Fewer students are dropping out of universities than ever before, according to new data.
But there are huge differences between subjects, with more than one in 10 computer science students dropping out, compared with fewer than two in 100 taking medicine and dentistry.
The latest data from the Higher Education Funding Council for England show that in 2011-12, 6.6 per cent of full-time UK students doing a first degree in England had quit after their first year.
This is almost one percentage point less than the previous year, and is the latest in a series of declines since 2003-04, when the dropout rate was 9.2 per cent.
But differences in dropout rates between subjects remain stark. Eleven per cent of computer science students dropped out in 2011-12, according to the data. This is an improvement on the previous year, but since 2003-04 dropout rates have generally stayed above 12 per cent, despite almost all other subject areas having seen significant improvements in retention.
A detailed breakdown of the figures shows that software engineering has a particularly poor retention record, with nearly 17 per cent of students dropping out after the first year. Artificial intelligence courses, on the other hand, do much better.
Digital Skills for Tomorrow’s World, a report released earlier this month by the UK Digital Skills Taskforce, suggested that computer science courses are “extremely varied” and that “some students arrive at university to find that the courses do not match their expectations”.
The data also show that men (7.6 per cent) are more likely to drop out than women (5.9 per cent). Students from areas with the lowest levels of participation in higher education also had higher dropout rates than those from other neighbourhoods. Neither of these differences could be fully explained when controlling for age, subject and qualifications on entry.
Black students and those from state school also have higher than average dropout rates, but these differences did largely disappear when other factors were taken into account.