The word “bibliometrics” is used to designate a set of quantitative methods of analysis of scientific publications. Every aspect of a publication that can be quantified may form the subject of a bibliometric study: the number of words in a paper, the delay time between submission and publication, etc. While the quantitative data concerning a specific paper may be rather boring, they become more interesting when comparing different publications or for a statistical study of large sets of publications.
Bibliometrics and research evaluation
During the last decades bibliometric methods have become quite fashionable for the evaluation of scientific research and for the assessment of individual researchers. The most frequently used bibliometric measures are the following:
- The number of papers published by a given researcher or research group, as an indication of his/its productivity.
- The frequency with which a published paper is cited in later publications by other researchers, as an indication of the interest this paper has raised.
- The frequency with which an electronically available paper is downloaded by readers, as an indication of its importance.
- The average frequency with which the papers in a given journal are cited during a given time span after publication, as an indication of the scientific quality of the journal or of the thoroughness of its peer review.
Advantages of bibliometric methods and drawbacks
The advantages of bibliometric methods for scientific evaluations are rather obvious:
- The methods are straightforward, since based on simple counting. Many techniques have become especially simple in the digital age, because their application can be automated.
- On first sight they are objective and unbiased.
At the same time, there are some obvious drawbacks in these methods:
- As quantitative methods they may completely miss the point for a qualitative evaluation.
- They may be manipulated (e.g., unnecessary citations by your colleagues).
- The number of citations depends more on the number of people working in the same domain, rather than on the intrinsic quality or originality of the published results.
- Since the number of citations is a driving force for evaluating researchers in their career promotion, researchers will tend to cling to “trendy research” in fields where many other researchers are active and where scientific funds can more easily be obtained. The result may be a trivialization of research subjects instead of an active search for original research ideas.
- In practice it has lead to an Anglo-Saxon bias and a strengthening of the big players in the publication sector, to the disadvantage of small publishers, of Southern countries and, e.g., Hispanic, Japanese, Chinese, Russian, Arabic and Francophone publications.
This means that one should be very careful in drawing conclusions from bibliometric methods. They are at their best as statistical methods, and therefore also prone to big errors when applied to individual cases. Even if it could be proven that there is a strong correlation between the number of citations and the scientific quality of a paper, it would be very dangerous to conclude that a paper without citations is necessarily of low scientific value.
- The most widely used bibliometric instrument is formed by the databases of Thomson Reuters (formerly ISI) with its Web of Knowledge, containing citation indices since 1900 and covering 23,000 journals, and the derived Journal of Citation Reports with statistical data such as impact parameters for more than 10,000 journals. We discuss them below in more detail. Access to these databases, however, requires a very expensive subscription.
- SCOPUS is an alternative citation database from Elsevier, covering 19,500 journals, also by subscription.
- Google Scholar offers a good and free alternative. By searching, e.g., with the author name, you not only obtain a list of publications, but for each of them the number of citations and even the link to all citing papers. Google Scholar may be a better alternative for the Social Sciences.
- CiteSeerX offers an independent free citation database for computer and information sciences.
- MESUR (MEtrics from Scholarly Usage of Resources) was a big project in which the access to e-journals was logged at a large number of US university campuses, and combined with the above mentioned bibliometric data. One of their conclusions, reported in J. Bollen et al. (2009) is that the journal impact factor is only of marginal importance and should thus be used with caution.
The Thomson Reuters databases
Since the Thomson Reuters databases and their contained indices and parameters are most widely used in Western universities and research funding agencies, we describe them here in some more detail.
The history of these databases goes back to the (paper format) citation indices published by the Institute of Scientific Information (ISI) since 1955, as started by Eugene Garfield. The original intention was not to evaluate research but to offer an instrument where researchers could discover the most recent publications by subject. It was soon realized that the citations could serve as an additional help for discovering relevant papers for specialized subjects, and the first citation index version was published in 1964 as the Science Citation Index (SCI). Two years later it became available on magnetic tape, later on CD-Rom and now – much extended with data from the Social Sciences – on the Internet.
The core of the system is still a large index database in which all papers from more than 13000 journals are registered with their full metadata and their list of citations, going back to 1955. Free format searches can be performed on title, subject, author, journal, author address and more. In this sense, it still fulfils its original role as an indexing instrument. For each of the search results, not only the full metadata are reproduced, but also the list of cited references, and the list of later papers that have cited this one. From the references, the database is extended with data about cited journals outside the core of 13000 analyzed journals and before 1955 (going back to 1900). As of March 2012 they claim to have 87 million source items, with 700 million cited references.
For the evaluation of individual researchers and research teams, the important aspect here is the number of citations received for each article.
From the data available in this large database, each year a special report is produced containing a detailed analysis of the citations per journal for the previous year, and classified by broad subject categories. We mention some of them:
- The total number of articles published in the journal during the concerned year.
- The total number of citations to each journal during the concerned year.
- The impact factor: the number of citations during the year to articles that appeared in this journal during the two previous years, divided by the number of articles that appeared in this journal during these two years. (E.g.: the impact factor of journal X for 2011 is the number of citations during 2011 to articles that appeared in journal X in 2009 and 2010, divided by the number of papers published in journal X in these same two years.) The impact factor therefore expresses the average number of times that an article (published in the previous 2 years) has been cited during the concerned year.
- The 5-year impact factor: the same as above, but for 5 years instead of 2. (This is important for subject fields that evolve more slowly.)
- The immediacy index: number of citations in the previous year to articles in the journal published in the same year, divided by the number of articles published in this year.
- Journal cited half-life: In order to calculate this half-life all citations during the year concerned are counted by year of publication of the cited paper, and the half-life is set such that there are as many citations to papers before that time span as to papers after that time span. (A short half-life indicates that on average publications in this journal may be cited well, but that over time these citations rapidly diminish.)
At present journal impact factors (JIF) are playing a very important role in research evaluation procedures in Western countries, in spite of the fact that they are widely believed to be overestimated. Because of this JIF many researchers continue to publish in overly expensive commercial journals and they often neglect more suitable Open Access publication channels. What are the drawbacks of the impact factors?
- In absolute terms, a JIF reflects for a large part the number of people working in a given field. The highest JIF attributed for 2010 to a journal in oncology was 94.3, whereas the highest in ornithology was 2.3. To conclude that the research quality in ornithology is so much lower than that in cancer research would be preposterous: there are clearly more people active in the latter domain. For this reason, the ranking lists of journals per category are more important than the individual JIF values.
- In the same way, it is easier to obtain a high JIF for a journal in a domain in full expansion than in a domain where scientific evolution has reached a point of quiet maturity. This should not discourage researchers from working in this more quiet field, where still important scientific work may be done, even if this will not lead to a high number of citations.
- By its definition, a JIF is an average value for all papers published during 2 (or 5) years in the given journal, and does not guarantee the quality of an individual paper.
We refer to the References for further reflexions about the value of this and other bibliometric evaluation parameters. It should be clear that they must be treated with care: they may be welcome as additional information, but they can never replace a good qualitative evaluation of the work performed by a researcher or research group (Georges Stoops, 2009).