UK universities, and worldwide, are increasingly making decisions about undergraduate admissions with reference to various contextual indicators which are intended to identify whether or not an applicant comes from a disadvantaged family, neighbourhood or school environment. In the US this has been primarily focused on ethnicity, rather than social background as such. This is perhaps the most prevalent current approach to widening participation. However, the indicators used are sometimes chosen because they are readily available, without much consideration of the possible alternatives and their comparative quality. This paper is part of an ESRC-funded study to assess the potential indicators of context with respect to their quality, availability, and their relationship to outcomes in UK higher education. The paper involves a scoping review of existing international evidence, yielding around 120,000 reports initially, and 28 categories of indicators. Each indicator was assessed on the basis of existing evidence concerning its relevance, reach, availability, accuracy, reliability, and completeness –and in terms of whether its use might inadvertently create a different kind of injustice or lower the student outcomes for universities. It also involves looking at the National Pupil Database for England linked to university student records for all students, in terms of variables that could be used by universities to help them assess undergraduate applications. Considerations include missing data, and what is known about students for whom data is missing, changes in indicators of potential disadvantage over time, and the relationship between all indicators and student attainment and progress at school and beyond.

Many possible indicators are not readily available, or not accurate enough for use in policy and practice. In general, indicators concerning individual circumstances are more relevant than area-based or school characteristics. As expected, there is no ideal single indicator. And it is not clear that combining indicators leads to the advantages rather than the deficits of all – minimising both false positives and false negatives. The paper presents the results for the best available indicators. The safest and clearest indicators (such as sex and age in year) are not currently considered in widening participation, or affect very few students (such as living in care), while those that are currently used require considerable care and disaggregation (such as family income or special needs) or should not be used (such as area measures). The paper looks at the policy and practice implications of the findings of the review and secondary analyses before proposing a discussion with the audience about more radical solutions to the stratification of universities.

The analyses in this paper are based on HESA records for students at universities (HEIs) in England from 2008/9 to 2011/12, linked to the National Pupil Database (NPD) with records for all pupils in England who ended Key Stage 4 (KS4) in 2006. The linkage means that this paper is chiefly about the 93% of pupils from state-maintained schools and colleges in England. There were around 600,000 students in the KS4 cohort, and around 200,000 of these continued to HE. The records include the student background characteristics, university attended, degree results, school details, attainment while at school and post-16 in Key Stage 5 (KS5). The quality and completeness of most variables in the HESA data are considerably worse than for the equivalent students in the NPD dataset. Indicators of student sex, disability, ethnicity and all other variables that it is possible to compare directly with NPD are confused and have considerable missing data. For example, of 594,762 at school, a total of 392,332 do not appear in the HE data, leaving 202,420. Of these, only 188,735 (93%) have a known ethnicity that is compatible across the two measures, and 12% of HE students have unknown or unrecorded ethnicity (compared to only 2.5% in NPD). Therefore, where NPD and HESA datasets contain the same variables, the NPD version is used in this paper. Each possible context variable is considered in terms of its missing data, its links to all other potential context variables, and to variables representing attainment and progress at KS1, KS2, KS4, KS5, admission to HE, type of HEI attended and eventual degree outcome. Some indicators are recoded to simplify them and some are presented in several ways (e.g. ethnicity may be presented both as a binary flag and in terms of major ethnic groups). The Key Stage 5 (post-compulsory) attainment data is considered in terms of who stays on in education after the age of 16, who attains the equivalent of ABB, CCC, and EE or better at A-level, who enters HE, and who attains a first or 2.1 class degree (where relevant). A regression model is also created for each stage of progression from KS5 to degree outcome, to estimate the possible impact of contextual variables after prior attainment has been accounted for.

The overwhelming majority of students with appropriate KS5 qualifications continue to HE, and they do so largely without systematic differences in their background or family characteristics. The stratification of HE is almost entirely the stratification of pre-university attainment. Therefore, where contextualised admissions are used it is not because similarly qualified students with different backgrounds are being treated differently on entry to HE. It can only be because the prior qualifications of students are deemed unfair. In this case, it is pertinent to consider whether we should use qualifications in this way at all.

None