This fiasco shows that, despite their flaws, exams work

Never mind what teachers’ unions may say, there is no better way to strip bias and unfairness out of grading

There is usually some sort of system for gaming an exam. If it’s history, a student can strategise by covering one or two historical debates on the bare minimum number of topics and twist the material to fit a question. For Shakespeare, certain quotes are applicable to almost any question likely to come up. If it’s maths, it’s harder, but questions can often be sorted into one of several categories that repeat themselves, even if the format changes.

This is the sort of gaming students are meant to spend time on. This year, however, there was another way to increase the chance of success. Study a niche subject in a small class at a small, high-performing school, impress the teacher in the first term and then take a long holiday. This strategy would have been most effective in music or German, where the proportion of A/A* grades given out this week soared by 15-17 per cent compared to last year. For those studying maths or English, however, Ofqual’s magic eight ball, known officially as “the standardisation process”, had the data to ensure this could not happen. There was no rise in the percentage of A/A*S in these subjects.

Despite the fury, the debacle over A-level results may be one of the less damaging educational events in the lives of this year’s school leavers. Our universities, desperate for cash and bereft of foreign student fees, have extended a record number of offers to British students. So most of them are likely to be more concretely affected by missing four months’ education and suffering social distancing rules at university than by downgrades to some imaginary exam results. The debacle does highlight a salient truth, however: exams work and, despite years of educational culture wars about them, there is no better alternative.

That is not the message that teaching unions are taking from this mess. Some union leaders were pleased when national exams were cancelled in the spring. The National Education Union suggested, opportunistically, that it was time to “open up a longerterm conversation about the impact on our pupils of unnecessary high-stakes exams”, because “exams are not the way to assess someone’s ability”. Patrick Roach, head of NASUWT, the teachers’ union, suggested “there are lessons from this crisis” – namely, that teacherawarded grades ought to play a bigger role in assessing pupils and exams should be reduced in importance. Tempting fate, the NEU even went as far as to claim: “Teacher assessment is just as reliable and stable.”

The data, alas, now proves otherwise. The Government agency in charge, Ofqual, reported this week that if it had not applied an algorithm to adjust grades awarded by teachers, the proportion of A-levels achieving an A or A* would have shot up by an unprecedented 12.5 per cent compared to last year. With adjustment, it has risen 2.4 per cent instead. Teachers interviewed by Ofqual admitted that they were giving pupils the benefit of the doubt and imagining how they would perform on “a good day”. This is human nature. It’s kind-hearted and inevitable – but it’s not very “reliable” or “stable”.

Beyond the aggregate effect, there is ample evidence that teachers, like most of us, harbour biases that affect their pupil assessments. The Education Policy Institute highlights three studies suggesting that, when they assess ability or predict exam grades, teachers systematically stereotype their pupils based on ethnicity, socio-economic status, gender, their native language and their special needs status. There are nuances to the stereotyping, too. Pupils from more well-off families tend to be judged less harshly than their poorer peers, but all groups can be affected. If you look just at overachievers getting the top grades, teachers systematically predict worse grades than they go on to achieve. The effect is especially marked for those from deprived backgrounds.

It is no wonder that the Government felt it ought to avoid relying wholly on teacher assessments. In swerving that bear trap, however, it has fallen into an even bigger one. The algorithm Ofqual used tries to protect faith in the exam system by undoing teachers’ overoptimism and adjusting grades according to how well a pupil’s school has tended to perform in the past. So the performance of past cohorts affects the results gained by the class of 2020, however hard the students or their teachers have worked since then.

Ofqual claims that the effect is the same for all pupils regardless of their socio-economic background, but the process also attempts to avoid skewing results unfairly by relying on historic data from extremely small class sizes. Hence, the fewer students there are in a class, the less their grades have been subjected to the voodoo of this “standardisation process”. The effect, naturally, is to favour those studying niche subjects in small groups, a feature more common to private schools than to big sixth-form colleges.

But the algorithmic approach has an even bigger problem. It prioritises protecting the integrity of the grading system over the right of each pupil to be assessed as an individual. It does nothing to adjust for the unconscious bigotry of teachers’ expectations and instead adds another layer of hardwired bigotry on top, based on how your school has ranked in the past. So if you come from a bad school and a bad home, you get hit twice.

This, in a nutshell, is why we have exams. It is not because they are some sort of definitive evaluation of a person. There is more to character and intellect than taking a test on a certain day. But policymakers have yet to devise a better way to strip prejudice and cheating out of the grading system. There is one test; everyone takes it and they are all marked blind by examiners who don’t even know their names. Tests can be brutal, but their blindness to personality and circumstance is part of what makes them fair. Some minds may be more or less suited to them, but almost everyone can improve their results through effort. Just like Cv-blind job applications and fitness tests, exams work.

Teachers, parents and pupils love to criticise tests for being stressful and reductive. Perhaps it takes a shambles like this one to show that, whatever their flaws and limitations, they are an awful lot fairer than the alternatives. If there is any “lesson” to come out of this fiasco, it should be that we learn to value exams more, not less.

Tests can be brutal, but their blindness to personality and circumstance is part of what makes them fair