Earlier today I wrote a post disputing some of the claims made in the Guardian by David Lammy about bias in the Oxbridge admissions system, particularly ethnic bias. A number of people, including David himself, commented on the post or contacted me to contribute to this debate and I promised to follow it up when I could.
After spending some time this evening looking at the data that's currently available on Oxbridge admissions I've realised that it's not possible to draw any definite conclusion as to whether there's an admissions bias at Oxbridge on the basis of ethnicity. The data on the Guardian website, on David Lammy's own site and even the analysis that Oxford's admissions office kindly sent me earlier today simply don't support a conclusion either way. (That's not to say the Oxford analysis is wrong - merely that I don't have the raw data that I believe went into it, without which I can only say that it looks plausible enough. The admissions press office has however been very helpful and we are still discussing how that data might be made available - if they are, I will be able to give you some firmer conclusions as to what's going on.)
"The statistic simply shows that fewer than average (BME candidates) secured access to a university that is known to have tougher academic standards than other Universities.
Whether that is because of bias by the decision makers, or because secondary education is failing to educate a section of society to the required standard is not proven by the statistics.
All the statistics do is raise the question - they do not provide an answer."
Indeed. In his response to my article, David writes,
"Why is it that 25 of 84 Black applicants received offers from Keble College but just 5 of 64 Black applicants received offers from Jesus College over the same 11 year period? Why is it that at Robinson College Cambridge, Black applicants have only had a quarter of the success rate of their white applicants over 7 years when at St. Catharine’s Black applicants are marginally more successful?"
These are all good questions, but the answer is that, given the data we have, we simply don't know. David speculates "it is impossible with the data that has been collected to truly isolate race factors from economics", but economics isn't necessarily the issue here. The data available are cut by college, by subject, by ethnic origin, by LEA and by school type but without cross-referencing those categories (to see, for example, applications by college by ethnic origin) it's not possible to say whether there's a bias. For example, the analysis I've seen from Oxford points out that,
"BME students apply disproportionately for the most competitive subjects, contributing to a lower than average success rate for the group as a whole:
Oxford’s three most competitive courses (E&M, Medicine and Maths) account for 43% of all BME applicants – compared to just 17% of all white applicants."
It may be that so many more black students than white apply for courses that are oversubscribed that fewer of them get in because of it, and that discrepancy may account for the variation in total and college-level acceptances observed. Oxbridge colleges vary significantly in size and in the range of subjects they offer so it may be that variations in acceptance on a college by college basis are because of those differences. The problems with taking an aggregate data set, disaggregating it, cutting it in multiple ways and then using the results to make conclusions about the set as a whole are numerous and one of those problems is that we cannot draw the conclusions at which some commentators have apparently arrived by doing so.
Of course, it's possible that Oxbridge, or some set of Oxbridge colleges and/or tutors, are systematically racist. We can't discount that possibility using the data either, only say that the data do not support that conclusion. What I can say from personally meeting dozens of Oxbridge tutors in my three years studying as an undergraduate and from subsequently knowing some as friends is that this is neither my personal experience nor something that strikes me as a plausible conclusion. Oxbridge tutors care about all sorts of things, including the intellectual capabilities of the students they spend much of their time teaching, the performance and reputation of their college, rowing, being left in peace to get on with their own research and political advancement within the university or their field...but I've yet to meet one who cared to embroil herself in a conspiracy to restrict entry to a specific ethnic, social or geographic group. There are presumably genuine racists in almost any population but a population of intellectually brilliant academics seems one of the least likely places to find them in numbers, given the priorities and motivations of the group.
It is not a terribly satisfying conclusion, that numerous large datasets which obviously took time and effort for researchers at two universities to assemble do not let us answer these questions. But when that is the answer, it is important we acknowledge it. Admission to Oxbridge of BME British students is lower than admission of white British students. Admission of Black Caribbean British students is especially low. These are legitimate concerns, and ones that deserve to be understood. However, it remains that we do not understand them yet, and pointing the finger at the admissions policy as if it is at fault is unhelpful and does not move us closer to either an answer or a solution, when in fact we do not know whether the problem is a biased admissions policy on the part of the universities or the individual colleges; sub-optimal application strategies on the part of specific groups, particularly BME students; failure of one sort or another by particular schools or categories of school to prepare their pupils for the rigours of an Oxbridge degree; or, and this is still quite possibly the explanation, an ultimately meaningless statistical coincidence.
Update: a press release from Oxford reiterates some of the details I published last night, particularly their observation that the reason fewer than average black students are accepted is a far greater than average propensity to apply for the most popular courses. Again, without looking at the raw data I cannot say whether this observation alone is sufficient to entirely explain the phenomenon but I am happy to take it on good faith and it is a perfectly reasonable explanation. Here is an actual statistician (thoughtful observers will of course have noticed that I am no such thing), Dr John Bithell, Emeritus Fellow in Statistics at St Peter's College, Oxford, commenting on the college-level breakdown:
"At the level of individual Oxford colleges, numbers of black students getting a place are going to be very much subject to the vagaries of chance.
If there were 50 or 60 students of black origin at Oxford, they would not be distributed with one or two at every college. Just by chance, some colleges could have 4 or 5 black students, while others may have no acceptances in a particular year.
For the numbers to be big enough to be "statistically significant" - meaning that the data would seem reasonably unlikely to have occurred purely by chance in that way – we need to look at acceptance rates at the University level across all colleges."
Which perhaps goes some way to explaining why the university was apparently so reluctant to hand over the college-level data to David Lammy; in the form requested, or at least in the form published on his website, it cannot be used to generate a meaningful statistical analysis of the reason for the relative success of different groups going through the Oxbridge application process, only to try and score unsubstantiated political points.
Update: I recommend all readers take a look at the Royal Stats Society publication Significance, where Michael Wallace has analysed the college-level variation between black and white admissions and found it to be well within the expected range, the most interesting and important finding so far in this debate.
(Picture from misslectures on wikimedia commons)