Exam appeals made free as fears grow over GCSE results

ALL exam appeals will be free this year, the Education Secretary announced last night in an attempt to see off rising anger ahead of GCSE results day.

Gavin Williamson said that his department would cover the costs so that schools were not left out of pocket, as he announced that he would set up a Government task force to oversee the appeals process.

His intervention came as statisticians predicted that GCSE results day on Thursday would be an even bigger disaster for students than A-levels were. The exams regulator’s own analysis of its algorithm being used to calculate most results this summer found that it was able to predict GCSE grades less accurately than it could predict A-level grades.

Prof Guy Nason, chair in statistics at Imperial College London, told The Daily Telegraph that he was struck by Ofqual’s admission. “I am extremely worried about next week,” he said. “If people think A-level results were bad, what is going to happen then?”

Ofqual should have been more open about its model so that a “mature discussion” could have taken place earlier, he added.

Statisticians pointed out that the more grades there were, the less accurate any estimate would be since it increased the likelihood of falling on the wrong side of a boundary. Since there are more grades in GCSES than at A-level, this will inevitably lead to the algorithm being less accurate, they said.

A technical report published by Ofqual said that while the model accurately predicted 93 per cent of A-levels that were at most one grade out, it only predicted 55 per cent of GCSES with the same precision.

“Based on the testing of the approaches applied this summer using results data from 2019, 51 of the 55 Alevels tested had accurate predictions for more than 90 per cent of students within plus or minus one grade,” the report said. This figure was lower for GCSES, with 12 out of 22 subjects.

“We extensively tested possible variations of the model to ensure we selected the one which gives students the most accurate results possible,” an Ofqual spokesman said.

Last night, it emerged that Nick Gibb, the schools minister, was ordered to set up a task force to monitor the appeals process, meeting daily with Ofqual and exam boards.

THE A-level algorithm predicts inaccurate grades at least a third of the time, the exam regulator’s own analysis reveals.

The accuracy of A-level grades, calculated by Ofqual’s statistical model, range from 68 per cent accurate, at best, to just 27 per cent accurate at worst.

Some of the most popular subjects, such as Maths, Biology and English Literature are only accurately predicted 61, 65 and 64 per cent of the time.

The figures, published in a technical report by Ofqual, come amid growing anger about this year’s exam results, which were largely calculated by an algorithm after exams were cancelled due to coronavirus.

Robert Halfon, education select committee chairman, yesterday joined the growing chorus of Tory MPS, as well as head teachers, statisticians and social mobility experts to voice concerns.

Gavin Williamson, the Education Secretary, faces calls to address the “huge injustice” for this year’s pupils after results showed poorer students in England were more likely to have marks downgraded by the algorithm.

“I am worried about it because some figures suggest that disadvantaged students have been penalised again,” Mr Halfon told BBC Radio 4’s The World at One. “If the model has appealed disadvantaged groups this is very serious and if it has disadvantaged colleges that has to be looked at. Ofqual will have to adjust the grades.”

The exam watchdog published an analysis of the accuracy of its algorithm where it tested it out by comparing its results to the 2019 data set of A-level results in every subject.

The algorithm was able to predict History A-level grades with the most precision of all subjects, at 68 per cent accuracy, followed by English Language with 67 per cent and Psychology with 65 per cent.

But its predictions for A-level Italian, German and Music were only 27, 42 and 47 per cent accurate respectively.

Meanwhile, the algorithm’s accuracy of producing a grade that is one grade out was far higher, at the most 99 per cent for History and at the least 75 per cent for Italian.

On Thursday thousands of students received their A-level results but many felt they had been unreasonably punished by the statistical mechanism that was used to calculate grades.

Mr Williamson is under growing pressure to change the grading system to that of Scotland, where grades created by an algorithm were withdrawn and teachers’ predicted results will take their place.

Robert Cuthbert, an emeritus professor in higher education management at the University of West England, said: “All these problems were entirely predictable… but it has not been well understood. Ofqual and the Government keep saying it is robust and fair, which is baffling. How could it be robust and fair?”

Ofqual’s technical report explains that all exam systems contain an element of unreliability, including during a normal year when there are

‘All these problems were entirely predictable but it has not been well understood’

exams. “While a degree of assessment unreliability always exists in any qualification system, and leads to students being awarded grades other than their ‘true grade’, it is likely that in the absence of exams and assessments, it will be more challenging to award reliable grades this summer,” the report says.

Barnaby Lenon, chairman of the Independent Schools Council and a former headmaster at Harrow School, said it was a mistake for the Government to

“make claims about the grading system that were never going to be feasible”.

He explained: “What we have now is quite large numbers of individuals whose grades are wrong. It might have been better to acknowledge that there was going to be a significant group whose grades will need to be adjusted.”

Mr Lenon, who is a former chairman of Ofqual, said that while the regulator is committed to fairness and accuracy, they were asked to produce something that “could never be fair or accurate”. Ofqual said it was important to remember that students had not actually taken exams this year and so results “can only ever be our best estimates”.

“We extensively tested possible variations of the model to ensure we selected the one which gives students the most accurate results possible,” a spokesman said. Schools can appeal or encourage students to take exams in the autumn, it added.

‘What we have now is quite a large number of individuals whose grades are wrong’