Schooling effects on degree performance: a comparison of the predictive validity of aptitude testing and secondary school grades at Oxford University

Abstract

This article examines the cause of school type effects upon gaining a first class degree at Oxford University, whereby for a given level of secondary school performance, private school students perform less well at degree level. We compare the predictive power of an aptitude test and secondary school grades (GCSEs) for final examination performance, using data from the Oxford Admissions Study. Both metrics are predictive of final degree performance but the school effects are only statistically robust for arts students. Private school students perform less well in final examinations relative to their GCSE results when compared with state school students, but they do not under perform relative to their aptitude test scores or in gross terms. It is therefore argued that teaching effects, associated with private school students, distort secondary school grades as an indicator of academic potential in higher education when compared to state school students.

Notes

1. The Russell Group is composed of the UK's elite universities, similar to the Ivy League in the USA.

2. For example, if comparing only the achievement of first class degrees with achieving a 2:1, a given group may seem to do well on this metric whilst also be more likely to achieve less than a 2:1. Using only a first/2:1 comparison would therefore give a misleading impression of examination performance. The same issue applies, of course, to dropping out from university altogether. Complete multinomial regression tables are available from the authors upon request.

3. The correlation between the proportion of firsts awarded and the proportion of lower than 2:1s awarded across Honour Schools is only 0.10. The results of regressions for the likelihood of gaining less than a 2:1 should therefore be regarded as highly provisional.

4. Model 2 also suggests that a portion of the female protection against lower than a 2:1 is the result of gender segregation by subject.