Predicting A-level performance

How well do AS marks predict A-level grades? Jeff Searle of Durham University has taken a detailed statistical look at this perennial question.

List of data and tablet showing graphsA-level qualifications are made up of two parts: the Advanced Subsidiary level (AS), usually taken by 17-year-old students after one year of study, and the full Advanced Level qualification, usually referred to as A2, normally taken by 18-year-old students after two years.

The Uniform Mark Scale (UMS) aims to maintain the standard of assessment grades awarded following different examination sittings. The A-level information system (Alis), which is run by the University of Durham’s Centre for Evaluation and Monitoring (CEM), makes predictions of A2 grades on the basis of a candidate’s GCSE grades. Since 2010, Alis has collected UMS data at both AS and A2 level. This data was used to investigate whether GCSE or AS performance is the more reliable predictor of A2 grades.

The data set used in the analysis in this investigation relates to candidates who sat A2 examinations in the summer of 2011, for whom the Alis system holds a score based on GCSE grades, predicted A-level grades and actual A-level grades awarded. The data set comprised approximately 40,000 candidates from 500 schools and colleges, representing about 100,000 subject entries. About half sat examinations in three subjects, but the number of subjects taken varied between one and six. The number of subject entries from schools and colleges showed wide variation, varying between 1 to over 1,500 with a mean of about 250. Only the 53 subjects with a minimum entry of 100 candidates were included in the analysis.

The correlations between the UMS scores at AS and A2 are of the order of 0.8, with a mean of 0.83, and between the GCSE grades score and A2 are of the order of 0.6, with a mean of 0.63. The higher correlations for the UMS scores suggest that using UMS as a predictor of A2 outcomes is likely to be more reliable than the GCSE grades-based score. This is not a surprising result as the AS UMS score is based on attainment at the end of the first year of study at advanced level, so the student is focused on a particular subject of his/her choice, and also the AS UMS score contributes to the final grade.

For each of the 53 subjects in the analysis, Table 1 shows the comparison between actual A-level grade achieved and the predicted grade based on GCSE grades, while Table 2 shows the comparison between the actual A-level grade achieved and that predicted by the AS UMS score.

With the exception of the anomalous-looking subject of environmental science, the UMS score at AS level consistently predicted the actual grade achieved by a candidate at A2 more reliably than did GCSE grades. For 34 subjects, the UMS system was at least 10 per cent more successful in predicting the actual grade than GCSE grades, and in 11 subjects it was at least 20 per cent more successful. 

The analysis indicates that a candidate’s predicted A-level grade in a subject is likely to be more reliable if it is based on the candidate’s AS UMS score in that subject than on GCSE grades. This should be of interest to all those who make use of predicted A-level grades.

Dr Jeff Searle is Research Associate at the Centre for Evaluation and Monitoring, Durham University.

Share this page