A Mystery Solved

One of the more interesting elements in the Guide to the World's Top Universities by John O'Leary, Nunzio Quacquarelli and Martin Ince, published by QS Quacquarelli Symonds at the end of 2006, is the information about student faculty ratio provided in the directory of over 500 universities and the profiles of the world's top 100 universities.

These are, even at first sight, not plausible: 590.30 students per faculty at Pretoria, 43.30 at Colorado State University, 18.10 at Harvard, 3.50 at Dublin Institute of Technology.

Scepticism is increased when the the Guide's data for student faculty ratio is correlated with that derived from the scores out of 100 for this measure in the 2006 rankings and cross-checked with the data on individual universities on QS's topuniversities site. The correlation for 517 universities is negligible at .057 and statistically insignificant (2-tailed .195).

Comparing the two sets of data on student faculty ratio for the British universities in the rankings shows that the problem is with the information in the Guide, not that in the rankings. The rankings data correlates highly with that provided by the Higher Education Statistics Agency (HESA: see earlier post) (.712, sig = .000) and that taken from the web site williseemy tutor (.812, sig = .000). There is no significant correlation between the data in the Guide and the HESA data (.133, sig = .389) and that derived from williseemytutor (.179, sig = .250).

So, where did the Guide's student faculty data come from?

First, here are the most favourable student faculty ratios calculated from the scores in the rankings (they can be cross-checked at the topuniversities site) and rounded to the first decimal place.

Duke 3.5

Yale 3.7

Eindhoven University of Technology 3.8

Rochester 3.8

London Imperial College 3.9

Paris Sciences Po 4.0

Tsing Hua 4.1

Emory 4.1

Geneva 4.3

Vanderbilt 4.3


Now, here are the most favourable ratios given in the Guide.

Dublin Institute of Technology 3.5

Wollongong 3.7

Ecole Polytechnique 3.8

Rio de Janeiro 3.8

Llubljanja 3.9

Oulu 4.0

Trento 4.1

Edinburgh 4.1

Fudan 4.3

Utrecht 4.3


Notice that the ratio of 3.5 is assigned to Duke university in the rankings and to Dublin IT in the Guide. If the universities are arranged alphabetically these two would be in adjacent rows. Likewise, the other scores listed above are assigned to universities that would be next to each other or nearly so in an alphabetical listing.

Next are the least favourable ratios derived from the rankings data.

Pune 580

Delhi 316

Tor Vergata 53

Bologna 51

Cairo 49

Concordia 42


Now the ratios in the Guide.

Pretoria 590

De La Salle 319

RMIT 53

Bilkent 51

Bucharest 49

Colorado 42

Notice again that, except for Tor Vergata and RMIT, the ratio in the two data sets is shared by universities that are close or next to each other alphabetically.

The conclusion is unavoidable. When the Guide was being prepared somebody created a new file and made a mistake, going down one or two or a few rows and inserting the rankings data in the wrong rows. So, every university in the Guide's directory acquired a new and erroneous student faculty ratio.

Since this piece of information is the one most likely to interest future undergraduate students, this is not a trivial error.

Is this error any less serious than QS's getting the two North Carolina business schools mixed up?