Oxford or Cambridge? A West-East divide!

I've been around Oxford and Cambridge for a decade now: I did my undergrad and PhD at Cambridge before jumping ship for a Junior Research Fellowship at Oxford. You meet people from all over the UK at both universities, but I've always had an impression that people from different parts of the UK were more likely to apply to one or the other. As a teenager in Scotland it felt obvious that I'd apply to Cambridge, while it was equally obvious that the 'posh' South-Englanders would apply to Oxford.

I've recently been experimenting with the Bokeh interactive visualisation library so I decided to put my subjective impressions to the test. I took the number of applications from each region over 2017-2019 from the admissions websites of both universities, then computed the fraction of Oxbridge applicants from each region who applied to Cambridge. The visualisation below is surprisingly interactive: you can zoom, pan, and hover over regions to see the exact numbers.

Bokeh Plot

There is a clear geographic trend with those on the West favouring Oxford and those on the East favouring Cambridge. The most extreme region is the South West where only 37% of Oxbridge applicants choose to apply to Cambridge. Scotland wasn't as extremely Cambridge-biased as I was imagining with only 54.3% of applicants choosing Cambridge. One thing to bear in mind is that Oxford admits a larger number of undergraduates each year, causing the average of 47% to be biased towards Oxford.

If I had to guess at a reason for the geographic divide, then my first thought would be the British rail network. The mainline railway routes in the UK all terminate in London causing the British railway system to look like a hub-and-spoke. It is much easier to travel to destinations along your spoke than destinations on other spokes, because you don't need to go via London! This phenomenon inspired the 'British Rail metric', which says that the distance between Point A and Point B is the distance from Point A to London plus the distance from London to Point B. It is easy to imagine reasons that make people more likely to apply to a university that they are closer to 'as the train flies'.

The code I used to generate that visualisation can be found on my GitHub here.

Comments

Post a Comment

Popular posts from this blog

Gaps between primes

Image
One of my favourite YouTube channels is Matt Parker's Stand-up Maths . The videos are full of fun mathematical oddities and I always come away wanting to play with those oddities myself. In one recent video he discussed the gaps between consecutive prime numbers, with those gaps getting on average longer as the prime numbers get bigger. In the screenshot below, Matt is showing a histogram of the gaps between all consecutive primes less than N=150 million. The most common gap size in this histogram is 6. My mind was then blown when Matt revealed that 6 is only the most common gap until N=1.74x10^35, when it is overtaken by 30. 30 itself is overtaken by 210 at N=10^425, and so on. These are the primorial numbers: 6=2x3, 30=2x3x5, 210=2x3x5x7. There is an open conjecture that every primorial number eventually becomes the most common gap size as N tends to infinity! I wanted to visualise that transition for myself. Rather than computing the gaps between all the consecutive primes less...
Read more

Interactive HEALPix maps with Bokeh

Image
One of the most common plots in astrophysics is the 'all sky map', which visualises the value of some quantity across the entire sky. For instance, you can visualise the number of stars per square degree catalogued by the European Space Agency's Gaia space telescope by downloading a dataset from CDS and plotting it with the Python healpy library. These visualisations are attractive, but really not that useful. You can compare two points on the sky and tell which has a greater density of stars, but you can't answer detailed questions like 'what are the coordinates of this interesting cluster of stars?' or 'how many stars are at this precise location?' They are also quite large files, requiring a large number of image pixels to represent the shapes of the much smaller number of HEALPix pixels (which are misaligned with the x-y pixels on your screen). I decided to address these issues by creating an interactive version of these plots using the Python Bok...
Read more