ANALYSIS: Colleges with more women get more firsts

Paper based on Tompkins Table data proposes “reducing the proportion of male students”.

academics tripos women

An analysis of Tompkins Table data has found evidence to suggest that “a greater proportion of women in fact increases a college’s performance”.

The analysis was drawn up by N. Hulbert (Gonville and Caius 2012) and D. Hulbert (Gonville and Caius 1969).

In a press release, they wrote: “Following Priscilla’s announcement last term that CUSU would formally oppose the publication of the Tompkins Table, we thought it would be worthwhile demonstrating some of the interesting trends that come out when the data is properly analysed”.

The paper provides “some evidence that women at Cambridge underperform in Tripos not because of bias against their sex or gender, but because the colleges which take a large fraction of the women are worse performers”. They suggest this could be because of a “bias” against female colleges, possibly by fellows.

When colleges with a high proportion of women and those originally founded to admit women, as well as Trinity and John’s, are excluded, though, the data show a shallow positive correlation between the proportion of female undergraduates and the percentage of firsts awarded.

Men: Just Not That Great.

One of the radical suggestions the paper makes is reducing the proportion of male students to improve academic performance.

In 2015, the all-women Lucy Cavendish languished at the bottom of the Tompkins Table.

In 2015, the all-women Lucy Cavendish languished at the bottom of the Tompkins Table.

They note, however, that their models explain only 60% of variation across colleges and would be more precise if their analysis took into account the STEM proportion of each college.

“The evidence this data provides towards claiming that female underperformance in Tripos is due to longstanding bias against female majority colleges rather than female candidates is both important and contentious enough to deserve further investigation – in particular, examination of the performance of similar female cohorts at different colleges would shed light on this, as would a breakdown by subject”.

Other provocative insights from the data include:

  • It is false to claim that fundraising campaigns improve academic performance. Having roughly £200,000 in the endowment per undergraduate member of the college predicts only 1% more firsts across the college.
  • Overall performance might be improved by consolidating smaller, richer colleges with larger, poorer ones.
  • Distance from St Mary’s in central Cambridge is positively correlated with academic performance.
Caius caught in the anti-intellectual shadow of Great St Mary's.

Caius caught in the anti-intellectual shadow of Great St Mary’s.

They argued that “students and applicants should have access to this information to enable them to reach their own conclusions, as well as to place pressure on underperforming colleges to improve”.

Rob Cashman, CUSU’s Education Officer, responded to The Tab’s request for comment:

“This research highlights the complexity of assessing differences across the Colleges. CUSU will continue to campaign for greater parity across Colleges, and for all students to have equal opportunity to succeed. We recognise the value of Colleges working together collaboratively, rather than in direct competition with one another. This research acknowledges the limitations of the Tompkins Table, which provides only a ‘slice’ of the information in a retrospective manner, and highlights the importance of a more holistic approach that Colleges should take when assessing their provision.

“We advocate further investigation into the performance of different women cohorts across Colleges. It is notable that this research does not address or consider other attainment gaps, including the BME attainment gap.

“We thank the authors for this research which supports the work CUSU is doing to achieve greater parity across Colleges.”

The University of Cambridge declined to comment.

You can request the full paper, including its methodology and sources, by emailing Nicholas Hulbert (CRSID: nmh42).