What they teach you and what you learn ...
Via Jamie, an Oxford PPEist writes:
Let’s be clear: Oxbridge accepts, roughly, the same proportion of state school applicants as apply. The problem therefore is that not enough bright state school kids are applying, not that they are being discriminated against once the UCAS forms are in
Per the University of Oxford[1] website, we discover that "roughly" covers a multitude of sins. Taking the "UK domiciled applications" (note that the non-UK acceptances which make up 15% of the total are going to be socially a lot more similar to UK independent than UK state school applicants).
"Maintained sector" accounted for 6485 applications and 1456 acceptances - an acceptance rate of 22.5%.
"Independent" accounted for 4173 applications and 1244 acceptances, an acceptance rate of 29.8%
How important is this? Well, eyeballing it suggests that it's the difference between just over 1 in 5 and just under 1 in 3, which makes a practical difference. And of course the smaller number is applied to a bigger applicant base. The overall acceptance rate ex "others" was (1456+1244)/(6485+4173)=25.3%. If the acceptance rates were actually the same, then there would have been 187 more state and fewer private school acceptances. Or alternatively, 15% of independent school acceptances are the result of differential acceptance rates. The figures for Cambridge are a bit less user friendly to deal with, but broadly comparable - state schools 23.3% acceptance versus independent schools 32.5%; I calculated 225 excess independent school pupils at Cambridge in 2009.
To be honest, I don't think this can actually be fairly considered to be with in the bounds of the phrase "broadly the same", particularly if you're then going to base whole other steps of your argument on having completely ruled out admissions bias as a possibility. This is before you get into the obvious point that applications are going to be endogenous with respect to acceptance rates.
Is this what they teach you on PPE courses? Yep it is. Specifically a) you don't learn statistics, econometrics or many quantitative skills at all unless you specifically search them out (this is why the Bank of England, among others, didn't count PPE graduates as "economists"). And b) the way the course is set up absolutely teaches you to make broad brush statements, preferably in a loud and confident tone of voice, about things that you haven't checked up on but have sort of picked up are reasonably there or thereabouts, in the sense that you might be badly wrong but are probably not so wrong that it's gonna be awfully embarrassing if you're picked up.
When I say "the course teaches you" to do these things, I don't mean that PPE has a paper on applied bullshitting, or that you're instructed to do it, of course. I just mean that the workload involved means that there's a constant temptation to cut corners, and the assessment method (involving weekly tutorials with people who aren't always experts on the specific area involved and who are typically on friendly terms with you) is practically designed to not pick up this sort of corner-cutting. What you learn in the lessons is always a very small subset of what you learn on the course.
This is true of MBA-school too. Lots of the things for which MBAs are widely reviled are things that they learnt at business school, but not in the classrooms. I never did the MBA track, of course but I watched them and as far as I could see, the two year MBA was almost entirely an exercise in workload management. Pushing off responsibility onto other people and dropping them in it wasn't on the syllabus, but it was more or less an essential survival tactic. And lessons learned, of course, are carried on into future life.
[1] If anyone knows why not "Oxford University", they are politely requested by mgmnt to keep it to themselves.
ah that's interesting, I hadn't bothered to check the numbers. Although we still need more information to know what this says about bias.
ReplyDeleteI posted this link on the LC thread, and chances are you're already familiar with it, but just in case not you might be interested in Simpsons Paradox and how it applies to admissions bias data. One of my favourite examples of how to be mislead by statistics. I wonder if there are any patterns in how state versus private school kids apply to hard/easy to get into courses.
In my university there is endless hand-wringing about "widening participation" and just about no progress on it. This is in large part because the senior management are also obsessed with "reputation" and so just about any effective measures are met with massive media counterstrikes by the independent schools and the middle classes. The one thing that would sort all this is admission by a suitably designed lottery. This is, however, politically impossible.
ReplyDeleteThe numbers don't really prove anything either way in terms of discrimination. What you need to get is the acceptance rates for academically similar candidates from private and state schools.
ReplyDeleteIf it was the case that admissions committees looked at applications and said "this one's an oik, bin him", then you'd expect a higher percentage of successful applications from private schools. But if it was widely believed that they tended to do this, then most state school kids wouldn't bother applying. The only exceptions would be absolute geniuses who were confident that their brilliance would override this anti-oik bias; so in that case you might actually see a higher percentage of applications being successful from state schools.
It is also totally possible to transfer into PPE in the 2nd year (from E&M or Hist & Pol etc), only ever study two branches from Year 2 but graduate with the full degree. I.e. have a degree in economics from Oxford never having studied one paper in it.
ReplyDeleteI say this as someone who transferred from PPE to Hist and Pol (with high E scores) in Year 2, in part because other PPEists were darn painful to study with.
Also, maintained includes a disproportionate number of grammar schools, and the ethnicity statistics don't represent the massive underrepresentating of black kids.
And, of course, there's another possibility: maybe the top x% of both groups are all applying, but private schools just tend to produce much more qualified students, through a combination of higher investment per pupil and favourable family backgrounds, so they're more likely to get in.
ReplyDeleteI wonder if there are any patterns in how state versus private school kids apply to hard/easy to get into courses.
ReplyDeleteWell, there is Classics.
Thanks for this. I've been feeling like I've been going mad listening to the "same proportion" argument.
ReplyDelete"hence, when my mates at other Unis were doing three essays a term, I was doing two a week"
ReplyDeleteThis is (and was) not obviously a great thing.
Looking at those admission stats, we have 'independent' men - 31.5%, 'independent women' - 31.0%, 'maintained men' 27.3%, and 'maintained women' 20.7%.
Splitting it up increases the risk of what Luis warned about, I guess.
Luis - in answer to your point if I take the admissions data by school and course, and course-by-course take the state school pupils application numbers but multiply by the private school acceptance rate, and the private school numbers multiplied by the state school acceptance rate, and sum them, then the prior % acceptance of 23.5% state and 29.2% private changes to 24.8% to 28.4%, so some narrowing but it doesn't disappear. Is that the right way to do it?
ReplyDeleteAnd b) the way the course is set up absolutely teaches you to make broad brush statements, preferably in a loud and confident tone of voice, about things that you haven't checked up on but have sort of picked up are reasonably there or thereabouts, in the sense that you might be badly wrong but are probably not so wrong that it's gonna be awfully embarrassing if you're picked up
ReplyDeleteIt's not excessively cynical, I think, to point out that this is an invaluable life skill if you want to move into any kind of knowledge-based work that involves directly meeting your client face-to-face every so often.
Also, developing the ability to bluff your way through a small meeting while shit-faced has proved useful on a (very few) occasions.
Matthew,
ReplyDeleteI'm not sure - I remember being shown Simpson's Paradox but not what to do about it. Isn't the solution just to look at disaggregated data, if you suspect that aggregation is causing a problem? Perhaps that paper by Pearl linked to in the wiki article has suggestions. Sorry not to be much help.
The way it was always explained to me, was that the PPE essay system trained you (wannabe Sir Humphrey) to prepare a brief for your minister at incredibly short notice.
ReplyDeleteThere's no real solution to Simpson's Paradox except eyeballing the disaggregated data and looking for confounding factors and I don't think that's what's driving it here; from table 7 of http://www.ox.ac.uk/document.rm?id=1502 , I can only see a few lines where the success rate of "Maintained" is better than of "Independent". Some of these are things like "Classical Arch & Ancient History" which IIRC was designed specifically to attract Classics students from schools that didn't do Greek, or "European & Mid Eastern Languages" where I would bet quids that it's picking up the demographics of native Turkish and Arabic speakers.
ReplyDeleteIn terms of Simpson's Paradox, in general state school applicants are much more likely to apply to do sciences than arts (57/42 vs independentts 65/34), where the total success rate is higher. Within the arts, they're proportionately more likely to apply to do English and (massively) Law. These are slightly more difficult subjects to get in on (English has an overall 22.7% success rate, Law 21.0% vs overall Arts 24.8%), but I don't think it's likely to be enough to make the differential success rates in the data likely to be a freak, particularly given the arts/sciences effect the other way.
Re: toff degrees, I'll see your Classics and raise you Land Economy.
ReplyDeleteMm yes, I'm quite often distressed at the innumeracy that you find in the Oxbridge house mag, the London Review of Books. For example, David Runciman, who teaches politics at Cambridge, and who absolutely ought to know better, had this to say about the polling during the 2008 US election:
ReplyDelete"The recent New York Times poll that gave Obama a 12 per cent lead was based on interviews with just 283 people. For a country the size of the United States, this is the equivalent to stopping a few people at random in the street, or throwing darts at a board." (http://www.lrb.co.uk/v30/n11/david-runciman/the-cattle-prod-election).
Presumably if he's comfortable going public with drivel like that he's also comfortable with his students knowing sweet FA about stats. Here's the response of Andrew Gelman (who does, of course, know about polling):
"Hmmm . . . I wonder why they don't just throw darts at a board, then? This would save them lots of money! For n=283, sqrt(p(1-p)/n)=.03, so +/- 2 standard errors is +/- 6 percentage points in the vote share for one candidate, or +/- 12 percentage points in the vote differential between two candidates.
"Is this as good as a "dartboard" (i.e., an estimate based on prior information)? Maybe. It depends how good the dartboard is. The best solution is probably a weighted average of the "dartboard" and the poll, maybe weighting the dartboard more than the poll if the sample size is that small. But the size of the U.S. is not the relevant issue here. (The relevant formula is 1/n - 1/N.)" (http://www.stat.columbia.edu/~cook/movabletype/archives/2008/06/but_viewed_in_r.html).
The two degrees subjects which are a) large and b) have a huge gap between state and private acceptance in Oxford are physics and maths, make of that what you will.
ReplyDeleteOn the LRB, I think it deserves credit for from time-to-time actually having a few statistical reviews, but it does contain some shockers. Andrew O'Hagan surely should have wondered about this figure?
"The motor industry makes £200 billion a year, accounting for 8 per cent of GDP. It employs close to one million people, which explains why successive governments have spent so much public money propping it up"
http://www.lrb.co.uk/v31/n11/andrew-ohagan/a-car-of-ones-own
John Lancaster also managed to mess up the balance sheet example he gave in his (otherwise very good) series of articles, IIRC.
ReplyDeleteI don't know that Oxford actually "trains" anybody to do anything. Holden's remarks on Pencey Prep's prospectus would be relevant in this regard.
ReplyDelete