CRELL: critiquing global university rankings and their methodologies

This guest entry has been kindly prepared for us by Beatrice d’Hombres and Michaela Saisana of the EU-funded Centre for Research on Lifelong Learning (CRELL) and Joint Research Centre. This entry is part of a series on the processes and politics of global university rankings (see here,  here, here and here).

Since 2006, Beatrice d’Hombres has been working in the Unit of Econometrics and Statistics of the Joint Research Centre of  the European Commission. She is part of the Centre for Research on Lifelong Learning. Beatrice is an economist who completed a PhD at the University of Auvergne (France). She has a particular expertise in education economics and applied econometrics.

Michaela Saisana works for the Joint Research Centre (JRC) of the European Commission at the Unit of Econometrics and Applied Statistics. She has a PhD in Chemical Engineering and in 2004 she won the European Commission – JRC Young Scientist Prize in Statistics and Econometrics for her contribution on the robustness assessment of composite indicators and her work on sensitivity analysis.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

The expansion of the access to higher education, the growing mobility of students, the need for economic rationale behind the allocation of public funds, together with the demand for higher accountability and transparency, have all contributed to raise the need for comparing university quality across countries.

The recognition of this fact has also been greatly stirred  by the publication, since 2003, of the ‘Shanghai Jiao Tong University Academic Ranking of World Universities’ (henceforth SJTU), which measures university research performance across the world. The SJTU ranking tends to reinforce the evidence that the US is well ahead of Europe in terms of cutting-edge university research.

Its rival is the ranking computed annually, since 2004, by the Times Higher Education Supplement (henceforth THES). Both these rankings are now receiving worldwide attention and constitute an occasion for national governments to comment on the relative performances of their national universities.

In France, for example, the publication of the SJTU is always associated with a surge of articles in newspapers which either bemoan  the poor performance of French universities or denounce the inadequacy of the SJTU ranking to properly assess the attractiveness of the fragmented French higher education institutions landscape (see Les Echos, 7 August 2008).

Whether the intention of the rankers or not, university rankings have followed a destiny of their own and are used by national policy makers to stimulate debates about national university systems and ultimately can lead to specific education policies orientations.

At the same time, however, these rankings are subject to a plethora of criticism. They outline that the chosen indicators are mainly based on research performance with no attempt to take into account the others missions of universities (in particular teaching), and are biased towards large, English-speaking and hard-science institutions. Whilst the limitations of the indicators underlying the THES or the SJTU rankings have been extensively discussed in the relevant literature, there has been no attempt so far to examine in depth the volatility of the university ranks to the methodological assumptions made in compiling the rankings.

The purpose of the JRC/Centre for Research on Lifelong Learning (CRELL) report is to fill in this gap by quantifying how much university rankings depend on the methodology and to reveal whether the Shanghai ranking serves the purposes it is used for, and if its immediate European alternative, the British THES, can do better.

To that end, we carry out a thorough uncertainty and sensitivity analysis of the 2007 SJTU and THES rankings under a plurality of scenarios in which we activate simultaneously different sources of uncertainty. The sources cover a wide spectrum of methodological assumptions (set of selected indicators, weighting scheme, and aggregation method).

This implies that we deviate from the classic approach – also taken in the two university ranking systems – to build a composite indicator by a simple weighted summation of indicators. Subsequently, a frequency matrix of the university ranks is calculated across the different simulations. Such a multi-modeling approach and the presentation of the frequency matrix, rather than the single ranks, allows one to deal with the criticism, often made to league tables and rankings systems ,that ranks are presented as if they were calculated under conditions of certainty while this is rarely the case. 

The main findings of the report are the following. Both rankings are only robust in the identification of the top 15 performers on either side of the Atlantic, but unreliable on the exact ordering of all other institutes. And, even when combining all twelve indicators in a single framework, the space of the inference is too wide for about 50 universities of the 88 universities we studied and thus no meaningful rank can be estimated for those universities. Finally, the JRC report suggests that THES and SJTU rankings should be improved along two main directions:

• first, the compilation of university rankings should always be accompanied by a robustness analysis based on a multi-modeling approach. We believe that this could constitute an additional recommendation to be added to the already 16 existing Berlin Principles;
• second, it is necessary to revisit the set of indicators, so as to enrich it with other dimensions that are crucial to assessing university performance and which are currently missing.

Beatrice d’Hombres  and Michaela Saisana