Most people in the UK higher education sector probably want the research excellence framework to deliver both generous funding for world-leading research and reasonable funding for the broad base that underpins it.

In practice, however, the formula that turns REF scores into allocations of ¡°quality-related¡± (QR) funding already puts most of its emphasis on work that is rated 4* (¡°world-leading in terms of originality, significance and rigour¡±). That focus ¨C which sounds good to politicians ¨C is only likely to intensify further. So the incentive for universities to maximise their submission of 4* work is absolute. And the increased flexibility in the rules for 2021 gives them greater ability to do?so.

Instead of four outputs being required for every researcher selected for submission, the draft , published in July, requires universities to submit a minimum of one and a maximum of five for every research-active staff member. This will allow institutions to make more selections from the strongest publication portfolios.

Moreover, the required average of 2.5 outputs per full-time equivalent staff member is not demanding for research-intensive institutions. A humanities department in which most staff produced one 4*?monograph (double-weighted in assessment) plus one 4*?article over the seven-year period of assessment would enjoy a near-100?per cent 4*?outcome.

³ÉÈËVRÊÓƵ

Of course, some individuals will produce fewer outputs, or outputs that do not fit neatly within a unit of assessment, and it will not be possible for departments to ¡°hide¡± them completely this time around by not submitting them. But most elite institutions will not struggle to find one excellent output per researcher over seven years ¨C and there are mechanisms for taking into account extenuating individual circumstances that make even that impossible.

Grade-point averages approaching the maximum 4.0 are all the more likely given that assessors are likely to interpret the definition of ¡°world-leading¡± broadly. While directing the course of a whole discipline is a rare achievement, doing so in a tightly defined subfield might be realistic. Since assessors¡¯ judgements affect their discipline¡¯s relative standing, they have an incentive to adopt such flexible criteria.

³ÉÈËVRÊÓƵ

The result is likely to be that QR funding becomes even more concentrated in research-intensives than it is already. There are caveats. Small units with a mixed portfolio might struggle to present all their research as world-leading even on a generous interpretation. And elite research units might not produce 4*?impact as consistently as 4*?outputs. But, in most cases, they only need one impact case study for every 15 researchers.

While a glut of GPAs approaching 4.0 might sound great for those institutions awarded such scores, drawing a tight boundary between the research elites and the rest would have obvious drawbacks for institutions that do excellent research with an applied focus, or only in pockets. It would also fail to distinguish meaningfully within that elite, focusing on tiny differences in GPA that are unlikely to reflect material differences in quality but that would nevertheless occupy inordinate amounts of institutional attention.

Such an outcome also risks rendering invisible the broad base of research that exists even at leading institutions, making it appear that all researchers can consistently produce genuinely world-leading work. This will only serve to ratchet up the pressure on individuals further, while doing scholarship a disservice by failing to value more incremental or locally focused work.

So what is the solution? Increasing the volume requirement would increase the burden on assessors and researchers alike. But if the REF is to continue, it should adopt the broader grading scale that the funders originally advocated. That would have included an extra 5*?category that could have been reserved for world-leading research, strictly defined.

³ÉÈËVRÊÓƵ

Perhaps that would only have resulted in further grade inflation, but this could be policed by strict enforcement of standards through the auditing of panels awarding excessive numbers of top grades. A less punitive but probably more effective measure would be to adjust the funding formula to better reward research graded?4* or even?3*.

That way, we could have a less elitist set of assessment outcomes and a healthier research culture.

Dominic Dean is a research quality officer at the University of Sussex. He writes in a personal capacity.