The Economist's British university rankings

WHICH university is best? Most published rankings tell you a great deal about an institution’s reputation, or how often its researchers publish in top academic journals. However, they don’t necessarily answer the question that prospective students really care about: what they have to gain by attending. For example, it is possible that graduates from the London School of Economics do well in the job market because of the university’s superior teaching. However, it is also possible that they earn so much simply because they were bright to begin with, or because they tend to choose courses that lead to lucrative careers.

To estimate which universities make the biggest positive impact on their students, we have built a statistical model that estimates how much graduates from each university can expect to earn. It is based on how selective a university is, what subjects its students study, the share of students from lower-income areas, how many students enter after age 20, what share of students attended private schools and where the university is located. We then compare these estimates of expected earnings with the actual figures. This difference, we argue, measures the value added by the institution.

The table below displays the results. We list both the actual and expected earnings of students from each institution, and rank them by the gap between the two figures. For the technically inclined, we offer a fuller explanation of our methodology here.

The Economist has published an interesting new league table which ranks universities by the 'value added' to your career 5 years after graduation by attending certain universities.

The Economist:


Here is the top 10:


You can also see how individual courses at universities far. The most lucrative course was E&M at Oxford, where expected earnings are £47,605 5 years after graduation, but actual earnings are £71,700, giving a value added of £24,095. Some of the unexpectedly poorly performing universities include St Andrews and Durham (and the methodology controls for location so that's not the reason).

Note: the actual earnings uses median not mean, as to not be skewed by the very high earners. That could be considered a weakness as it may under-emphasise the performance of the richest graduates from elite universities.

I cant see the point in the table? All its doing is showing a small variation between however they decided on the average perceived v average actual earnings.

Most people would still prefer to go to the uni that offered 100% greater actual earnings than a possible 5% increase of a low estimated earning.

According to this table, taking Computer Science at Oxford Brookes has a value added of +£6,134, whereas taking Computer Science at the University of Oxford has a value added of -£5,061. I don't know, perhaps there is some truth to that, but it makes it difficult for me to take this seriously as a measure of value. There are a lot of other very strange figures that come up. I can also imagine that this might fluctuate significantly year-on-year and suffer from sample-size effects, particularly for individual subjects.

I can see the universities doing it already... getting their PR team to plaster this shiny new ranking all over their website and clearing advertisements.

I can see the universities doing it already... getting their PR team to plaster this shiny new ranking all over their website and clearing advertisements.

Yes definitely. I think that as a ranking, this is basically useless, but as a tool it is useful. I think it is better to compare similar universities. There's no point comparing Brookes and Oxford when the students are of a completely different standard. But for someone trying to compare Oxford Brookes and Robert Gordon it would be useful to see that students from Oxford Brookes largely outperform what their backgrounds would suggest they would earn, whilst students from Robert Gordon do not. And yes they may suffer from sample-size issues or random variations as it is only based on data from 1 year, so its a WIP. They hope to start compiling new data and adding it each year.

Yes definitely. I think that as a ranking, this is basically useless, but as a tool it is useful. I think it is better to compare similar universities. There's no point comparing Brookes and Oxford when the students are of a completely different standard. But for someone trying to compare Oxford Brookes and Robert Gordon it would be useful to see that students from Oxford Brookes largely outperform what their backgrounds would suggest they would earn, whilst students from Robert Gordon do not. And yes they may suffer from sample-size issues or random variations as it is only based on data from 1 year, so its a WIP. They hope to start compiling new data and adding it each year.

According to this table, taking Computer Science at Oxford Brookes has a value added of +£6,134, whereas taking Computer Science at the University of Oxford has a value added of -£5,061. I don't know, perhaps there is some truth to that, but it makes it difficult for me to take this seriously as a measure of value. There are a lot of other very strange figures that come up. I can also imagine that this might fluctuate significantly year-on-year and suffer from sample-size effects, particularly for individual subjects.

Some data is pretty odd though. Expected earnings of Brighton medics is £51,561, whereas expected earnings of Cambridge medics (who have substantially higher UCAS points) is £42,000, a full 23% lower. That's quite a pay cut!

I'd look in the methodology but paywall...

OUR British university rankings are based primarily on “longitudinal education outcomes data” released in June by the Department for Education, which breaks down earnings for graduates by university, course and sex. We then match these figures to data on factors such as how selective each university is, as measured by UCAS entry standards. Wherever possible, we use course-specific entry requirements.

The model itself is an ordinary-least squares regression, where each observation is a specific course at a specific university and the dependent variable is median graduate earnings. The regression is weighted by the number of graduates in each course. Our independent variables are entry standards, field of study, the share of students from low-income areas, the share of older students, the share of students who attended private schools, the gross-value-added per person of the region the university is located in (eg, the north-east) and how far the university is from London, as measured by driving distance on Google Maps.

To produce an estimate of the value added by each particular university course, we first take the projected earnings numbers from our model and subtract them from the actual earnings numbers. We then aggregate these estimates across all courses to derive a university-level average. The advantage of this approach is that it allows us to compare like-for-like; Imperial College, for example, offers courses mainly in the more-lucrative fields of science and engineering. We argue that universities which offer students more choices in the arts and humanities should not be penalised for doing so.

We look forward to improving these calculations as more data become available. Currently, earnings figures five years after graduation are only available for a single graduating class. Oxford, for instance, is ranked far higher than Cambridge in our ranking because their graduates earned £3,640 a year more in our dataset—it’s possible that other classes of Cambridge graduates have outperformed their Oxford peers. Another potential pitfall is our use of median earnings, which compresses the variance in earnings across universities, and may under-emphasise the performance of the richest graduates from elite universities.

Finally, our analysis rests on the assumption that the residuals from our regression are a reliable proxy for the value added by a university. However, formally, all we can say is that our independent variables could not explain these remaining differences in earnings. They could well have been caused by some factor other than the education at a given university, or simply be the result of random chance. Future researchers would be served well by looking at individual-level statistics, rather than at data aggregated at the course level.

On a more conceptual level, it’s impossible to say how much of a graduate’s earnings can truly be attributed to value provided by the university they attended. A sceptic would argue that universities do not increase their students’ human capital, and simply serve as a filtering mechanism for employers to weed out weak job candidates. Another interpretation is that British university admissions are extremely meritocratic, and only the best pupils earn the privilege of enjoying superior educations at elite universities.

I haven't had a chance to read through the methodology, but one of the issues that strike me with this is that the 'geographical area' thing is bs and disadvantages London unis in favour of those in the South. Hence LSE and UCL's negative ratings. Studying E&M at Oxford isn't going to make you magically set your career in that city as opposed to London - London Paddington is barely an hour away from Oxford train station. Most E&M grads aren't going to be deterred from living in London by the mere location of their university - that's where all the juicy, high paying grad jobs are! Take this principle and apply it to any sector which is London-based, and you realise that the 'benefit' that London unis have is illusory. Everyone in banking/law/consultancy/accounting wants to work in London, irrespective of where they went to uni.

Regional contextualisation also ignores the location of one's parents/family. I've met plenty of kids at Durham who, despite living in what is nominally the poorest area in England, come from Home Counties, with families that can accommodate for them living at home while working, or with high enough salaries that the Bank of Mum and Dad can fund a mortgage. The fact that they happened to live up North did not pose any structural barriers to them moving back into higher paying, more prosperous areas upon graduation (the high relative presence of Durham grads in the City is testament to this).

Comparing value added also implies that the better unis have as much room to 'add value' to their students as the worse ones. I would humbly suggest that that's not the case - the relationship between value added and grades isn't linear. Top students who can get into any top RG aren't going to rely on that uni to save their careers - they already have the connections, grades, and likely know what they're doing. In fact, the benefits they'll get from that uni will come from the people they will study and network with, as opposed to anything the institution itself is responsible for providing (e.g. teaching). The uni could literally be a glorified library card with the occasional reading list and they'd still get into their grad scheme of choice. On the other hand, the CCCs who get into e.g. Bishop Grosseteste are likely to have far more untapped potential (be it academic or professional), that a uni can help build on, through higher numbers of contact hours, smaller seminars, etc.

And now for the good old critiques: this table is useless insofar as it doesn't have subject specific data, it doesn't compare like to like (no one has a dilemma between UCL and Aston, even for something like Engineering), and ultimately shows value added differences that are trivial (someone facing an Oxbridge dilemma, or a UCL vs LSE vs KCL dilemma, isn't going to be swayed by a few hundred pounds difference in value added salary - not absolute earnings - at age 27).