ÖйúAƬ

In his first major policy speech as universities and science minister, Jo Johnson spoke about the importance of graduate employability.

The idea that you could use this as an indicator of the educational quality of a university is superficially attractive but deeply problematic. A recent predecessor of Johnson¡¯s, David Willetts, had read evidence about which indicators of quality were valid and useful, and which were not, and understood the difficulty of using any ¡°outcome¡± measure.

Outcome measures of all kinds ¨C whether degree classifications, retention rates or employability measures ¨C are strongly influenced by a raft of other variables that tell us nothing useful about institutional quality. The confounding variables that are most relevant to employability measures are the quality of the students themselves ¨C which varies hugely between institutions independently of their competence at developing students¡¯ employability ¨C and the reputation of institutions, which employers are influenced by but which predicts nothing at all about how much students have learned.

Currently available employability data are also much too short-term and unreliable to interpret, let alone use as a crude metric. I support both evidence-based policy and the use of quality evidence to direct funding. However, you would have to use valid and reliable data about a university¡¯s achievements in making whatever students they have more employable than they otherwise would have been, independently of institutional reputation.

No such data exist. Using current employability data will simply reinforce traditional reputation-based hierarchies. Perhaps that is what this policy is really about.

Graham Gibbs is former director of the Oxford Learning Institute at the University of Oxford.