Strong case study in a large submission could bring in £2 million, says Simon Kerridge
When research impact was introduced in the first Research Excellence Framework in 2014, it was partly to address concerns that universities were not visibly contributing to the UK economy and wider society.
That first REF allocated 20 per cent of the total mark to impact. Case studies accounted for 80 per cent of this, so 16 per cent of the total. The remaining 4 per cent depended on the approach to and support for impact.
The latter was generally perceived as, shall we say, less useful. So for REF 2021, when impact share rose to 25 per cent, assessment was based entirely on case studies. This meant that each was worth significantly more than in 2014.
The initial decisions for REF 2028, published on 15 June, shows that what is now called impact and engagement retains its 25 per cent slice. But as well as case studies, each submission will have to include a “disciplinary level evidence statement”. This will account for between 20 per cent, for the largest submissions, and 50 per cent, for the smallest, of the impact slice.
So case studies will surely be worth less next time? Well, as with all things REF, it depends.
Shifting weights
While the reintroduction of a more narrative element to impact has diluted the contribution of case studies, the weighting of outputs has also fallen, from 65 per cent in REF 2014 to 60 per cent in REF 2021; it is down to 45 per cent for REF 2028, allowing for the mark allotted to explanatory statements for each unit of assessment.
So how much is a case study worth, compared with a research output, and should institutions continue to support them?
As with most things, size matters. In REF 2028, the relative value of a case study increases as submission size rises, so the easiest way to compare them to outputs is to chart them over changing submission sizes. This line zigzags because, as I have previously reported, the value of a case study dips sharply when the threshold for an additional impact case study is reached.
The take-home message is that impact still deserves a lot of attention. For submissions above around 30 full-time equivalents in REF 2028, a case study will be worth at least 10 times a research output. When you get to 150 FTE or so, it will be 20 times an output.
What also jumps out is that the difference between REFs grows as submission size increases. For submissions of around 50 full time equivalents or more, impact case studies will be worth still more relative to outputs than they were for REF 2021. Compared with REF 2014, almost every submission to REF 2028 will see the worth of case studies relative to outputs increase.
Pervasive effect
To put this in context, in REF 2021, a 4-star output was worth around £10,000 a year in quality-related funding, with most falling between £5,000 and £25,000. So an impact case study from a large submission in 2028 could easily be worth £250,000 a year until the next REF—close to £2 million in total.
As the graph also shows, most submissions (54 per cent in REF 2021) are smaller than 30 FTE and only a few (21 per cent) are larger than 50 FTE. However, these larger submissions account for 53 per cent of the academic population, and those in submissions over 30 FTE make up 78 per cent of the population. So changing the value ratio of case studies to outputs will have a pervasive effect.
Impact is clearly worth continued investment from institutions—and particularly for those with large submissions.
Simon Kerridge is an independent research consultant and honorary staff at the University of Kent. He was a panel adviser to main panel C for the 2021 Research Excellence Framework
A version of this article appeared in Research Fortnight
Correction 28/6 – This article originally stated that weighting of outputs would make up 50 per cent of the mark in REF 2028. In fact, this section of the REF, contributions to knowledge and understanding, will also include assessment of an explanatory statement that will account for at least 5 per cent of the total mark. The text and graph have been corrected to account for this. SK would like to thank Laura Bates at the University of St Andrews for pointing this out.