show that academic ability and potential can account for the varied acceptance rates of candidates from different backgrounds.
Accordingly, one of the major alms of an Admissions Study is to provide a more detailed investigation of admissions to see if the varied success rates of applicants from different social and educational backgrounds can be explained by measures of academic ability and potential of the sort alluded to by EPSC. The meaning and measurement of ability and potential are notoriously slippery, but our expectation was that performance in public exams such as GCSE, AS and A2 would be reasonably good proxies for ability and potential, and would therefore largely account for differences in acceptance rates. We expected that any remaining differentials, after taking account of prior examination success, would be substantively small. Of course, there could be no guarantee that comforting results of this sort would actually emerge. But our hope was that, if uncomfortable results did emerge, the collegiate university would address the issue.
As well as studying admissions, we also intend to follow up successful candidates once they arrive at Oxford and to see if we can shed any light on the sex inequalities in Finals results. The key question here is whether women's lower chances of obtaining firsts reflect features of Oxford's own teaching and examining procedures or are largely to be explained by prior differences in the character of male and female applicants. For example, women students at Oxford might, on average, be less ambitious or more risk averse than their male peers, and these traits might be predictive of success in Finals. Alternatively, it might be that men and women students are similar in relevant respects when they enter the University, in which case it might be worth considering whether aspects of Oxford's teaching and examining procedures might disadvantage women students. At any rate, our study also includes instruments designed to test theories about differential performance in Finals.
The ideal strategy for tackling these research questions would have been to hire an authoritative and independent outside body to carry out the research. An internal study runs the risk that, if it obtains 'comfortable' results for the University, it might be accused of lacking independence and its results might then be questioned. However, the costs of an independent investigation would have been prohibitive. We also had some reason to suspect that an outside body would fail to secure acceptable response rates.
The research that we have undertaken is funded by the HEFCE-sponsored Widening Participation Initiative through the Admissions' Office. The project was carried out with the support of the Admissions Executive and was supervised by a panel consisting of Jane Mellanby, Peter Clifford, Geoffrey Walford, Liz Frazer and Janet Howarth. I am very grateful for the support of the Admissions Executive and of the Advisory Panel, but as principal investigator I must rake full responsibility for the design and conduct of the research, the analysis of the results and their interpretation. I would also like to record my appreciation of the work of my research officers, Susanne Choi and Anna Zlmdars.
We decided on the following research design:
There were many initial concerns from Colleges that this research design would be unduly onerous on Colleges and on candidates. We therefore carried out a pilot study in four Colleges in 2000. This showed the feasibility of our research design, and did not lead to any complaints from candidates or their schools. The Admissions Executive therefore gave us their backing in proceeding with the research, which we carried out in December 2002. We obtained ethical approval for this research design, and we obtained the written agreement of all participants. We emphasized that the research was unrelated to admissions decisions; that all data would be kept confidential; that all results would protect the anonymity of candidates, schools and Colleges. When we have finished the analysis, we will make the data available so that they can be re-analysed by other researchers, but we shall ensure that the publicly available data do not permit the identification of individual candidates, schools or Colleges.
Some comments on the research design may be in order. Most importantly, this research, like much of social science, is non-experimental and therefore strong causal conclusions cannot be drawn. In principle it might have been possible to conduct a field experiment in which actors, with matched academic records but different ethnicities or class origins, applied for places; field experiments of this sort have been carried out regularly in the field of ethnic disadvantage in the labour market. However, such experiments do involve deception; nor can they be double-blind and so their results too can be challenged. I doubt if field experiments would have been acceptable to the Admissions Executive. In the absence of field experiments, we cannot obtain 'proof' that candidates from different backgrounds are treated differentially. It will always be possible for a critic of our findings to say, quite reasonably, that some unmeasured aspect of potential such as candidates' performance in interview accounts for any difference in success rate. However, in reply we would argue that non-experimental work, of the sort we carry out, can lead to important findings which it is sometimes unwise to ignore. The most famous example is of course Snow's study of the relationship between outbreaks of cholera and the sources of drinking water in London (J. Snow, On the Mode of Communication of Cholera in London, 1849).
We secured a good level of co-operation from candidates and obtained a sample of 1929 respondents, 54 per cent of all candidates attending for interview. Candidates who were not called for interview, including those from overseas, are excluded from the research. We are now following up the successful candidates and obtaining their results in the First Public Examination.
Our main analysis so far has concentrated on the 'pre-A level' candidates, who make up the largest single group of applicants. We will next look at the post-A level candidates and at candidates taking other school examinations such as the International Baccalaureat, in order to see if the patterns found among pre-A level candidates are replicated among these other groups.
Turning to our findings for the pre-A level candidates, we begin by distinguishing between the 'gross' differentials in acceptance rates for candidates from different backgrounds and the 'net' differentials after taking account statistically of various measures of candidates' ability and potential.
The gross differentials that we found in our study tally closely with those reported by the Admissions Office for the full set of applicants in 2002. Thus we found in our sample that 37.1 per cent of pre-A level applicants from independent schools obtained an offer of a place compared with 35.2 per cent of applicants from state schools; 37.7 per cent of white applicants secured an offer compared with 19.2 per cent of South Asian applicants. (There were too few applicants in our sample from other ethnic backgrounds for sensible analysis.) We were also able to look at the gross social class differences, which are not available in the published figures from the Admissions Office but which feature prominently in public debate. We found that 37.6 per cent of applicants from professional and managerial backgrounds obtained an offer compared with 24.7 per cent from working class backgrounds. (Since we exclude 'desummoned' candidates, the acceptance rates in our sample are higher than those shown in Admissions Office statistics, but the pattern of differentials is similar.)
There are a number of reasons why working-class and South Asian candidates may have poorer chances of gaining an offer. First of all, they may apply to more competitive subjects such as medicine, Law or PPE. Second, they may have poorer grades at GCSE or AS level or poorer predicted A2 grades. The information on grades and predictions is available to selectors, and would normally be regarded as reasonable and defensible criteria to be used for estimating the ability and potential of candidates. We therefore carried out a series of multivariate (logistic regression) analyses in which we controlled statistically for subject choice and examination grades. The key results were as follows:
Now it could, of course, be argued that GCSEs and so on, important though they are, are not the only measures of ability that selectors use. Selectors have available written work, references, performance in the interview, and in many subjects standardized tests. It is perfectly possible that this additional information accounts for the 'net' differences that remain. However, a critic of the University is unlikely to be satisfied by this and could retort that other rival hypotheses could also account for the net differences and that there are no grounds for privileging the selectors' accounts rather than the alternative hypotheses (of which discrimination is one, but by no means the only, possibility).
In principle we could make some progress at measuring these additional factors; for example we could get independent graders to assess the strength of references or personal statements according to some agreed set of criteria. We could also obtain selectors' own gradings of interview performance, although the critic could well counter that these were more likely to be justifications of decisions rather than independent and valid measures of potential.
Given the difficulty of obtaining valid and reliable measures, we have not attempted to incorporate references, personal statements or interview performance. However, in the case of medicine and maths we have included the standardized and centrally-marked tests that candidates take. Including these as additional controls enables us to explain much of the net disadvantages experienced by South Asian and female candidates in these two subjects. But performances in the medicine and maths tests do not explain the independent school disadvantage. (We should note that the number of applicants involved becomes rather small when we disaggregate our analyses to examine individual subjects and these conclusions should therefore be treated with considerable caution.)
We are left therefore with some unexplained disadvantages for candidates from independent schools and (in subjects other than maths and medicine) for South Asian and women candidates. It may well be that unmeasured aspects of ability and potential account for these unexplained differentials: perhaps candidates from independent schools have had more intensive teaching at school, thus giving them unduly flattering GCSE grades. Perhaps women's good GCSE grades reflect their more conscientious approach to coursework, whereas men may show greater potential for performing well at unseen exams. Perhaps South Asian candidates come across in interviews (or in their personal statements) as having an instrumental approach, whereas candidates with the highest potential may exhibit a more intrinsic love of their subject. There is some (weak) circumstantial evidence for all of these suppositions, but alternative explanations that rely more on prejudice or stereotyping on the part of the selectors cannot be ruled out either. We cannot adjudicate between these rival explanations at present.
One important planned aspect of our research, which may give some leverage on these questions (although not definitive answers), is to follow up the candidates' examination performance at Oxford. If potential is the key selection criterion, then examination success as an undergraduate becomes a valuable yardstick for assessing admissions decisions. For example, if appropriate discounts have been made for the intensive teaching at independent schools, then there should be no differences in the examination performance of candidates from state and independent schools once they arrive in Oxford. But if candidates from independent schools do better in Finals, then this would suggest that selectors may have discounted too much. Of course, there could always be additional influences on performance in Finals, but at least it would be another piece of evidence that could assist our interpretation of the results.
The Admissions Study is not, therefore, going to give us definitive answers to our questions about the influence of social factors on admissions. But it has, at the very least, called into question the popular belief that Oxford selectors are biased in favour of candidates from independent schools and against those from working-class backgrounds, and it has demonstrated that, in keeping with the University's stated policies, admissions decisions are very largely meritocratic. On the other hand, speaking personally, I find the South Asian net disadvantage rather troubling. Under the Race Relations (Amendment) Act 2000 the University has a statutory duty to promote good race relations. I think that, at the very least, the collegiate University needs to undertake a more in-depth study of ethnic minority applications than has been possible in the present research.
Technical details of the methodology are available in the full, published report (www/sociology.ox.ac.uk/swp.html) and a more detailed paper is available on request from Anna Zimdars at the Department of Sociology. [Duff link - A.M.]