‘Bias’ in predicted A-level grades casts doubt on admissions role

Teachers are “systematically” predicting better A-level grades for girls, with the “unconscious” practice suggesting it could be time to revisit the debate on post-qualification admissions, according to a new study.

The analysis of some 351,373 pupils in England finds girls may be benefitting from a “halo effect”, in which their broader approach to classroom engagement and their academic behaviour influences their expected exam results – even when their male peers went on to achieve the same results in the end.

When analysing three university cycles between 2017 and 2019, on average, girls received predictions between one tenth and one fifth of a grade higher than boys – a difference which could in many cases could tip them over a grade boundary.

Economist Konstantina Maragkou, an assistant research professor at Cambridge’s Faculty of Education and the study’s lead author, said: “It is very unlikely that teachers deliberately predict better results for girls, but they may be relying on an impression of student characteristics which can lead to gendered predictions.

“We need to think carefully about how fair that is for young people when the stakes are so high.”

The findings could reignite conversations around the fairness of the UK’s admissions system, in which those going to university receive offers on the basis of predicted grades – an unusual approach compared to other sectors.

The Conservatives had initiated plans to move away from the model, suggesting it was “letting down the brightest pupils from the most disadvantaged backgrounds”, but ultimately scrapped the change to post-qualification admissions in 2022.

University admissions service Ucas has in recent years suggested the UK is now leaning towards a “best-of-both-worlds” situation, with the popularity of clearing offering thousands of students the chance to reassess their options after receiving their A levels.

Nonetheless, the new paper, published in the Economics of Education Review, finds that “teacher assessments capture information beyond standardised achievement measures”.

This is despite official guidance advising “predicted grades should be independent of student behaviour or background”.

The study is based on analysis of “cross-disciplinary competence” – considering how students who went on to study STEM subjects fared in English earlier on in their school career, and the maths performance of those who went on to study non-STEM A levels.

This allowed researchers from both the universities of Cambridge and Bristol to look at an overarching picture of pupils’ academic attainment, and to understand how it might manifest in later grade predictions.

Across 10 subjects, including English literature, economics, history, biology, and psychology, girls were between 3 per cent and 11 per cent likelier than boys to have their grades overpredicted, and between 15 per cent and 41 per cent less likely to have them underpredicted. The majority of the sample (75 per cent) applied to at least one course at a selective university, compared to 62 per cent of the full population of applicants.

The study finds the gender gap in predictions is even larger among lower-attaining students – though researchers note that the differences between the sexes operate according to a “subject-specific pattern”.

“Girls tend to receive relatively lower predicted grades in STEM subjects, whereas boys tend to receive relatively lower predictions in non-STEM subjects,” the paper notes.

Maragkou said the findings should prompt further discussion.

“Teachers are not supposed to base predicted grades on students’ character, but they are probably not doing so consciously,” she said. “We also need to ask whether it is always wrong, conceptually, to consider broader qualities that affect academic success. These are difficult questions and they deserve careful consideration.”

georgia.luckhurst@timeshighereducation.com