Impact is widely reported as the success story of the REF 2014 exercise. Stern recommends that consistency should be maintained where possible with the 2014 impact assessment process and the number of impact case studies required should remain roughly equivalent to total numbers assessed in 2014. However, it would seem that many of the proposed modifications to the REF process currently under consultation could undermine these principles:

Required number of impact case studies

The potential requirement to return all ‘research-active’ staff may result in institutions that returned a smaller cohort of staff to REF 2014 seeing an increase in the total numbers of impact case studies required for 2021, even if the case study to staff ratio is reduced. A very crude calculation based on REF 2014 UoA structure, 60% staff submission and a minimum of one case study for smaller units would realise a small increase in the number of case studies required in 2021.

In addition, if the minimum requirement at UoA level is reduced to one impact case study for smaller units and the impact template element is moved to the environment submission, the rating received for those single case studies would be made openly available once ratings are announced. Institutions with a number of smaller UoAs would be open to a considerable number of individual case study ratings being made publicly available. If the one case study minimum was observed in REF 2014, the University of Portsmouth would have submitted one case study in 7 units of assessment.

These scenarios also do not take into account any overall increase in staff numbers over time or potential increase in the range of UoAs in which we may be required to return.

HESA mapping to UoAs

The requirement to submit in UoAs mapped to HESA return could see impact case studies required for UoAs not previously returned and areas where pockets of research are carried out, this has the potential to ‘water down’ the excellent research reported through impact case studies. Forcing research into discrete silos undermines the greater emphasis being placed on inter/multidisciplinary research.

Institutional level impact case studies

Institutional case studies could potentially top-slice the best examples of impact from the total institutional pool of excellent research. A 10-20% institutional case studies requirement could mean Portsmouth would need up to 10 case studies in this element (based on the number of required case studies in 2014). These ‘best’ examples of impact will be used for individual UoAs without the institutional requirement; it can therefore be questioned whether institutional level impact case studies are even required as an element in the REF process.

It seems counterproductive to introduce a new institutional partition part way through the REF cycle. This will put additional burden on institutional selection and assessment processes that are already in place, and potentially require new processes to recognise input and rewarding mechanisms.

Limits to number of resubmission of case studies returned in 2014

As with most institutions, many of our case studies returned in 2014 are being considered for resubmission in 2021 as excellent research and subsequent impact is continuing to be developed in these areas. Any limits on the numbers of resubmissions could dilute continued excellence being reported through the REF process.

The question as to what constitutes a resubmission also needs to be addressed; if new research or new streams of impact have been realised would this be classed as a resubmission or could it be articulated as a new case study?

Impact template as part of environment submission

As an element of the environment submission the impact template requirements may become more metricised and the broad range of research support mechanisms may be lost without the provision to provide a supporting narrative.


Dee Summers is Research Impact Officer for the University of Portsmouth and writes here in a personal capacity. If you would like to talk about impact please contact me: