Statistical modelling to decide exam grades
Doubts over the reliability of teachers’ predictions for A-levels and GCSEs have brought an about-turn
MOST A-level and GCSE results will be decided by statistical modelling rather than teachers’ predicted grades, in a major about-turn by the Government.
Teachers’ predicted grades will serve “little or no purpose” in the modelling that determines the majority of pupils’ results, sources have told The Sunday Telegraph.
Concerns over the reliability of teachers’ predictions – in particular, their tendency to inflate pupils’ grades – led to a decision by Ofqual, the exam regulator, not to rely on them.
Hundreds of thousands of students will receive their A-level and GCSE results next month, despite all exams being axed this year.
Ofqual has promised that “no child will be penalised” by this year’s system and that all exam grades will be just as fair and valid as those in previous years.
But head teachers warned last night that this will be a year where some pupils are “bound to feel victims in a process which is not of their own making”.
Geoff Barton, general secretary of the Association of School and College Leaders (ASCL), said Ofqual has done the best it can but added: “This was always going to be a problematic year. There will be people who think the system isn’t fair this year, but the reality is that Covid isn’t fair.”
In March, the Department for Education (DfE) said teachers who “know their students well” will be asked to send exam boards the grade they believe the student would have received if exams had gone ahead.
Teachers were told to take into account pupils’ marks in mock exams, as well as the quality of their work throughout the course when coming up with the grade.
Ministers said exam boards would combine this information with other data, to produce grades for each student, but this will no longer be the main factor for most students.
“If Ofqual are able to use statistical standardisation to produce grades, then in effect the teachers’ predicted grades serve little or no purpose,” a source told The Sunday Telegraph.
“Teachers by nature are optimists; they want the best for their students. I don’t think anyone has ever thought they would be sufficiently accurate that they could be implemented as results,” the source continued.
An analysis by Ofqual, published earlier this week, found that teachers bumped up predicted A-level marks by 12 per cent on average and GCSE marks by 9 per cent. The watchdog said it was “not surprising” that grades predicted by teachers were optimistic since teachers “naturally want to do their best for their students”.
Ofqual said their researchers tested out 12 different statistical models and chose one which drew on a number of factors, including data on pupils’ previous education attainment as well as previous results of students at the same school.
Under the model, the “rank order”, which teachers drew up for this year’s students, will also play an important role in determining grades.
Mr Barton said it is “inevitable” that teachers’ predicted grades are optimistic, adding the model was “the fairest thing they could do”.
Ofqual insists that teachers’ predicted grades are an “important component” as they helped teachers to draw up the rank order.
The predicted grades were also useful for testing out different statistical models, and for “quality assurance” of the end results, the watchdog said.
For new schools, which do not have historical data, as well as small schools or those where low numbers of students are taking particular subjects, teachers’ predictions will be relied on as the “primary source” of evidence for grades.
A DfE spokesperson said: “The vast majority of students will receive a calculated grade this summer that enables them to move on to the next stage of their education or training. Ofqual has developed a robust process that will take into account a range of evidence.”