Ofqual's 2020 A-Level Standardisation Algorithm — the Grading Model Withdrawn After 39% of Results Were Downgraded

With exams cancelled in 2020 because of COVID-19, England's regulator Ofqual asked schools for centre-assessed grades and teacher rankings, then applied a statistical standardisation model to prevent grade inflation, using each centre's historical results.
Exam boards' moderation lowered about 39.1% of A-level grades, roughly 280,000 entries adjusted down. Independent analyses, including work at the University of Bristol, examined the effects, and commentators noted the model tended to reproduce existing inequalities, disadvantaging pupils from disadvantaged backgrounds.
The approach was withdrawn within days, with grades reverted to teacher-assessed ones. The Education Select Committee had earlier called on Ofqual to be fully transparent about its standardisation model, publish a programme of evaluation, test for bias against protected characteristics and provide clear appeal routes; its chair voiced serious doubts about Ofqual's independence, accountability and transparency.
The case is a widely cited cautionary example: an algorithmic decision with direct life consequences for individuals, deployed without enough transparency, bias testing or appeal design, and reversed under public pressure.

Read the full analysis: https://www.bristol.ac.uk/cmm/research/grade/