(The paper below was submitted to Scientometrics in August 2007. A revised version was submitted in September 2007. On September 25 I received notice that the paper was accepted. I announced it in the blog and emailed a few reprints to people who requested it. However, on October 12 I received an email from the editor Tibor Braun, that the paper was rejected. This was the first time that I heard of something like this ever happening. To me it seemed strange and rather inappropriate, not to say unethical on the part of a journal editor. Of course I appealed the decision, but to no avail. In the end I gave up. I am not a professional bibliometrics jock anyway.)
Ranking of universities has lately received considerable attention. However, ranking of departments would give a higher resolution picture of the distribution of quality within each university. In this work the Hirsch (h) index of each faculty in Greek Chemistry, Chemical Engineering, Materials Science, and Physics departments was calculated using the Web of Science and the mean value was used to rank them. This ranking refers to the research performance of each department and thus is most relevant to its doctoral program. The results seem highly meaningful. If performed on a pan-European basis, such rankings could spur healthy competition and could provide a strong motive for meritocratic hiring practices. Technical difficulties and possible extension of this approach to social science and humanities departments is discussed.
University rankings have attracted considerable interest over the past few years. Rankings of this type originated in the United States. For over 25 years US News and World Report has been publishing annual rankings of research universities, liberal arts colleges and professional programs. The National Research Council has also been publishing rankings of doctoral programs every 10 years or so; the latest one is in progress. Worldwide interest was kindled by the work at Shanghai Jiao Tong University (http://ed.sjtu.edu.cn), that in 2003 started publishing annual rankings of world universities. Similar efforts are taking place in Europe, by the London Sunday Times, the Center for Science and Technology Studies at the University of Leiden (http://www.cwts.nl), the German Center for Higher Education Development, etc. While the difficulties of university rankings have been discussed 1, these efforts will likely persist and amplify. A recent comparison of currently available university rankings can be found in Ref. 2.
Ranking a university gives an overall picture of its quality. However, many universities are quite heterogeneous, containing excellent as well as mediocre departments. University assessment fails to give proper credit to those pockets of excellence. Thus, ranking of individual departments is a worthwhile endeavor. Such efforts are still quite rare. An attempt to compare two Greek Mathematics departments using bibliometric indicators was published in 1991 3.
Due to its simplicity and meaningfulness, Hirsch’s h-index 4 has created quite a stir in the scientific community. It has been used to rank information scientists 5, 6, Spanish biologists 7, business scholars 8, and chemists (http://www.rsc.org/chemistryworld/News/2007/April/ 23040701.asp). It has been found to correlate well with peer judgment 9. Self-citation corrections have been proposed 10, although another study found no pressing need for doing so 11. A number of extensions and improvements of the h index have already been proposed 12-17.
Here we propose ranking university departments using the mean h index of their faculty. The approach is applied to Chemical Engineering, Chemistry, Materials Science, and Physics departments in Greece. This study has been conducted over the last year and a half in the context of a grass roots effort to promote reform of the Greek higher education system (GreekUniversityReform.org). We found that the mean h index correlates well with qualitative perceptions of the quality of each department and could thus be used in large-scale studies over the entire European continent.
The Web of Science (WoS) was used for all data, except for the astronomers in the Physics departments for whom the ADS was used (publication and citation numbers in ADS are somewhat higher than in the WoS). The dates of the searches were: April 2006 for Chemical Engineering, October 2006 for Materials Science, May 2007 for Chemistry, and August-September 2007 for Physics. We recorded all hits for a given name, including abstracts and reviews. The h index is defined as follows 4: “A scientist has index h if h of his or her Np papers have at least h citations each and the other (Np – h) papers have h citations each.” It was calculated by sorting the hits according to times cited and counting manually or using the «Citation report» feature of the more recent version of the WoS. Common challenges were encountered: the transliteration of Greek names, names shared by more than one person, names spelled in different ways, etc. The preliminary data were sent to the departments involved and faculty members were asked to send corrections. All data are available in Excel files on the web (http://GreekUniversityReform.org/axio.html).
Tables 1-4 show the results for the departments of Chemical Engineering, Materials Science, Chemistry, and Physics.
Table 1. Ranking of Greek Chemical Engineering departments based on the mean h index. For comparison, data are given for one of the top Chemical Engineering departments in the USA, at the Univ. of Wisconsin. P is the total number of publications, and h is the Hirsch h index.
1. Univ. Patras 29 61 14.7
2. Univ. Thessaloniki 35 37 8.8
3. Natl Tech Univ Athens 89 33 7.1
Wisconsin 18 96 19.7
Table 2. Ranking of Greek Materials Science departments based on the median h index.
1. Univ. of Crete 10 81 17.3
2. Univ. of Patras 14 43 11.3
3. Univ. Ioannina 18 29 8.3
Table 3. Ranking of Greek Chemistry departments based on the mean h index. For comparison, data are given for arguably the top Chemistry department in the world, at Harvard. P is the total number of publications, and h is the Hirsch h index.
1. Univ. Crete 23 56 16.6
2. Univ. Patras 44 61 12.6
3. Univ. Thessaloniki 100 41 10.4
4. Univ. Ioannina 61 48 10.3
5. Univ. Athens 76 33 9.0
Harvard 23 219 44.5
Table 4. Ranking of Greek Physics departments based on the mean h index. P is the total number of publications, and h is the Hirsch h index.
1. Univ. Crete 31 68 16.3
2. Univ. Ioannina 52 43 11.5
3. Univ. Athens 117 55 11.1
4. Univ. Thessaloniki 100 47 9.3
5. Univ. Patras 54 27 6.9
Numerous personal communications showed that the results are in agreement with the qualitative perception in the engineering and scientific community. It is widely accepted within the sizeable expatriate Greek Chemical Engineering community that Patras has the best Chemical Engineering department in Greece. They have striven to bring in the best people from abroad and have managed to do so to a great extent. In contrast, despite its erstwhile great reputation, the NTUA department fares badly. The main reason for this seems to be inbreeding, which is a large problem in Greece and other European countries. Over half of the faculty at NTUA obtained their PhD at that department. This is certainly not due to the lack of candidates. There is clearly favoritism towards internal candidates. Inbreeding is extensive in the Chemistry departments as well. There is very little cross-fertilization, which is unjustifiable, given the availability of an adequate number of departments.
Another interesting observation is the large variance in the number of faculty among departments. The NTUA Chemical Engineering department could be the largest department of its kind in the world. The same could be true for the Univ. of Thessaloniki Chemistry department or the Univ. Athens Physics department. This should be attributed to the desire of a department to provide public employment to its graduates and grossly inadequate oversight by the Greek Education Ministry. No one sets a rational target for the size of a department; it evolves as a result of internal pressures and political expediencies rather than educational needs. A department has no motive to limit its size, as all salaries are paid by the central state budget.
Rankings are valuable in many ways: they allow an assessment of «where we stand», they provide recognition to those who do well, they spur competition between departments, and they provide a strong motive for meritocracy in faculty hiring. The latter is especially important in many European countries, such as Greece, which are plagued by favoritism and political interference. Bibliometric department ranking is fast and cheap, especially when one uses simple measures like the h-index. With relatively little effort this activity can be extended to all of Europe and could contribute considerably to the improvement of higher education and research prowess in the continent.
The caveats of the h index have been extensively discussed 18. It is obvious that the h index ranks departments on the basis of the lifetime achievement of their faculty. If we want to rank the departments on the basis of current activity, say in the last 3 years, we need to use different measures such as impact factor-weighted publication counts. The most serious caveat of the h index in our opinion is that it does not correct for co-authorship so that scientists with many collaborations are favored. Someone who is the second or third author in numerous publications will be favored over someone who is the sole author in fewer publications. The problem is especially severe in experimental high-energy physics where papers are signed by hundreds of names and the contribution of each one is unclear. Ways to correct for this are still lacking. One idea would be to divide the number of citations of each paper by a function of the number of coauthors. For example, the functions ln(#) or f=1 + #/50, where # is the number of authors, would reduce an h index of 50 for an experimental high-energy physicist to less than 20. However, such arbitrary corrections seem unwarranted without justification, so they were not employed in this work. It was confirmed that if these fictitiously high h values were omitted the ranking would not be affected.
Another issue is that the citation patterns vary not only among different fields but also among different subfields. For example, environmental chemistry does not gather as large a number of citations as organic chemistry. If a department emphasizes such a low-citation subfield, it will be disfavored by the ranking. In addition, because the h index depends on the age of the researcher, the performance of a department will depend on the age and rank distribution of its faculty. A new department with relatively young faculty in lower ranks will be disfavored in comparison with a mature department with many full professors. For example, the majority of faculty at Harvard are established, full professors. Data on the age of the faculty in the departments studied here were not available, but there is no reason to suspect that there are large differences in age distribution in these departments.
Aside from the appropriateness of the h index for individual scientists, is the mean or the median value a good measure for the quality of a department? An alternative that has been proposed is the meta-h index, which is defined as the number of faculty in a department with h index greater or equal to that number (http://www.cs.utah.edu/~shirley/hindex/). There are two problems with the meta-h: it does not take into account the size of the department and it is not highly discerning. For example, the meta-h index of the 5 chemistry departments in Greece is 15 for Patras and Thessaloniki and 14 for Crete, Athens, and Ioannina. The meta-h index of Harvard is 18. It clearly does not work. A second alternative would be to calculate the h-index of the entire department. Such an approach has been recently taken to rank institutions in the field of conservation biology 19. The problem with this idea is that the departmental h index can be dominated by the contribution of one very productive faculty member (e.g., in the Athens Chemistry department the highest h is 45 and the next highest 22. In this case the departmental h will be close to 45). For the above reasons, taking the mean or the median h index seems to be the best idea.
Is the WoS the best database to use? It seems so, at least for Science and Engineering. Google Scholar suffers from two deficiencies: a) many important Science journals are not included (e.g. J. Chem. Phys.) b) it includes publications that are not peer reviewed, e.g. information on web sites. One issue in Physics is the preprint archive Arxiv.org. Google Scholar includes citations to and from the preprints, while WoS does not. Whether this is a blessing or a curse is open for discussion. While the WoS is appropriate for Science and Engineering, it may be rather inadequate for the humanities and social sciences. In many of these fields scholarship takes the form of books, not articles, and books are not included in the WoS. For these fields Google Scholar may be more appropriate because it gives information on books as well.
1. Van Raan, A. F. J., Fatal attraction: Conceptual and methodological problems in the ranking of universities by bibliometric methods, Scientometrics, 62 (2005) 133.
2. Buela-Casal, G.; Gutierrez-Martinez, O.; Bermudez-Sanchez, M. P.; Vadillo-Munoz, O., Comparative study of international academic rankings of universities, Scientometrics, 71 (2007) 349.
3. Zachos, G., Research Output Evaluation of 2 University Departments in Greece with the Use of Bibliometric Indicators, Scientometrics, 21 (1991) 195.
4. Hirsch, J. E., An index to quantify an individual’s scientific research output, Proceedings of the National Academy of Sciences of the United States of America, 102 (2005) 16569.
5. Cronin, B.; Meho, L., Using the h-index to rank influential information scientists, Journal of the American Society for Information Science and Technology, 57 (2006) 1275.
6. Oppenheim, C., Using the h-index to rank influential British researchers in information science and librarianship, Journal of the American Society for Information Science and Technology, 58 (2007) 297.
7. Imperial, J.; Rodriguez-Navarro, A., Usefulness of Hirsch’s h-index to evaluate scientific research in Spain, Scientometrics, 71 (2007) 271.
8. Saad, G., Exploring the h-index at the author and journal levels using bibliometric data of productive consumer scholars and business-related journals respectively, Scientometrics, 69 (2006) 117.
9. Van Raan, A. F. J., Comparison of the Hirsch-index with standard bibliometric indicators and with peer judgment for 147 chemistry research groups, Scientometrics, 67 (2006) 491.
10. Schreiber, M., Self-citation corrections for the Hirsch index, EPL, 78 (2007).
11. Thijs, B.; Glanzel, W., The influence of author self-citations on bibliometric meso-indicators. The case of European universities, Scientometrics, 66 (2005) 71.
12. Egghe, L., Theory and practise of the g-index, Scientometrics, 69 (2006) 131.
13. Egghe, L., How to improve the h-index, Scientist, 20 (2006) 15.
14. Egghe, L., Dynamic h-index: The Hirsch index in function of time, Journal of the American Society for Information Science and Technology, 58 (2007) 452.
15. Banks, M. G., An extension of the Hirsch index: Indexing scientific topics and compounds, Scientometrics, 69 (2006) 161.
16. Braun, T.; Glanzel, W.; Schubert, A., A Hirsch-type index for journals, Scientometrics, 69 (2006) 169.
17. Sidiropoulos, A.; Katsaros, D.; Manolopoulos, Y., Generalized Hirsch h-index for disclosing latent facts in citation networks, Scientometrics, 72 (2007) 253.
18. Bornmann, L.; Daniel, H. D., What do we know about the h index? Journal of the American Society for Information Science and Technology, 58 (2007) 1381.
19. Grant, J.; Olden, J.; Lawler, J.; Nelson, C.; Silliman, B., Academic Institutions in the US and Canada ranked according to research productivity in the field of Conservation Biology, Conservation Biology, on the web (2007).