David Kernohan is Deputy Editor of Wonkhe

The idea of the A level to university pathway is crashing into the less elegant reality of the process in a way that is likely to have repercussions far beyond the plague year.

In the popular imagination an A level (or Higher) grade is a perfect encapsulation both of a students attainment at the end of the compulsory education pathway and of their potential to succeed in their future academic endeavours. Next only maybe to the choice of university, it is one of the last moments in life that an externally applied label will overshadow more realistic expressions of personal potential.

Gavin Williamson says:

We know that, without exams, even the best system is not perfect.”

The system with exams used in every other is far from perfect. It may not even be the best.

A levels unmasked

A levels are actually a better measure of where you are from and how well off your parents are. There’s a great deal of evidence that your level 3 attainment does not predict how well you do as you continue.

The university “required grades” that loom so large when aged 18? Largely for show. In reality, a university that has already made you a firm or insurance offer will happily lower grades for a student they already know – or even, in many cases, those they don’t. The tariff of a course bears little relationship to the qualifications owned by those actually studying it.


Teachers are notoriously – but dependably – not good at predicting A level grades. Every year your average student will do two-to-three grades worse in their top three A levels than predicted on the UCAS form. This gap has widened over the years, suggesting either that teachers are getting worse at predicting grades – or that exams are getting worse at demonstrating the potential that students display.

So of course the predicted grade component of this year’s A levels and highers would produce a national grade pattern that is out of sync with recent years. In 2019 75.8 per cent of A level entries in England resulted in a grade of C and above. If the predicted grades in 2020 became the final results, OfQual tell us that 82 per cent of A levels would be graded C or above.

So, in most cases, OfQual have not used teacher predicted grades in the final calculation. Instead this relies on previous performance by students from a school, weighted by the attainment backgrounds of the students from this year compared to those for previous years. This weighted curve is laid on top of a ranked list of entrants for that subject from that school, compiled by teachers. If the model says the top 20 per cent of students in a subject at a school got an A* – that’s what the top 20 per cent of students are awarded.

The law of the curve

Does this sound arbitrary or unfair? A levels are graded to a curve every year, meaning a good performance in a bad national year will get a better grade than the exact same mark in a good year. An A in 2019 is not directly comparable to an A in 2018, an A in 2012, or an A in 1996. Somehow we manage.

The pressures placed on the government in Scotland concern the differences between the predicted and awarded grades – in particular the difference for students from less advantaged backgrounds. What’s missing from press coverage is the uncontested fact that exams do a very similar job of causing less advantaged students to do less well than teachers may hope and expect every other year.

Are students from these backgrounds less likely to succeed in higher education? POLAR4 quintile 1 – a decent but not perfect proxy – students are slightly more likely not to continue with their studies (about three percentage points difference from all other students). There’s also a pattern where students who did less well at the end of their compulsory education are more likely to fail to complete their course – remember that there is a correlation.

What happens next?

The results of 2020 foreground an inequality of attainment that has existed for a long time. It feels worse because we can no longer tell ourselves comforting stories about a good or bad day in the exam hall (that are themselves under-examined expressions of privilege).

I write before the announcement of measures to rectify this 2020 problem from John Swinney, and as Michelle Donelan writes to vice chancellors to ask offers to be held while students appeal. In England applicants that appeal,  and end up with better grades that meet offer conditions, will not be counted in student number controls – a kind gesture marred only by the fact that students in England cannot appeal directly unless they feel their school has made a mistake. In Scotland the changes are likely to be wholesale to forestall a great number of student appeals and to attempt to reinvigorate public confidence – even though the resulting grades are likely to be significantly higher on average than in comparable years.

But the problem is not with the 2020 arrangements. The problem is with A levels, and how poorly they serve less advantaged students. Those arguing, as UCU has done, that we need to move to a system that removes teacher predictions from admissions entirely, need to watch what happens next very carefully.

One response to “What if the problem is A levels and Highers, not 2020?

  1. Minor point compared to the main thrust of the article, but I’d suggest the inaccuracy of predicted grades is baked into the system, rather than just a sign of teachers being bad at predicting. We actively encourage an optimistic approach – what would they get ‘on a good day’, ‘if they put in the work’ etc. – so it’s no surprise to see the predictions average out higher than the results.

    Which is not to say, of course that that accounts for all of the gap, or explains the widening…

Leave a Reply