Market-based “choice” arguments tend to model choice at the point of individual decision – can this 18-year-old pick between provider A and provider B.
What they don’t model well is choice as a feature of the landscape over time – whether there continues to be a viable provider within reach of a given student in five or ten years.
These aren’t the same thing, and a system can improve on the first while destroying the second.
Uncapped numbers plus a recruitment market where reputation, marketing spend, and league table position are self-reinforcing (successful recruiters can invest more in the things that drive further recruitment) tends to produce exactly a concentration dynamic – it’s a standard winner-takes-more pattern in markets with strong network/reputation effects that is not unique to HE.
But if the system is increasingly reliant on a growing pool of commuter students for whom the choice set isn’t “any university in England” but “whichever provider is within commuting distance of where I already live,” then course or provider closures don’t just mean those specific students face disruption – they mean choice itself becomes geographically stratified.
Someone in a city with three universities retains real choice under a concentration dynamic. Someone whose only local option shuts a course doesn’t, regardless of what the national market looks like.
It’s a nasty asymmetry in who bears the cost of “the market working as intended,” and it maps onto affordability in a way that compounds rather than offsets existing inequalities. The winners, in other words, create losers. Those that can move away to university ultimately rob course viability from those that can’t.
Are courses closing?
Back in October 2024, my colleague David Kernohan made the best of a bad job with the only publicly available list of undergraduate courses in the UK – the Discover Uni dataset, the once-and-future Unistats of lore.
He found plenty of churn, some suspicious course recoding, and no obvious subject story. One year of comparison, he warned, could only ever be indicative.
I’ve been playing with four snapshots – the publications feeding the 2024, 2025 and 2026 application cycles, plus the live dataset as of the end of June 2026.
With a longer run of data it is possible to deal with some of the problems DK flagged, and to ask another question – not just how many courses have gone, but where they have gone from, and who is left with less to choose from.
Dealing with the code problem
Two adjustments. First, a course here is a unique combination of provider and Unistats course code, with modes of study collapsed so a course dropping its part-time route doesn’t count as a closure, and the totals are a little lower than the ones DK used.
Second, where a “removed” course code belongs to a provider that still lists a course with an identical title the following year, it is treated as a relabel rather than a closure. This deals imperfectly, but transparently, with re-coding artefacts at Bristol, Kingston, Norwich University of the Arts and friends that bedevilled the one-year comparison.
There are sharp edges in this data – part-time routes sometimes carry their own course codes, mode coding changes format between releases, and a handful of colleges are recorded at coordinates in the North Sea. The full methodology and code are available for anyone who wants to check the working, and the headline holds under the strictest sensitivity test we could construct – excluding every part-time-only course code entirely gives a decline of 8.2 rather than 8.4 per cent.
On that basis the catalogue has shrunk from 31,140 unique courses in the 2024 cycle to 28,512 today – a net loss of 2,628, or 8.4 per cent, in three years.
The gross flows are far bigger – around 16,000 course records have disappeared over the period, of which roughly 10,200 look like genuine closures, offset by about 13,300 new arrivals. The pace of net decline is slowing – minus 1,635, then minus 852, then minus 141 – but the churn underneath is not slowing much at all.
An obvious objection that someone made in the comments last time is that closing courses is not necessarily a bad thing, especially courses that have under-recruited for years. The data gives that objection real support – with a twist.
Strictly course-level entry populations only exist where cohorts are big enough to report, which is under half of courses. Within that subset, closed courses skew clearly smaller – a median measured entry cohort of 15, against 25 for survivors – and closed courses are nearly twice as likely to have cohorts too small to report at course level at all. A lot of what has gone really is small provision being rationalised away.
But that is the point of the geographic story that follows. Small courses are how less popular subjects exist outside the big cities at all – a fifteen-student cohort is invisible in national statistics and decisive for what a commuter student in Barnstaple can study. Rationalising small courses and narrowing local choice are not competing explanations. They are the same event, described from two ends.
I should say at this point that this is also one of the huge downsides of a culture that rocks thick prospectuses of hundreds of courses with little elective choice, rather than a much smaller number of “courses” with a huge amount of choice within. There may not be enough students wanting to commit to 360 credits in music or a language at your university. There may well be enough who would choose those as elective pathways to sustain a team of research active academics.
My favourite module back in 1995? Ray Billington teaching philosophy. Did UWE offer a full philosophy bachelors? It did not. Would I have enrolled onto it if it did? Are you serious?
Subjects churn and one-way streets
Last time round there was “no noticeable subject trend”. Over three years, there is. Net decline is concentrated in languages (down 827 courses, or 18 per cent), social sciences, creative arts, history and philosophy, media, education and geography. Computing, subjects allied to medicine and medicine itself grew.
But the more interesting signal is the relationship between churn and net change, because it separates two different phenomena. Computing churns furiously – nearly half its 2024-cycle catalogue has turned over – and replaces everything it removes, and then some.
Churn itself is neutral – in a fast-moving field like computing, high turnover with full replacement looks like a portfolio keeping up with the labour market. What churn measures is how stable the offer is from an applicant’s point of view. The worrying combinations are at the extremes. Languages have the lowest churn rate of any subject group, so what closes there is simply not replaced – less refresh, more one-way attrition. Education and teaching combines heavy turnover (47 per cent) with an 18 per cent net decline – the offer is being restructured and shrunk at the same time.
Where the shrinkage is
Every teaching location in the dataset carries coordinates, which means the national numbers can be taken apart geographically – something the sector-level releases from regulators never quite let you do.
Assigning each location to a nation or English region (distance-learning-only courses excluded), the decline turns out to be noticeably uneven. The courses available in London, Scotland and the East of England have actually grown slightly since the 2024 cycle.
The South East is down 16 per cent, Wales down 12.6 per cent, Yorkshire and the South West down more than 8 per cent each. The North West’s headline fall of 28 per cent is dominated by Liverpool Hope collapsing its combined honours matrix – but even excluding Hope entirely the region is down 12.8 per cent.
Cross the regions with subjects and the languages story gets worse. Language and area studies provision has fallen by 30 to 50 per cent in the North West, the East Midlands, the South West, Wales and the South East, while barely moving in London and growing in Scotland.
A subject that was already scarce outside the big cities is retreating to them. London’s share of all course availability has risen from 13.1 to 14.5 per cent in three years, and its share rose in 16 of 21 subject areas.
What you can reach from where you live
Regions are big, and most students – particularly the growing number who commute – do not choose from a region. So for around a hundred major towns, I’ve counted how many of the 167 detailed (CAH3) subject areas had at least one undergraduate course taught within 25 miles, in the 2024 cycle and now.
Two thirds of towns have a shorter menu than they did three years ago; the median town has lost two subjects. The biggest losses in the country are clustered in Kent, where the contractions at Kent and Canterbury Christ Church compound in the same catchment – Canterbury has lost 24 subject areas from its 25-mile radius, Folkestone 20, Maidstone 16.
And the losses hurt hardest where the menu was thinnest to begin with – Barnstaple has gone from 15 subjects in reach to eight, Hastings from 25 to 16, Yeovil from 28 to 22.
Meanwhile London, Manchester and Leeds are essentially untouched, and the clearest growth is in Essex and Suffolk.
A note for the member for Ashford
Towns are how students experience this. Constituencies are how politics does. So here is the same 25-mile logic translated into Westminster’s geography – for each of the 650 constituencies, the number of subject areas with at least one undergraduate course taught in the constituency itself or a neighbouring one, in the 2024 cycle and now.
Two thirds of constituencies have lost subjects. And the incidence is strikingly cross-party – 64 per cent of Labour seats, 62 per cent of Conservative seats and 61 per cent of Liberal Democrat seats are down, with an identical median loss of two subjects. Nobody’s voters are exempt.
But the deep cuts are not evenly spread – six of the ten worst-hit seats are Labour, including a contiguous Kent bloc – Ashford, East Thanet, Dover and Deal and Folkestone and Hythe, alongside Conservative Faversham and Herne Bay, and Canterbury itself – where constituents have lost around a fifth of the subjects previously within reach. All four Plaid Cymru seats are down. And Reform UK’s eight seats start from the shortest menus in Britain – 57 subjects within reach on average, against 91 in Green-held seats.
This is not a story universities are rushing to tell MPs. The sector’s public affairs output runs on openings, civic agreements and economic impact assessments – course closures travel as internal restructuring announcements, if they travel at all. No one sends the member for Ashford a note explaining that school leavers in their constituency now choose from 23 fewer subject areas than the cohort three years ago.
But somebody should – because the people this lands on are constituents first and applicants second, and the MPs now fielding questions about opportunity in coastal and post-industrial England are mostly unaware that the local opportunity set is shrinking on a spreadsheet nobody publishes.
It matters more, not less, as the skills agenda gets more local. The Greater Manchester Baccalaureate is the template everyone cites – a map from school subject choices to gateway industries and the local routes into them. Every version of that idea, MBacc or otherwise, assumes the destination end of the pathway exists within reach – that the young person who follows the signposted route to engineering, or health, or the creative industries will find a local course at the end of it.
This data is the first systematic look at whether that assumption still holds, constituency by constituency – and in a growing number of places it is fraying exactly where the pathway rhetoric is loudest. If mayors are building pipelines and MPs are selling opportunity, someone needs to be watching the other end of the pipe.
What a Mayor
Cut the same data by England’s devolution geographies and the pattern lands directly on mayoral desks – existing mayoral authorities have collectively lost close to one in ten of their advertised university courses since the 2024 cycle, and toggling to all provision (including the further education colleges that serve most cold spots, and over which mayors have rather more power) redraws the league table in instructive ways.
If government is serious about cold spots, shortage subjects and regional access to provision, it is flying with instruments nobody designed for the job – and if Metro Mayors wanted data on what’s on offer, this bodge is it.
I’ve included a STEM toggle – not because other subjects don’t matter, but because all the evidence points to a serious shortage of STEM grads in regions. By the time devolution happens, there might not be much left depending on which chunk of geography we’re looking at.
The usual caveats, and the usual conclusion
Everything above measures the shop window, not the shop. This data captures what is marketed to applicants at a point in time – not whether a course actually ran, how many optional pathways survive inside it, or how much of it shares modules with the programme next door.
A provider consolidating twelve named degrees into three broad ones with pathways shows up here as nine closures, while a surviving degree quietly stripped of half its options shows up as no change at all. Those two errors run in opposite directions, which is why the provider-and-subject presence measures above are more trustworthy than raw course counts.
Twenty-five miles is measured as the crow flies, and English regions are assigned from campus coordinates, so a handful of border cases will sit one region out.
But the direction of travel is consistent across every way I’ve cut it – the advertised undergraduate offer is smaller than it was three years ago, the shrinkage is concentrated in particular subjects and particular places, and the places losing most are disproportionately those that had least.
Students in Barnstaple, Hastings or the Kent coast now choose from a visibly shorter list than their equivalents three years ago – students in London do not.
All of which sharpens the point DK ended on last time. This dataset built to power a comparison website, full of recoding quirks, missing anything about capacity or delivery remains the only instrument anyone has, regulators included, for watching the national course portfolio change.
The obvious fix has been sitting there all along. UCAS holds a complete, current, structured course catalogue – and treats it like a commercial secret. It is a public asset, and it should be published as open data. That still needs to change.
