Gavin Williamson was offered evidence of serious flaws in the grading of this year’s exams two weeks before A-level results were published in England, the Guardian has been told.
But Williamson and the Department for Education were reassured by the exam regulator Ofqual that the flaws could be managed by allowing schools to appeal in cases of unusual results.
It is the most compelling evidence to date that Williamson, the DfE and Ofqual had detailed warning of the problems that would later turn into a fiasco, disrupting the lives and education of hundreds of thousands of young people.
The details are likely to raise difficult new questions for Ofqual, whose chiefs go in front of MPs on Wednesday – but also for Williamson. At the time of the U-turn on A-levels, he suggested he had only known of the potential problems after the initial results were published on 13 August.
The new disclosures come from Cambridge Assessment, which operates OCR, one of the three big examination boards that administer most GCSE and A-level qualifications.
It said it approached ministers and the DfE in July, two weeks before A-level results were published, and again days before GCSE results were released, after it identified major problems in the way grades were being allocated by Ofqual’s statistical model.
Tim Oates, Cambridge Assessment’s group director of research, told the Guardian DfE officials including Williamson’s advisers and the schools minister, Nick Gibb, were eager to hear about the problems his organisation had uncovered.
“I cannot fault their reaction, the DfE responded swiftly and energetically, taking our analysis extremely seriously,” Oates said.
Cambridge Assessment’s internal research team identified anomalies that were punishing “outlier” students – talented individuals in large schools with poor track records – as well as middle-ranking students who were assessed as gaining Cs but were inexplicably being given a failing U grade by Ofqual’s model.
“This year’s approach – like all assessment – had technical limitations. We investigated those we identified through targeted research quickly and forensically. Two issues – outliers and the odd pattern of results – passed an evidential threshold and we contacted the DfE the moment this happened,” Oates said.
Cambridge Assessment’s warnings in July spurred the DfE into action, pushing Ofqual to allow schools to appeal if “the grades of unusually high or low ability students been affected by the model” or if they received a “very different pattern of grades” compared with previous years.
Sources within the government say that Ofqual assured the DfE that the enhanced appeals process would be enough to cope with what it said would be a small number of unusual cases generated by its algorithm.
In a statement to parliament on Tuesday, Williamson admitted that the appeals process that was to fix Ofqual’s model would have been inadequate.
He told MPs: “Ofqual had put in place a system for arriving at grades that was believed to be fair and robust. It became clear however that there were far too many inconsistent and unfair outcomes for A-level students, and that it was not reasonable to expect these to be dealt with, even through a boosted and enhanced appeals process.
“This situation has I know caused a great deal of stress and uncertainty and I’m deeply sorry that those who have borne the brunt of it have been students themselves. I can only apologise to them for this.
“We took immediate action to provide certainty, as soon as it was clear that if we didn’t, too many students would receive grades that didn’t reflect their hard work and their ability.”
But evidence submitted by Cambridge Assessment to the House of Commons education committee – ahead of its questioning of Roger Taylor, the Ofqual chair – shows that it approached the regulator and DfE a second time after it had found a significant flaw within a database used by Ofqual’s model that caused grades to be dragged down.
A timeline included in the evidence states: “14 August: Cambridge Assessment informs secretary of state’s office and Ofqual’s chair [Roger Taylor] that the likely cause of the anomalous centre grades appears to be the way in which Ofqual had implemented its national standards correction to maintain standards. The same issues were likely to arise with GCSE results.”
The timeline explains Williamson’s comments that he only learned of the serious problems within Ofqual’s model the weekend following A-levels, because it was then that the flaw was identified as likely to also affect GCSE results.
Cambridge Assessment’s evidence reveals: “We suggested to the DfE that GCSE results should be delayed and rerun, and we made Ofqual aware of the detail of our analysis and of our policy recommendations. Ultimately, Ofqual took another course of action.”
Taylor later announced that both A-levels and GCSE grades would be decided by school assessments.
Ofqual said: “These are astonishing comments. Cambridge Assessment and OCR were totally key to the development, testing and quality assurance of the algorithm right from the beginning through to delivery. Their warning about A-levels came a day after results were released and was incorrect.
“It also came as Ofqual itself was developing a possible appeals process. Ultimately, it’s clear that no appeals process would have provided a satisfactory outcome to this affair while still retaining national standards. That is why Ofqual ended up returning to centre-assessed grades.”
Relations between Ofqual and the DfE have deteriorated throughout the summer, culminating in Taylor’s threat to resign just days before GCSE results were published unless he was backed by the government. Last week Sally Collier, the chief regulator, stepped down.
Williamson has been the target of derision after he insisted there would be no U-tun on awarding grades through assessments, following school closures and exam cancellations prompted by the coronavirus pandemic. But Cambridge Assessment’s revelations lend support to Williamson’s insistence that he did everything he could to press Ofqual on its methods.
In its evidence to the education committee, Cambridge Assessment will say that confidence in GCSE and A-levels “has been eroded by events this summer and it needs to be urgently rebuilt”.
The exam board said the fiasco could have been avoided “if extensive checking of individual school and college results had been conducted prior to A-level results being awarded”.
But Ofqual appears to have done little or no modelling of its algorithm’s impact on individual schools, relying instead on the national distribution of exam results.
A spokesperson for the DfE declined to comment.