Elsevier

by

Bibliometrics, global rankings, and transparency

Why do we care so much about the actual and potential uses of bibliometrics (“the generic term for data about publications,” according to the OECD), and world university ranking methodologies, but care so little about the private sector firms, and their inter-firm relations, that drive the bibliometrics/global rankings agenda forward?

This question came to mind when I was reading the 17 June 2010 issue of Nature magazine, which includes a detailed assessment of various aspects of bibliometrics, including the value of “science metrics” to assess aspects of the impact of research output (e.g., publications) as well as “individual scientific achievement”.

The Nature special issue, especially Richard Van Noorden’s survey on the “rapidly evolving ecosystem” of [biblio]metrics, is well worth a read. Even though bibliometrics can be a problematic and fraught dimension of academic life, they are rapidly becoming an accepted dimension of the governance (broadly defined) of higher education and research. Bibliometrics are generating a diverse and increasingly deep impact regarding the governance process at a range of scales, from the individual (a key focus of the Nature special issue) through to the unit/department, the university, the discipline/field, the national, the regional, and the global.

Now while the development process of this “eco-system” is rapidly changing, and a plethora of innovations are occurring regarding how different disciplines/fields should or should not utilize bibliometrics to better understand the nature and impact of knowledge production and dissemination, it is interesting to stand back and think about the non-state actors producing, for profit, this form of technology that meshes remarkably well with our contemporary audit culture.

In today’s entry, I’ve got two main points to make, before concluding with some questions to consider.

First, it seems to me that there is a disproportionate amount of research being conducted on the uses and abuses of metrics in contrast to research on who the producers of these metrics are, how these firms and their inter-firm relations operate, and how they attempt to influence the nature of academic practice around the world.

Now, I am not seeking to imply that firms such as Elsevier (producer of Scopus), Thomson Reuters (producer of the ISI Web of Knowledge), and Google (producer of Google Scholar), are necessarily generating negative impacts (see, for example, ‘Regional content expansion in Web of Science®: opening borders to exploration’, a good news news story from Thomson Reuters that we happily sought out), but I want to make the point that there is a glaring disjuncture between the volume of research conducted on bibliometrics versus research on these firms (the bibliometricians), and how these technologies are brought to life and to market. For example, a search of Thomson Reuter’s ISI Web of Knowledge for terms like Scopus, Thomson Reuters, Web of Science and bibliometrics generates a nearly endless list of articles comparing the main data bases, the innovations associated with them, and so on, but amazingly little research on Elsevier or Thomson Reuters (i.e. the firms).  From thick to thin, indeed, and somewhat analogous to the lack of substantial research available on ratings agencies such as Moody’s or Standard and Poor’s.

Second, and on a related note, the role of firms such as Elsevier and Thomson Reuters, not to mention QS Quacquarelli Symonds Ltd, and TSL Education Ltd, in fueling the global rankings phenomenon has received remarkably little attention in contrast to vigorous debates about methodologies. For example, the four main global ranking schemes, past and present:

all draw from the databases provided by Thomson Reuters and Elsevier.

One of the interesting aspects of the involvement of these firms with the rankings phenomenon is that they have helped to create a normalized expectation that rankings happen once per year, even though there is no clear (and certainly not stated) logic for such a frequency. Why not every 3-4 years, for example, perhaps in alignment with the World Cup or the Olympics? I can understand why rankings have to happen more frequently than the US’ long-delayed National Research Council (NRC) scheme, and they certainly need to happen more frequently than the years France wins the World Cup championship title (sorry…) but why rank every single year?

But, let’s think about this issue with the firms in mind versus the pros and cons of the methodologies in mind.

From a firm perspective, the annual cycle arguably needs to become normalized for it is a mechanism to extract freely provided data out of universities. This data is clearly used to rank but is also used to feed into the development of ancillary services and benchmarking capabilities that can be sold back to universities, funding councils, foundations, regional organizations (e.g., the European Commission which is intensely involved in benchmarking and now bankrolling a European ranking scheme), and the like.

QS Quacquarelli Symonds Ltd, for example, was marketing such services (see an extract, above, from a brochure) at their stand at the recent NAFSA conference in Kansas City, while Thomson Reuters has been busy developing what they deem the Global Institutional Profiles Project. This latter project is being spearheaded by Jonathon Adams, a former Leeds University staff member who established a private firm (Evidence Ltd) in the early 1990s that rode the UK’s Research Assessment Excellence (RAE) and European ERA waves before being acquired by Thomson Reuters in January 2009.

Sophisticated on-line data entry portals (see a screen grab of one above) are also being created. These portals build a free-flow (at least one one-way) pipeline between the administrative offices of hundreds of universities around the world and the firms doing the ranking.

Data demands are becoming very resource consuming for universities. For example, the QS template currently being dealt with by universities around the world shows 14 main categories with sub-categories for each: all together there are 60 data fields, of which 10 are critical to the QS ranking exercise, to be launched in October 2010. Path dependency dynamics clearly exist for once the pipelines are laid the complexity of data requests can be gradually ramped up.

A key objective, then, seems to involve using annual global rankings to update fee-generating databases, not to mention boost intra-firm knowledge bases and capabilities (for consultancies), all operational at the global scale.

In closing, is the posited disjuncture between research on bibliometrics vs research on bibliometricians and the information service firms these units are embedded within worth noting and doing something about?

Second, what is the rationale for annual rankings versus a more measured rankings window, in a temporal sense? Indeed why not synchronize all global rankings to specific years (e.g., 2010, 2014, 2018) so as to reduce strains on universities vis a vis the provision of data, and enable timely comparisons between competing schemes. A more measured pace would arguably reflect the actual pace of change within our higher education institutions versus the needs of these private firms.

And third, are firms like Thomson Reuters and Elsevier, as well as their partners (esp., QS Quacquarelli Symonds Ltd and TSL Education Ltd), being as transparent as they should be about the nature of their operations? Perhaps it would be useful to have accessible disclosures/discussions about:

  • What happens with all of the data that universities freely provide?
  • What is stipulated in the contracts between teams of rankers (e.g., Times Higher Education and Thomson Reuters)?
  • What rights do universities have regarding the open examination and use of all of the data and associated analyses created on the basis of the data universities originally provided?
  • Who should be governing, or at least observing, the relationship between these firms and the world’s universities? Is this relationship best continued on a bilateral firm to university basis? Or is the current approach inadequate? If it is perceived to be inadequate, should other types of actors be brought into the picture at the national scale (e.g., the US Department of Education or national associations of universities), the regional-scale (e.g., the European University Association), and/or the global scale (e.g., the International Association of Universities)?

In short, is it not time that the transparency agenda the world’s universities are being subjected to also be applied to the private sector firms that are driving the bibliometrics/global rankings agenda forward?

Kris Olds

by

Developments in world institutional rankings; SCImago joins the club

Editor’s note: this guest entry was kindly written by Gavin Moodie, principal policy adviser of Griffith University in Australia.  Gavin (pictured to the right) is most interested in the relations between vocational and higher education. His book From Vocational to Higher Education: An International Perspective was published by McGraw-Hill last year. Gavin’s entry sheds light on a new ranking initiative that needs to be situated within the broad wave of contemporary rankings – and bibliometrics more generally – that are being used to analyze, legitimize, critique, promote, not to mention extract revenue from.  Our thanks to Gavin for the illuminating contribution below.

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

It has been a busy time for world institutional rankings watchers recently. Shanghai Jiao Tong University’s Institute of Higher Education published its academic ranking of world universities (ARWU) for 2009. The institute’s 2009 rankings include its by now familiar ranking of 500 institutions’ overall performance and the top 100 institutions in each of five broad fields: natural sciences and mathematics, engineering/technology and computer sciences, life and agriculture sciences, clinical medicine and pharmacy, and social sciences. This year Dr. Liu and his colleagues have added rankings of the top 100 institutions in each of five subjects: mathematics, physics, chemistry, computer science and economics/business.

Times Higher Education announced that over the next few months it will develop a new method for its world university rankings which in future will be produced with Thomson Reuters. Thomson Reuters’ contribution will be guided by Jonathan Adams (Adams’ firm, Evidence Ltd, was recently acquired by Thomson Reuters).

And a new ranking has been published, SCImago institutions rankings: 2009 world report. This is a league table of research institutions by various factors derived from Scopus, the database of the huge multinational publisher Elsevier. SCImago’s institutional research rank is distinctive in including with higher education institutions government research organisations such as France’s Centre National de la Recherche Scientifique, health organisations such as hospitals, and private and other organisations. Only higher education institutions are considered here. The ranking was produced by the SCImago Research Group, a Spain-based research network “dedicated to information analysis, representation and retrieval by means of visualisation techniques”.

SCImago’s rank is very useful in not cutting off at the top 200 or 500 universities, but in including all organisations with more than 100 publications indexed in Scopus in 2007. It therefore includes 1,527 higher education institutions in 83 countries. But even so, it is highly selective, including only 16% of the world’s estimated 9,760 universities, 76% of US doctoral granting universities, 65% of UK universities and 45% of Canada’s universities. In contrast all of New Zealand’s universities and 92% of Australia’s universities are listed in SCImago’s rank. Some 38 countries have seven or more universities in the rank.

SCImago derives five measures from the Scopus database: total outputs, cites per document (which are heavily influenced by field of research as well as research quality), international collaboration, normalised Scimago journal rank and normalised citations per output. This discussion will concentrate on total outputs and normalised citations per output.

Together these measures show that countries have been following two broad paths to supporting their research universities. One group of countries in northern continental Europe around Germany have supported a reasonably even development of their research universities, while another group of countries influenced by the UK and the US have developed their research universities much more unevenly. Both seem to be successful in support research volume and quality, at least as measured by publications and citations.

Volume of publications

Because a reasonable number of countries have several higher education institutions listed in SCImago’s rank it is possible to consider countries’ performance rather than concentrate on individual institutions as the smaller ranks encourage. I do this by taking the average of the performance of each country’s universities. The first measure of interest is the number of publications each university has indexed in Scopus over the five years from 2003 to 2007, which is an indicator of the volume of research. The graph in figure 1 shows the mean number of outputs for each country’s higher education research institutions. It shows only countries which have more than six universities included in SCImago’s rank, which leaves out 44 countries and thus much of the tail in institutions’ performance.

Figure 1: mean of universities’ outputs for each country with > 6 universities ranked


These data are given in table 1. The first column gives the number of higher education institutions each country has ranked in SCImago institutions rankings (SIR): 2009 world report. The second column shows the mean number of outputs indexed in Scopus for each country’s higher education research institutions from 2003 to 2007. The next column shows the standard deviation of the number of outputs for each country’s research university.

The third column in table 1 shows the coefficient of variation, which is the standard deviation divided by the mean and multiplied by 100. This is a measure of the evenness of the distribution of outputs amongst each country’s universities. Thus, the five countries whose universities had the highest average number of outputs indexed in Scopus from 2003 to 2007 – the Netherlands, Israel, Belgium, Denmark and Sweden – also had a reasonably low coefficient of variation below 80. This indicates that research volume is spread reasonably evenly amongst those countries’ universities. In contrast, Canada which had the sixth highest average number of outputs also has a reasonably high coefficient of variation of 120, indicating an uneven distribution of outputs amongst Canada’s research universities.

The final column in table 1 shows the mean of SCImago’s international collaboration score, which is a score of the proportions of the institution’s outputs jointly authored with someone from another country. The US’ international collaboration is rather low because US authors collaborate more often with authors in other institutions within the country.

Table 1: countries with > 6 institutions ranked by institutions’ mean outputs, 2007

Source: SCImago Research Group (2009) SCImago institutions rankings (SIR): 2009 world report.

Citations per paper by field

We next examine citations per paper by field of research, which is an indicator of the quality of research. This is the ratio between the average citations per publication of an institution and the world number of citations per publication over the same time frame and subject area. SCImago says it computed this ratio using the method established by Sweden’s Karolinska Intitutet which it called the ‘Item oriented field normalized citation score average’. A score of 0.8 means the institution is cited 20% below average and 1.3 means the institution is cited 30% above average.

Figure 2 shows mean normalised citations per paper for each country’s higher education research institutions from 2003 to 2007, again showing only countries which have more than six universities included in SCImago’s rank. The graph for an indicator of research quality in figure 2 is similar in shape to the graph of research volume in figure 1.

Figure 2: mean of universities’ normalised citations per paper for each country with > 6 universities ranked

Table 2 shows countries with more than six higher education research institutions ranked by their institutions’ mean normalised citations. This measure distinguishes more sharply between institutions than volume of outputs – the coefficient of variations for countries’ mean institutions normalised citations are higher than for number of publications. Nonetheless, several countries with high mean normalised citations have an even performance amongst their universities on this measure – Switzerland, Netherlands, Sweden, Germany, Austria, France, Finland and New Zealand.

Finally, I wondered whether countries which had a reasonably even performance of their research universities by volume and quality of publications reflected a more equal society. To test this I obtained from the Central Intelligence Agency’s (2009) World Factbook the Gini index of the distribution of family income within a country. A country with a Gini index of 0 would have perfect equality in the distribution of family income whereas a country with perfect inequality in its distribution of family would have a Gini index of 100. There is a modest correlation of 0.37 between a country’s Gini index and its coefficient of variation for both publications and citations.

Table 2: countries with > 6 institutions ranked by institutions’ normalised citations per output

Sources: SCImago Research Group (2009) SCImago institutions rankings (SIR): 2009 world report; Central Intelligence Agency (2009) The world factbook.

Conclusion

SCImago’s institutions research rank is sufficiently comprehensive to support comparisons between countries’ research higher education institutions. It finds two patterns amongst countries whose research universities have a high average volume and quality of research publications. One group of countries has a fairly even performance of their research universities, presumably because they have had fairly even levels of government support. This group is in northern continental Europe and includes Switzerland, Germany, Sweden, the Netherlands, Austria, Denmark and Finland. The other group of countries also has a high average volume and quality of research publications, but spread much more unevenly between universities. This group includes the US, the UK and Canada.

This finding is influenced by the measure I chose to examine countries’ performance, the average of their research universities’ performance. Other results may have been found using another measure of countries’ performance, such as the number of universities a country has in the top 100 or 500 of research universities normalised by gross domestic product. But such a measure would not reflect a country’s overall performance of their research universities, but only the performance of its champions. Whether one is interested in a country’s overall performance or just the performance of its champions depends on whether one believes more benefit is gained from a few outstanding performers or several excellent performers. That would usefully be the subject of another study.

Gavin Moodie

References

Central Intelligence Agency (2009) The world factbook (accessed 29 October 2009).

SCImago institutions rankings (SIR): 2009 world report (revised edition accessed 20 October 2009).