A whole lotta cheatin’ going on? REF stats revisited

 

1.

The rankings produced by Times Higher Education and others on the basis of the UK’s Research Assessment Exercises (RAEs) have always been contentious, but accusations of universities’ gaming submissions and spinning results have been more widespread in REF2014 than any earlier RAE. Laurie Taylor’s jibe in The Poppletonian that “a grand total of 32 vice-chancellors have reportedly boasted in internal emails that their university has become a top 10 UK university based on the recent results of the REF”[1] rings true in a world in which Cardiff University can truthfully[2] claim that it “has leapt to 5th in the Research Excellence Framework (REF) based on the quality of our research, a meteoric rise” from 22nd in RAE2008. Cardiff ranks 5th among universities in the REF2014 “Table of Excellence,” which is based on the GPA of the scores assigned by the REF’s “expert panels” to the three elements in each university’s submission (outputs 65%, impact 20%, environment 15%)—just behind Imperial, LSE, Oxford and Cambridge. Whether this “confirms [Cardiff’s] place as a world-leading university,” as its website claims, is more questionable.[3] These figures are a minefield.

Although HEFCE encouraged universities to be “inclusive” in entering their staff in REF2014, they were not obliged to return all eligible staff and there were good reasons for those with aspirations to climb the league tables to be more “strategic” in staff selection than in previous RAEs. Prominent among these were (1) HEFCE’s defunding of 2* outputs from 2011, which meant outputs scoring below 3* would now negatively affect a university’s rank order without any compensating gain in QR income, and (2) HEFCE’s pegging the number of impact case studies required to the number of staff members entered per unit of assessment, which created a perverse incentive to exclude research-active staff if this would avoid having to submit a weak impact case study.[4] Though the wholesale exclusions feared by some did not materialize across the sector, it is clear that some institutions were far more selective in REF2014 than in RAE2008.

Unfortunately data that would have permitted direct comparisons with numbers of staff entered by individual universities in RAE2008 were never published, but Higher Education Statistical Authority (HESA) figures for FTE staff eligible to be submitted allow broad comparisons across universities in REF2014. It is evident from these that selectivity, rather than an improvement in research quality per se, played a large part in Cardiff’s “meteoric rise” in the rankings. The same may be true for some other schools that significantly improved their positions, among them Kings (up to 7th in 2014 from 22= in 2008), Bath (14= from 20=), Swansea (22= from 56=), Cranfield (31= from 49), Heriot-Watt (33 from 45), and Aston (35= from 52=). All of these universities except Kings entered fewer than 75% of their eligible staff members, and Kings has the lowest percentage (80%) of any university in the REF top 10 other than Cardiff itself.

Cardiff achieved its improbable rank of 5th on the basis of a submission that included only 62% of eligible staff. This is the second-lowest percentage of any of the 28 British universities that are listed in the top 200 in the 2014-15 Times Higher Education World University Rankings (of these schools only Aberdeen entered fewer staff, submitting 52%). No other university in this cohort submitted less than 70% of eligible staff, and half (14 universities) submitted over 80%. Among the top schools, Cambridge entered 95% of eligible staff, Imperial 92%, UCL 91% and Oxford 87%.

Many have suggested that “research power” (which is calculated by multiplying the institution’s overall rounded GPA by the total number of full-time equivalent staff it submitted to the REF) gives a fairer indication of a university’s place in the national research hierarchy than GPA rankings alone. By this measure, Cardiff falls to a more credible but still respectable 18th. But when measured by “research intensity” (that is, GPA multiplied by the percentage of eligible staff entered), its rank plummets from 5th to 50th. To say that this provides a more accurate indication of its true standing might be overstating the case, but it certainly underlines why Cardiff does not belong among “world-leading” universities.  Cardiff doubtless produces some excellent research, but its overall (and per capita) performance does not remotely justify comparisons with Oxford, Cambridge, or Imperial—let alone Caltech, Harvard, Stanford, Princeton, MIT, UC-Berkeley and Yale (the other universities in the THE World University Rankings top 10).  In this sense the GPA Table of Excellence can be profoundly misleading.

“To their critics,” writes Paul Jump in Times Higher Education, “such institutions are in essence cheating because in reality their quality score reflects the work produced by only a small proportion of their staff.”[5] I am not sure the accusation of cheating is warranted, because nobody is doing anything here that is outside HEFCE’s rules. The problem is rather that the current REF system rewards—and thereby encourages—bad behavior, while doing nothing to penalize the most egregious offenders like Cardiff.

The VCs at Bristol (11= in the REF2014 GPA table) and Southampton (18=, down from 14= in 2008) might be forgiven for ruefully reflecting that they, too, might now be boasting that they are “a top ten research university” had they not chosen to submit 91% and 90% of their eligible faculty respectively—a submission rate that on any reasonable criteria (as distinct from HEFCE’s rules) should itself be regarded as a mark of research excellence. Measured by research intensity Bristol comes in at 5= (jointly with Oxford) and Southampton 8= (jointly with Queen’s University Belfast, which submitted 95% of its staff and is ranked 42= on GPA). Meantime the VCs at St Andrews (down from 14= to 21=, 82% of eligible staff submitted), Essex (11th to 35=, 82% submitted), Loughborough (28= to 49=, 88% submitted) and Kent (31= to 49=, 85% submitted) may by now have concluded that—assuming they hold onto their jobs—they will have no alternative other than to be much more ruthless in culling staff for any future REF.

2.

The latest Times Higher Education World University Rankings puts Cardiff just outside the top 200, in the 201-225 group—which places it 29= among UK universities, along with Dundee, Newcastle, and Reading. Taking GPA, research power and research intensity into account—as we surely should, in recognition that not only the quality of research outputs but the number and proportion of academic staff who are producing them are also necessary elements in evaluating any university’s overall contribution to the UK’s research landscape—such a ranking seems intuitively to be just about right.

I have shown elsewhere[6] that there was, in fact, a striking degree of overall agreement between the RAE2008 rankings and the Times Higher Education World University Rankings. Repeating the comparison for UK universities ranked in the top 200 in the THE World University Rankings for 2014-15 and the REF2014 GPA-based “Table of Excellence” yields similar findings. The data are summarized in Table 1.

 

Table 1: REF2014 performance of universities ranked in the top 200 in Times Higher Education World University Rankings 2014-15

University THE World University Rankings 2014-15 REF2014 ranking by GPA (RAE2008) REF2014 ranking by research power REF2014 ranking by research intensity Percentage of eligible staff submitted
1-50
Oxford 3 4 (4=) 2 5= 87
Cambridge 5 5 (2) 3 2 95
Imperial 9 2 (6) 8 3 92
UCL 22 8= (7) 1 4 91
LSE 34 3 (4=) 28 7 85
Edinburgh 36 11= (12) 4 12= 83
Kings 40 7 (22=) 6 17 80
50-100
Manchester 52 17 (8) 5 26= 78
Bristol 74 11= (14) 9 5= 91
Durham 83 20 (14=) 20 24= 79
Glasgow 94 24 (33=) 12 15 84
100-150
Warwick 103 8= (9) 15 11 83
QMUL 107 11= (13) 22 34= 74
St Andrews 111= 21= (14=) 22 16 82
Sussex 111= 40 (30) 34 42= 73
York 113 14= (10) 23 32 75
Royal Holloway 118 26= (24=) 40 31 77
Sheffield 121 14= (14=) 13 33 74
Lancaster 131 18= (20=) 26 29 77
Southampton 132 18= (14=) 11 8= 90
Leeds 146 21= (14=) 10 34= 75
Birmingham 148 31 (26) 14 23 81
150-200
Exeter 154 30 (28=) 21 19= 82
Liverpool 157 33 (40) 19 46= 70
Nottingham 171 26= (24=) 7 28 79
Aberdeen 178 46= (38) 29 57 52
UEA 198 23 (35=) 36 37 75
Leicester 199 53 (51) 24 39 78

Seven UK universities make the top 50 in the 2014-15 THE World University Rankings: Oxford, Cambridge, Imperial, UCL, LSE, Edinburgh, and Kings. Six of these are also in the REF2014 top 10, while the other (Edinburgh) is only just outside it at 11=. Four of the leading five institutions are same in both rankings (the exception being UCL, which is 8= in REF 2014), though not in the same rank order. Of 11 UK universities in THE top 100, only one is outside the REF top 20 (Glasgow, at 24th). Of 22 UK universities in THE top 150, only two (Birmingham, 31 in REF, and Sussex, 40 in REF) are outside REF top 30. Of the 28 UK universities in THE top 200, only two (Aberdeen at 46= and Leicester at 53) rank outside the REF top 40.

Conversely, only two universities in the REF2014 top 20, Cardiff at 6 and Bath at 14=, do not make it into the THE top 200 (their respective ranks are 201-225 and 301-350). Other universities that are ranked in the top 40 in REF2014 but remain outside the THE top 200 are Newcastle (26=), Swansea (26=), Cranfield (31), Herriot-Watt (33), Essex (35=), Aston (35=), Strathclyde (37), Dundee (38=) and Reading (38=).

Table 2 provides data on the performance of selected UK universities that submitted to REF2014 but are currently ranked outside the THE world top 200.

Table 2. REF2014 performance of selected UK universities outside top 200 in Times Higher Education World University Rankings 2014-15

University THE World University Rankings 2014-15 REF2014 ranking by GPA (RAE 2008) REF2014 ranking by research power REF2014 ranking by research intensity Percentage of eligible staff submitted
Cardiff 201-225 6 (22=) 18 50 62
Dundee 201-225 38= (40=) 39 49 68
Newcastle 201-225 26= (27) 16 26= 80
Reading 201-225 38= (42) 27 19= 83
Birkbeck 226-250 46= (33=) 48 30 81
Plymouth 276-300 66= (75=) 47 59 50
Bath 301-350 14= (20=) 35 34= 74
Bangor 301-350 42= (52=) 59 51 63
Essex 301-350 35= (11) 45 22 82
Aberystwyth 350-400 58= (45=) 51 46= 76
Aston 350-400 35= (52=) 69 60 43
Portsmouth 350-400 65 (68=) 55 80 27
Swansea 26= (52=) 42 42= 71
Cranfield 31= (49) 61 64 37
Heriot-Watt 33 (45) 44 53 57

Dundee, Newcastle and Reading only just miss the THE cut (they are all in the 201-225 bracket). While all three outscored Aberdeen and Leicester, who are above them in the THE rankings (in Leicester’s case, at 199, very marginally so) in the REF, only Newcastle does substantially worse in the THE rankings than in the REF. It is ranked 26= in the REF with Nottingham and Royal Holloway, ahead of Leicester (53), Aberdeen (46), Sussex (40), Liverpool (33), Birmingham (31) and Exeter (30)—all of which are in the top 200 in the THE World Rankings. While there was a yawning gulf between Essex’s RAE2008 ranking of 11th and its THE ranking in the 301-350 group, the latter does seem to have presaged its precipitous REF2014 fall from grace to 35=. Conversely, the THE’s inclusion of Plymouth in the 276-300 group of universities places it considerably higher than its RAE rank of 66= would lead us to expect. This is not the case with most of the UK universities listed in the lower half of the THE top 400. Birkbeck, Bangor, Aberystwyth and Portsmouth also all found themselves outside the top 40 in REF2014.

The greatest discrepancies between REF2014 and the THE World Rankings come with Cardiff (6 in REF, 201-225 in REF), Bath (14= in REF, 301-350 in THE), Swansea (26= in REF, not among THE top 400), Aston (35= in REF, 350-400 in THE), Cranfield and Heriot-Watt (31= and 33 respectively in REF, yet not among THE top 400). On the face of it, these cases flatly contradict any claim that THE (or other similar) rankings are remotely accurate predictors of REF performance. I would argue, on the contrary, that these are the exceptions that prove the rule. All these schools were prominent among those universities identified above who inflated their GPA by submitting smaller percentages of their eligible staff in REF2014. Were we to adjust raw GPA figures by research intensity, we would get a much closer match, as Table 3 shows.

Table 3. Comparison of selected universities performance in THE World University Rankings 2014-15 and REF2014 by GPA and research intensity.

University THE 2014-15 REF2014 intensity REF2014 GPA
Cardiff 201-225 50 6
Bath 301-350 34= 14=
Swansea 42= 26=
Aston 350-400 60 35=
Cranfield 64 31=
Heriot-Watt 53 33

The most important general conclusion to emerge from this discussion is that despite some outliers there is a remarkable degree of agreement between the top 40 in REF2014 and the top 200 in the THE 2014-15 World University Rankings, and the correlation increases the higher we go in the tables. Where there are major discrepancies, these are usually explained by selective staff submission policies.

One other correlation is worth noting at this point. All 11 of the British universities in the THE top 100 are members of the Russell Group, as are 10 of the 17 British universities ranked between 100-200. The other six universities in this latter cohort (St Andrews, Sussex, Royal Holloway, Lancaster, UEA, Leicester) were all members of the now-defunct 1994 Group. Only one British university in the THE top 200 (Aberdeen) belonged to neither the Russell Group nor the 1994 Group. Conversely, only two Russell Group universities, Newcastle and Queen’s University Belfast, did not make the top 200 in the THE rankings.[7] In 2013-14 Russell Group and former 1994 Group universities between them received almost 85% of QR funding. Here, too, an enormous amount of money, time, and acrimony seems to have been expended on a laborious REF exercise that merely confirms what THE rankings have already shown.

3.

The most interesting thing about this comparative exercise is that the Times Higher Education World University Rankings not only make no use of RAE/REF data, but rely on quantitative methodologies that have repeatedly been rejected by the British academic establishment in favor of the “expert peer review” that is supposedly offered by REF panels. THE gives 30% of the overall score for the learning environment, 7.5% for international outlook, and 2.5% for industry income. The remaining 60% is based entirely on research-related measures, of which “the single most influential of the 13 indicators,” counting for 30% of the overall THE score, is “the number of times a university’s published work is cited by scholars globally” as measured by the Web of Science. The rest of the research score is derived from research income (6%), ‘research output scaled against staff numbers’ (6%, also established through the Web of Science), and ‘a university’s reputation for research excellence among its peers, based on the 10,000-plus responses to our annual academic reputation survey’ (18%).

The comparisons undertaken here strongly suggest that such metrics-based measures have proved highly reliable predictors of performance in REF2014—just as they did in previous RAEs. To be sure, there are differences in the fine details of the order of ranking of institutions between the THE and REF, but in such cases can we be confident that it is the REF panels’ highly subjective judgments of quality that are the more accurate? To suggest there is no margin for error in tables where the difference in GPA between 11th (Edinburgh, 3.18) and 30th (Exeter, 3.08) is a mere 0.1 points would be ridiculous. I have elsewhere suggested (here and here) that there are in fact many reasons why such confidence would be totally misplaced, including lack of specialist expertise among panel members and lack of time for reading outputs in the depth required.[8] But my main point here is this.

If metrics-based measures can produce similar results to those arrived at through the REF’s infinitely more costly, laborious and time-consuming process of ‘expert review’ of individual outputs, there is a compelling reason to go with the metrics; not because it is necessarily a valid measure of anything but because it as reliable as the alternative (whose validity is no less dubious for different reasons) and a good deal more cost-efficient. The benefits for collegiality and staff morale of universities not having to decide who to enter or exclude from the REF might be seen as an additional reason for favoring metrics. I am sure that if HEFCE put their minds to it they could come up with a more sophisticated basket of metrics than Times Higher Education, which would be capable of meeting many of the standard objections to quantification.  Supplementing the Web of Science with Publish or Perish or other citation indices that capture books as well as articles might be a start.  I hope James Wilsdon’s committee will come up with some useful suggestions for ways forward.

[1] Laurie Taylor, “We have bragging rights!” in The Poppletonian, Times Higher Education, 8 January 2015.

[2] Well, not quite. Cardiff is actually ranked 6th in the REF2014 “Table of Excellence,” which is constructed by Times Higher Education on the basis of the grade point average (GPA) of the marks awarded by REF panels, but the #1 spot is held not by a university but the Institute of Cancer Research (which submitted only two UoAs). This table and others drawn upon here for “research power” and “research intensity” are all compiled by Times Higher Education.

[3] “REF 2014,” Cardiff University website at http://www.cardiff.ac.uk/research/impact-and-innovation/quality-and-performance/ref-2014

[4] Paul Jump, “Careers at risk after case studies ‘game playing’, REF study suggests.” Times Higher Education, 22 January 2015.

[5] Paul Jump, “REF 2014 rerun: who are the ‘game players’?” Times Higher Education, 1 January 2015.

[6] See Derek Sayer, Rank Hypocrisies: The Insult of the REF. London: Sage, 2014.

[7] I have discussed Newcastle already. Queen’s came in just outside the REF top 40 (42=) but with an excellent intensity rating (8=, 95% of eligible staff submitted).

[8] See, apart from Rank Hypocrisies, my articles “One scholar’s crusade against the REF,” Times Higher Education, 11 December, 34-6; “Time to abandon the gold standard? Peer Review for the REF Falls Far Short of Internationally Acceptable Standards,” LSE Impact of Social Sciences blog, 19 November (reprinted as “Problems with peer review for the REF,” CDBU blog, 21 November).


3 responses to “A whole lotta cheatin’ going on? REF stats revisited

  • telescoper

    Reblogged this on In the Dark and commented:
    Here’s a scathing analysis of Research Excellence Framework. I don’t agree with many of the points raised and will explain why in a subsequent post (if and when I get the time), but I reblogging it here in the hope that it will provoke some comments either here or on the original post (also a wordpress site).

  • jamym

    There is also the outright lies, some departments and University put members of staff forward who never really worked for them. A lot of contracts to people who never actually turned up,but got paid so they could be submitted

  • Ninth Level Ireland » Blog Archive » A whole lotta cheatin’ going on? REF stats revisited

    […] “The rankings produced by Times Higher Education and others on the basis of the UK’s Research Assessment Exercises (RAEs) have always been contentious, but accusations of universities’ gaming submissions and spinning results have been more widespread in REF2014 than any earlier RAE …” (more) […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: