Thursday, September 25, 2014

How the Universities of Huddersfield, East London, Plymouth, Salford, Central Lancashire et cetera helped Cambridge overtake Harvard in the QS rankings

It is a cause of pride for the great and the good of British higher education that the country's universities  do brilliantly in certain global rankings. Sometimes though, there is puzzlement about how UK universities can do so well even though the  performance of the national economy  and the level of adult cognitive skills are so mediocre.

In the latest QS World University Rankings Cambridge and Imperial College London pulled off a spectacular feat when they moved ahead of Harvard into joint second place behind MIT, an achievement at first glance as remarkable as Leicester City beating Manchester United. Is this a tribute to the outstanding quality of teaching, inspired leadership or cutting edge research, or perhaps something else?

Neither Cambridge nor Imperial does very well in the research based rankings. Cambridge is 18th and Imperial 26th among higher education institutions in the latest Scimago rankings for output and 32nd and 33rd for normalised impact (citations per paper adjusted for field). Harvard is 1st and 4th for these indicators. In the CWTS Leiden Ranking, Cambridge is 22nd and Imperial 32nd for the mean normalised citation score, sometimes regarded as the flagship of these rankings, while Harvard is 6th.

It is true that Cambridge does much better on the Shanghai Academic Ranking of World Universities with fifth place overall, but that is in large measure due to an excellent score, 96.6, for alumni winning Nobel and Fields awards, some dating back several decades. For Highly Cited Researchers and publications in Nature and Science its performance is not nearly so good.

Looking at the THE World University Rankings, which make some attempt to measure factors other than research, Cambridge and Imperial come in 7th and 10th overall, which is much better than they do in the Leiden and Scimago rankings. However, it is very likely that the postgraduate teaching and research surveys made a significant contribution to this performance. Cambridge is 4th in the THE reputation rankings based on last year's data and Imperial is 13th.

Reputation is also a key to the success of Cambridge and Imperial in the QS world rankings. Take a look at the scores and positions of Harvard, Cambridge and Imperial in the rankings just released.

Harvard  gets 100 points (2nd place) for the academic survey, employer survey (3rd), and citations per faculty (3rd). It has 99.7 for faculty student ratio (29th), 98.1 for international faculty (53rd), and 83.8 for international students (117th). Harvard's big weakness is its relatively small percentage of international students.

Cambridge is in first place for the academic survey and 2nd in the employer survey, in both cases with a score of 100 and one place ahead of Harvard. The first secret of Cambridge's success is that it does much better on reputational measures than for bibliometric or other objective data. It was 18th for faculty student ratio, 73rd for international faculty, 50th for international students and 40th for citations per faculty.

So, Cambridge is ahead for faculty student ratio and international students and Harvard is ahead for international faculty and citations per faculty. Both get 100 for the two surveys.

Similarly, Imperial has 99.9 points for the academic survey (14th), 100 for the employer survey (7th), 99.8 for faculty student ratio (26th), 100 for international faculty (41st), 99.7 (20th) for international students and 96.2 (49th) for citations per faculty. It is behind Harvard for citations per faculty but just enough ahead for international students to squeeze past into joint second place.

The second secret is that QS's standardisation procedure combined with an expanding database means that the scores of the leading universities in the rankings are getting more and more squashed together at the top. QS turns its raw data into Z scores so that universities are measured according to their distance in standard deviations from the mean for all ranked universities. If the number of sub-elite universities in the rankings increases then the overall means for the indicators will fall and the scores of universities at the top end will rise as their distance in standard deviations from the mean increases.

Universities with scores of 98 and 99 will now start getting scores of 100. Universities with recorded scores of 100 will go on getting 100, although they might go up up a few invisible decimal points

In 2008, QS ranked 617 universities. In that year, nine universities had a score of 100 for the academic survey, four for the employer survey, nine for faculty student ratio, six for international faculty, six for international students and seven for citations per faculty.

By 2014 QS was ranking over 830 universities (I assume that those at the end of the rankings marked "NA" are there because they got votes in the surveys but are not ranked because they fail to meet the criteria for inclusion). For each indicator the number of universities getting a score of 100 increased. In 2014 there were 13 universities with a score of 100 for the academic survey, 14 for the employer survey, 16 for faculty student ratio, 41 for international faculty, 15 for international students and 10 for citations per faculty,

In 2008 Harvard got the same score as Cambridge for the academic and employer surveys. It was 0.3 (0.06 weighted) behind for faculty student ratio, 0.6 (0.53 weighted) behind for international faculty, and 14.1 (0.705 weighted) behind for international students, It was, however, 11.5 points. (2.3 weighted) ahead for citations per faculty. Harvard was therefore first and Cambridge third.

By 2014 Cambridge had fallen slightly behind Harvard for international faculty. It was slightly ahead for faculty student ratio. Scores for the survey remained the same, 100 for both places. Harvard reduced the gap for international students slightly.

What made the difference in 2014 and put Cambridge ahead of Harvard was that in 2008 Harvard  in fifth place for citations and with a score 100 was 11.5 (2.3 weighted) points ahead of Cambridge. In 2014 Cambridge had improved a bit for this indicator -- it was 40th instead of 49th -- but now got 97.9 points reducing the difference with Harvard to 2.1 points (0.42 weighted). That was just enough to let Cambridge overtake Harvard.

Cambridge's rise between 2008 and 2014 was thus largely due to the increasing number of ranked universities which led to lower means for  each indicator which led to higher Z scores at the top of each indicator and so reduced the effect of Cambridge's comparatively lower citations per faculty score.

The same thing happened to Imperial . It did a bit better for citations, rising from 58th to 49th place and this brought it a rise in points from 83.10 to 96.20 again allowing it to creep past Harvard.

Cambridge and Harvard should be grateful to those universities filling up the 701+ category at the bottom of the QS rankings. They are the invisible trampoline that propelled "Impbridge" into second place, just behind MIT.

QS should think carefully about adding more universities to their rankings. Another couple of hundred and there will be a dozen universities at the top getting 100 for everything.









No comments: