Explaining the use of prediction data

Friday, 12 August 2011 11:59

To help ensure that GCSE and A level results are comparable with the standards of previous years, awarding organisations use statistical information to predict the percentage of candidates expected to achieve the key grades (such as GCSE grades A and C) in each subject overall. This is a key tool for guiding awarders when they set grade boundaries and for maintaining standards. It is especially important to have this information when qualifications change - such as the new GCSEs being awarded for the first time in 2011. Awarding organisations also report against the predictions to the regulators (Ofqual in England, DfES in Wales and CCEA in Northern Ireland) to show whether outcomes are as expected. The predictions are used as a guide and are not targets or quotas.

All grade predictions are derived from the relationship between prior attainment and qualification results for a particular group of students overall. Predictions for A level outcomes for summer 2011 were derived from the relationship between A level attainment in 2009 and prior achievement at GCSE. This relationship is then used to predict the outcomes for the 2011 cohort. To map the expected outcomes for the 2011 GCSE cohort, the English awarding organisations looked at the relationship between GCSE performance in 2009 and that cohort's attainment at Key Stage 2. This allowed them to produce a model of the relationship they could use to produce expected outcomes.

Key Stage 2 data is used as the prior attainment measure as it is the most statistically reliable information available when predicting the expected achievement of the cohort of 16 year-olds taking their GCSEs. This is the average attainment overall at age 11, not achievement in any particular Key Stage 2 test subject. As there are no Key Stage 2 tests in Wales and Northern Ireland, awarding organisations also produce predictions based on common centre information. This is an analysis of GCSE performance data from a set of centres who entered candidates with the same exam board in 2008 and 2009.

It is important to stress that these predictions are used at whole cohort level, rather than at individual candidate level. These prediction data are part of a wider package of information used to support the judgments made by senior examiners where they consider candidates' performance during the awarding process. The use of both qualitative and quantitative information is required by the regulators when awarding organisations set qualification grade boundaries.

Tags: Ofqual, education, examination regulation, qualifications, assessments