Minority report: the impact of predicted grades on university admissions of disadvantaged groups

ABSTRACT

We study the UK's university application system, in which students apply based on predicted examination grades, rather than actual results. Using three years of UK university applications data we find that only 16% of applicants’ predicted grades are accurate, with 75% of applicants having over-predicted grades. However, high-attaining, disadvantaged students are significantly more likely to receive pessimistic grade predictions. We show that under-predicted candidates are more likely to enrol in courses for which they are over qualified. We conclude that the use of predicted rather than actual grades has important implications for labour market outcomes and social mobility.

KEYWORDS:

• Higher education
• university application
• information
• mismatch

JEL CLASSIFICATION:

• I22
• I23
• I28

Acknowledgements

We thank and acknowledge University and College Union for funding and support, particularly Angela Nartey for valuable guidance and comments. We also thank participants at the International Workshop on Applied Economics of Education (IWAEE) 2017, and the UCL Institute of Education Quantitative Social Science seminar series for useful comments. Wyness acknowledges funding from University and college Union. All errors are the authors’ own.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Note that practically all university degrees charge the maximum tuition fees allowable – meaning there is no variation in fees across degrees. However, disadvantaged students may want to attend their local university to continue living with parents. This would will result in them attending less selective universities on average compared to more advantaged students who are willing to travel further.

2 There are 187 universities in the UK, while a further 204 colleges provide higher education courses. Source: HESA (2020) and AoC (2019).

3 The General Certificate of Secondary Education (GCSE) is a set of compulsory exams taken by students aged 15–16, after two years of study. Most students take between 5 and 12 subjects.

4 During our period of analysis, the A-level was spread out over two years, with the AS level is taken in the first year, followed by the A2 level in the second year. AS levels could stand as a qualification on their own or could be carried on to A2 the next year to complete the full A level qualification. More recently, this system has been reformed so that results from AS levels no longer count towards the final A level grade.

5 Unfortunately, very little information is available on the precise reasons why students are not accepted at university. To our knowledge there is no available breakdown of applications versus acceptances by school type or Polar3. Even when a student is not accepted at university, they may still reapply in the following year. 8% of those who applied in 2015 were re-applicants (UCAS 2015), though there is no information on background characteristics of re-applicants.

6 93% of students attend publicly funded secondary schools in England (Table 2A; DfE 2010).

7 The equivalent figures for Scotland, Northern Ireland and Wales are 10% v 36%, 15% v 46%, 16% v 44% respectively.

8 Further education (FE) in the United Kingdom is typically undertaken after age 16, in order to achieve a range of vocational qualifications. In contrast students may choose to continue in secondary school to achieve more academic qualifications, such as A-Levels. FE colleges may also offer HE qualifications such as teaching qualifications, or degrees.

9 Given that this measure of quality is based on universities’ published tariff points, this measure of quality may be prone to error; it is well known that institutions accept students with grades below the advertised requirement (e.g. Sellgren, 2019).

10 We apply frequency weights in all calculations.

11 It is important to note that this measure of under-over prediction is rather blunt. The measure is based on the total points achieved from the applicants’ best 3 A-levels. This means that each teacher in each subject would have to correctly predict the applicants’ grade for the total points score to be accurately predicted. There is also room for error in the measure: a student could be predicted to achieve BBB=12 points, and actually achieve AAD=12 points, and appear to be accurately predicted according to this measure.

12 In Appendix A, we present regressions with an a binary dependent variable, showing the probability that students from different backgrounds or schools are underpredicted. These results show similar patterns, whereby high achieving students from low SES backgrounds and state schools have a greater probability of being underpredicted than those from high SES backgrounds and independent schools.

13 As a robustness test we replicate the results in Table 6 in the dataset without controls, defining high tariff as Russell Group institutions, and medium and low tariff as all other institutions. We find similar results to those in Table 6, suggesting the lack of control variables may not seriously impact these coefficients.

Additional information

Funding

This work was supported by University and College Union.