Back this time last year I fired up Tableau to search for "Britain's Worst Higher Education Courses". I wasn't able to find them in Unistats (our only course-level public dataset) https://wonkhe.com/blogs/britains-worst-higher-education-courses/

You can identify subject areas (note not courses) where graduates have historically seen poor salary returns. Nursing is one. Social Work another. History another.

But are these returns the fault of the subject?

You can also identify providers where graduates see low salary returns. But prior attainment, and multiple deprivations, and sex all have an impact on lifetime earnings. And some providers recruit disproportionally from these groups. Should they stop?

Region is another complicating factor. If you are a provider in, say, the North East - your graduates are likely to earn less than their peers in London. Should the University of Teeside move to Islington?

And there are no provider or subject specific data points that even mildly correlate with LEO data. If the difference is down to provider and subject choice only, what about the provider is causing it?

And then we deal with history. LEO data is from 2017-18, at latest, referring to graduates from the middle of the decade. Do the same courses exist at the same providers five years on? Are they taught the same way, by the same people?

Courses with low salaries aren't predicted by entry requirements either. Or by applicant demand. Or by any metric of "employer demand" that I've come across. Or student satisfaction. Or VC salary, for that matter.