Predicted Grades’ Gender Gap Sparks Doubts About University Admissions Accuracy

A large study of more than 350,000 students in England suggests that teacher grade forecasts may embed a gender pattern, even when boys and girls eventually achieve the same A-Level results. Conducted by researchers at the Universities of Cambridge and Bristol, the work points to systematic differences in how predicted grades align with later exam performance, raising stakes for admissions systems that rely on predictions.

Using data from three university application cycles between 2017 and 2019, the team examined whether teachers’ forecasts differed for boys and girls across ten A-Level subjects. They report a consistent “pro-girl” gap: depending on the subject, girls were more likely to receive overpredicted grades and less likely to receive underpredicted ones compared with boys.

Crucially, the study does not claim intentional favoritism. Instead, it argues that teachers may form expectations using a mixture of signals linked to academic trajectory—such as motivation, diligence, and classroom engagement—attributes that can vary across groups and influence judgment under uncertainty.

This mechanism could generate a “halo effect,” where positive impressions about students’ broader performance bleed into estimates of future achievement. In practical terms, even modest forecast differences matter: the researchers note that average prediction gaps of around one tenth to one fifth of a grade could push students across grade boundaries (for example, from B to A or from A to A*).

To test whether other factors explained the pattern, the researchers controlled for socioeconomic background, ethnicity, school attended, and other contextual variables; the results persisted. They then compared students who matched not only on A-Level outcomes but also on prior academic records.

When these prior profiles were aligned, the gender story changed. In non-STEM subjects, girls retained a smaller advantage, while in STEM subjects the pattern reversed, with boys receiving higher predictions. The authors interpret this as evidence that earlier academic signals—captured through cross-disciplinary competence—shape forecast formation.

The gender gap was also most pronounced among lower-achieving students, consistent with a broader idea in educational assessment: when performance is less predictable, judgment becomes more dependent on supplementary cues, increasing the risk of skewed estimates.

Published in Economics of Education Review, the findings feed into ongoing debates about university admissions. In the UK, initial offers are made months in advance and predicted grades play a central role, despite policy discussions about shifting decisions until after final results.

The researchers emphasize that official guidance says predicted grades should be independent of student behavior or background. They call for further study, but also for careful reconsideration of fairness—especially where high-stakes decisions hinge on forecasts that may reflect more than raw ability.