As preparations for the next research excellence framework gather momentum, I fear I’m in a dwindling minority who finds it discombobulating to hear colleagues confidently asserting that they have a handful of 3* or 4* papers, and looking down upon those with only 2*s.
Colleagues, the difference between a 2*, 3* and 4* is based on subjective judgements of vague criteria. My articles have been ranked internally as 3* or 4*, so this isn’t sour grapes, but there really is not a knowable distinction between – as Main Panel C puts it – “quality that is internationally excellent” (3*) and “quality that is recognised internationally” (2*). The latter implies the former.
Nor is there a meaningful difference between research that is a “major influence” (4*) and research that is “likely to have a lasting influence” (3*). Ditto. And even if there were, the judgement is going to be made by a REF panellist who is unlikely to be an expert in your sub-field and who will probably skim-read your article over breakfast on a flight.
That, of course, has always been the problem with the REF. But back in 1986, when the first forerunner exercise was run, it was excusable to think that the only way of splitting research funding between universities on the basis of some sort of demonstrable merit was through some bureaucratic process of this sort, however imperfect.
But, hey, times have moved on. There are now numerous internet-based data sources on academics' research performance – and many of them are free. Senior academics have used some of these to come up with rankings of universities and departments that are extremely closely correlated to those produced from the REF.
For instance, in 2017, Anne-Wil Harzing, professor of international management at Middlesex University, found only small differences between the REF rankings and those created using data from Microsoft Academic. Memorably, doing so took her just “one rainy Sunday afternoon”. And, in 2013, Dorothy Bishop, professor of developmental neuropsychology at the University of Oxford, looked at the data from the REF’s precursor, the research assessment exercise, and found that departmental h-indices in psychology predicted the results “remarkably well”. She suggested this may be true more broadly, too.
Meanwhile, Marcus Munafo, professor of biological psychology at the University of Bristol, found in 2015 that a “prediction market” closely mirrored REF outcomes for chemistry departments. Prediction markets arrive at the probability of an outcome occurring based on individuals betting on what they believe the outcome will be.
I’ve also had a go, with my colleague Barrie Craven. We found that university rankings compiled using ResearchGate, Google Scholar and Webometrics (which creates scores based on “link analysis”, looking at each university’s presence and impact on the web) were, again, extremely closely correlated with REF rankings compiled by both Times Higher Education (based on quality) and Research Fortnight (based on quality and volume).
Importantly, two of the approaches described, prediction markets and Webometrics, have nothing to do with citation indices. These sometimes get a bad press from those, especially in the humanities, who are reluctant to let go of the REF; in the sciences, by contrast, the case is more accepted that citation by colleagues who presumably are experts in the relevant field is a better mark of quality than the approval of stressed, non-expert REF panellists. Either way, it is hard to endorse the conclusion of the 2015 government-sponsored independent review that subjective judgement based on ambiguous criteria remains “the least worst form of academic governance we have” in the 21st century.
Let’s spell this out. The REF delivers data extremely slowly and infrequently, at great expense (the official estimate is £246 million) and with huge disruption to university life, resulting in rankings very similar but, arguably, inferior to those obtained simply and cheaply using a range of methods that don’t disrupt anyone.
Clearly the government is not going to replace the REF any time soon. An elephantine beast like this develops a life and purpose of its own, and loyalty to match. But there is a clear market opportunity for a sympathetic thinktank to create parallel league tables using the alternative, freely available resources. Because there are many of these, it would be easy to experiment to find an optimum combination of data that can’t be gamed and that offers no perverse incentives. A handful of supervised interns could easily handle it.
Regularly updated tables will be much more attractive to consumers of higher education – students and funders – than the quickly stale and out-of-date REF rankings. Hence, as Friedrich Engels might have put it, the demand for the REF will wither away. The interference of state power in research excellence will become superfluous. Universities will cease to see the need to participate. And, with that, the minister’s pen will easily do the needful and consign the REF to history.
All that universities would then need to do would be to make sure that their academics published high-quality research articles. Then they could stand back and let the private sector do the heavy lifting.
James Tooley is professor of education policy at Newcastle University.