How to make higher education admissions fairer, quicker and better

This blog was contributed by Rob Cuthbert, Emeritus Professor of Higher Education Management at the University of the West of England and Managing Partner of the Practical Academics consultancy. He wrote the 2020 HEPI blog ‘A-levels 2020: What students and parents need to know’ (which has had more hits than any other HEPI blog ever). You can find Rob on Twitter @RobCuthbert.

Many people would like to change the UK HE admissions system by moving to a system in which applications and/or offers would be made not before, but after applicants have received exam results and know their qualifications. A system based on post-qualification applications is the worldwide norm, and the DfE is now considering the results of a consultation on possible changes. But ‘the popular opinion has been that post-qualifications applications is desirable but impractical thus it has been placed in the “too hard to tackle” pile’ (Post-qualification admissions: how we can make it work A report for UCU by Professor Graeme Atherton). The DfE Consultation acknowledged that ‘Implementing PQA would involve major administrative changes and have practical implications for other parts of the education system too, not only HE’. Universities UK warned that a post-qualification applications system is ‘unworkable’. But after two Covid-affected years the status quo also seems increasingly unworkable: some universities have exceeded their intended capacity, others have seriously under-recruited. What is to be done?

The DfE Consultation noted particular dissatisfaction among students who gain a place through Clearing, and students from BAME backgrounds, but all applicants have to choose their course using guesswork about universities’ preferences and guesswork about their own examination grades. Equally, HE institutions’ admission strategies guess the rate at which offers will convert to students enrolled, leading increasingly to over- or under-recruitment. Institutions use unconditional offers, ‘conditional unconditional’ offers and other devices to try to make admissions more predictable.

These problems are widely recognised, but other deficiencies go unnoticed just because they are so deeply embedded in the present system. Why are applicants limited to just five choices? Why must they drop three of the five choices before exam results are known? Why can they apply either to Cambridge or Oxford, but not to both?

What if we had a system that guaranteed:

• every student gets the best place they can, given the competition from other students
• every university gets no more than the number of students they want on each course
• there is no need to change the school year or the higher education year

What if all this were possible, and

• candidates could apply to more HE providers than at present, if they wish. It may not be desirable or efficient (for students or HEIs) for most students to make more than about six applications, but some may have good reason to go beyond six. The system should allow it, and allow students to apply to both Oxford and Cambridge
• all students do is work out their own order of preferences, without revealing them to institutions
• all universities do is to put applicants in their order of preference, using whichever criteria they wish, thus protecting institutional autonomy

There is such a system, and it can work using technology already proven in the UK and worldwide to be effective on a large scale.

Although the broader marketisation of HE remains controversial, market competition for students has always been taken for granted.  In the admissions market, ensuring supply matches demand has always been a problem. The current system lacks an effective mechanism to coordinate offers across the whole system, and therefore suffers from market congestion. The practical consequences are seen in the frantic Clearing period, when a combination of computer searches, stressful telephone calls and decisions made at short notice leads to unsatisfactory outcomes for many of the students affected. The less visible consequence, for all students, is the obligation to make early decisions about firm and insurance choices and discard other potentially desirable options.

Market clearing, co-ordination and congestion are familiar economic problems and there are proven methods for dealing with them at large scale, through ‘matchmaking’ for market clearance. Economists have long been involved in designing centralised clearinghouses which rely on matchmaking; such methods have been successfully applied in allocating pupils to schools in the UK and the USA. For example, in London: ‘Parents can rank up to 6 schools in any area. Matching procedure: Local Authorities use the student-proposing Deferred-Acceptance algorithm. Priorities [school ranking of applicants] and quotas [course capacities] … Within the permitted provisions, schools that are their own admission authority are free to choose their admission criteria’. The same approach is also used in allocating newly qualified doctors to hospitals throughout the UK and USA (the National Resident Matching Program). They are also used for university admissions on a large scale in many other countries, including Germany, China, Turkey and Hungary. The method has been adapted in a model developed by Aytek Erdil (Cambridge) and Battal Dogan (Bristol) which in outline works as follows.

Students apply as usual in the Autumn/Winter preceding their intended start date. Universities assess applications and form preliminary admission scores for their applicants, based on whichever criteria they wish. International and special-needs students may receive early offers. After A-level or other examination results, students may make additional applications to any course which accepts late applications. Students privately submit to UCAS their preference ranking of the courses to which they have applied. Universities incorporate A-level results into the preliminary scores, finalise their course admission rankings which list students from first-to-admit to last, and submit these rankings to UCAS. Having collected the information on how students rank their choices, how universities rank their applicants, and the capacity for each course, UCAS runs a matchmaking procedure known as the deferred acceptance mechanism. This procedure is operated via repeated rounds which mimic an idealised scenario of students and courses freely communicating with a view towards finding a match. Each student begins with their top choice and moves down their preferences only if it is oversubscribed with applicants ranked higher by that course. Having students and universities actually processing these rounds would take a huge amount of time, but the matchmaker already knows students’ preferences, universities’ ranking of applicants, and course capacities. Thus, the matchmaker can operate on their behalf the idealised rounds of negotiations as follows.

In Round 1, each student ‘applies’ to their top choice. Each course keeps — but does not yet admit — the highest ranked applicants, up to its capacity; all other applicants are rejected. In the next round, all rejected applicants ‘apply’ to their next favourite course. Each course considers its applicants kept from the previous round together with the new applicants. From this combined set of applicants, it keeps the highest ranked ones up to its capacity; all others are rejected. These rounds are repeated until we reach the position that all students have either been placed or have no place because they have not been chosen by any of their preferred courses.

The mechanism operates transparently, based on students’ full preferences and universities’ ranking of their applicants. It is fairer because no student misses out on a course at the expense of a less-qualified applicant. Students can hold all their applications right to the end of the admissions cycle – no need to make firm and insurance choices. It is quicker: the model can work through multiple rounds in seconds once preferences are known. It is better, because: universities can focus their efforts on assessing students without needing to game the system; DfE and UCAS do not need to monitor uncompetitive offer making, because circumventing the system (say via conditional-unconditional offers) is not profitable for universities or students; and, students are guaranteed the best match they can qualify for and universities are guaranteed to meet realistic capacity targets, without risk of over-recruitment – most targets will be met exactly. The model is designed for optimal fairness to students, by allowing the students to make the moves and the universities respond; inverting the procedure to allow the universities to ‘apply’ to students and the students to respond would also lead to a stable solution, but would not put ‘students at the heart of the system’.

It should also be possible to allow late applications, before the matchmaking process begins. This is a necessary safety net for the minority whose A-level results are very different (whether better or worse) from their predicted grades, allowing them to apply for any courses which are open to late applicants. This process substitutes for adjustment and clearing, but more effectively. Students know they can choose courses more focused on their true preferences, rather than including an ‘insurance’ choice as a non-preferred fallback position if results do not go as planned.

Timetable

• Students apply by mid-January, for September entry of that year.
• Preliminary assessment: based on all available information, possibly including predicted grades, universities form preliminary course admission scores, to be updated after results are known. Universities can then assess candidates quickly in August without ignoring other factors such as interviews, reference letters, contextual factors, or teacher predictions.
• Early offers: universities can make early offers to some applicants, e.g. from overseas, or with low-quintile IMD scores as in Scotland. Applicants are not obliged to decide on these offers until the main matching round in August.
• BTEC, IB, A-level results: released in July/August as usual
• Main matching round: Mid-August
• Each university lists courses willing to accept late applications and applicants make additional applications if they wish.
• Applicants privately submit to the matchmaker their preference ranking of all courses to which they applied.
• Universities finalise their course admission rankings, listing applicants for each course from first-to-admit to last. They privately submit to the matchmaker their target course capacities and their course admission rankings.
• The matchmaker runs the deferred acceptance algorithm.
• The minority of students who are still unplaced enter a probably much smaller-scale Clearing exercise as at present

Yes, but … there are any number of practical issues and questions which must be answered to the satisfaction of all the key stakeholders: students and their parents and supporters; schools and colleges; examining agencies; HE providers and their staff and managers; government and national agencies such as UCAS; and a range of others. The Erdil/Dogan model considers and addresses many of these detailed issues, and it is compatible with many other practical proposals, such as those recently made by Hertfordshire VC Quentin McKellar. This is because it takes fair and efficient market clearing seriously, by insisting that both students and HE providers fully express their preferences and recruitment targets. The model can operate on any set of criteria for admissions, or on any combination of different criteria in different universities, and it does not require students to reveal their preference order to universities, or the reasons for their preferences. The matchmaker model requires only stated preferences.

Practical issues for applicants and their parents

At present most applications are submitted by late January, and applicants then hope to receive offers from their chosen universities, in some cases after interviews. Many offers are made by about mid-March and are conditional, dependent on grades achieved in exam results not published until July/August. A conditional offer may make applicants feel that the outcome is at least in their hands and may be motivational during their final stage of secondary education. When applicants have heard from their chosen universities, they make a ‘firm’ acceptance and a second or ‘insurance’ choice (if they have two or more offers – they may of course receive any number of offers from zero to five). Any further offers are discarded. At this stage applicants may begin to think about planning their accommodation and starting preliminary studies for their chosen programme.

Alternative approaches may seem to offer less definite information at that early stage. The UCU/Atherton proposals require applicants to express interest in some courses, but to apply only after actual results are known: ‘Expression of interest is not application. It does not lock students, especially those from widening access backgrounds who have the potential to achieve the highest grades, into choices on the basis of their predicted grades as the present system does’. UCU also suggest that applicants should make less than five expressions of interest, perhaps as few as two.

Under the ED model applicants may perhaps be advised at some intermediate stage on their chances of admission to some or all of their chosen universities, but such advice could only be framed in terms of broad probabilities – perhaps in terms such as ‘good/moderate/remote’. However the ED model does keep all five (or more) options open.

Both UCU and ED approaches inevitably delay some information about which choices are achievable until after exam results are known. However, in all post-qualification offer and post-qualification application systems, students only have offers confirmed after their exam results are known.

UCAS interpret their data as meaning that 73% of applicants currently get their first/firm choice, but some or many might not be true ‘first choices’ because of the strategising and guesswork forced on every applicant – first in choosing no more than five courses, and then in deciding which offers to take as firm and insurance choices. Under the ED model applicants must still assess their chances of acceptance in different courses before applying, universities must still publish entry requirements, and schools and teachers must still guide their students in identifying a reasonable set of targets. However, in contrast with the existing system, the deferred acceptance mechanism makes it completely safe for the applicants to express their preferences over those choices honestly – it is what economists call ‘strategy-proof’. And there is no need to make a second strategic decision about firm and insurance choices. All applications remain live, and applicants may change their preference ordering, if they wish, at a late stage before the matchmaking process begins. The ED model may not improve the proportions of those getting first choices, but it will be fairer in the sense that any applicant missing their first choice has been displaced by students explicitly preferred by the university on the grounds of their stated entry criteria.

Practical issues for schools and colleges

Demands on schools and colleges may not change significantly, but the ED model might in practice lead to less dependence on predicted grades. The removal of unconditional offers might help sustain motivation for exams; there is some evidence of underperformance by unconditional offer holders.

Practical issues for HE providers

The ED model can accommodate intermediate stages at which universities and colleges can, if they wish, adjust their stated capacities to reflect initial provisional demand. Universities’ aims for diversity in the student group can be accommodated in the same way as at present, since the ED model allows providers to use whatever admissions criteria they wish.

The UCU proposals argue there is wasted effort for universities and students in handling as many as five choices. Increasing the number of choices would increase university workload, but universities need only identify their preferred applicants up to their capacity. Beyond that they might, for example, simply sort applicants into three lists: a small reserve list; a list of possibles (unranked); and a list of almost-certainly or definitely rejected (unranked).

The process of post-results offer/acceptance will work much as it is now. Applicants must confirm their final applications and preference order before the final matching. Institutions will then still be able to rely on their preferred candidates coming through, with greater confidence than at present.

Practical questions for government

The ED model addresses the supposed problem of ‘undermatching’ by expanding the options for applicants and enabling them to ‘take more risks’ in making applications. Crucially, the ED model relies on applicants’ explicit preferences, which does not patronise applicants by telling them they have made the ‘wrong’ choices.

Conclusion

Without changing the school year or the HE year we can move to an admissions system which is fairer, quicker, and better than at present. Applicants can have more choice and keep all their options open longer. Universities can have more security that they will hit target numbers. Schools continue to advise applicants as at present, and government can be reassured that student choice is at the heart of the system.