ABSTRACT
Usually, scientific production is measured by indicators based on two types of data: the number of publications, mainly in refereed international scientific journals, articles in collective books, proceedings and books; and the citations received by published articles. Studying the use of citations needs a database that includes the references of articles; only two bibliographic multidisciplinary databases offer this: Thomson Reuters’ Web of Science (WoS) or Elsevier’s Scopus. These two databases were created for bibliometric purposes. Other specialized databases or other large multidisciplinary databases (for example, the French PASCAL) are usually not used in bibliometric analysis – with some exceptions. These two databases provide a “stable” reference required in order to do comparative analyses. But none of these databases explain, in any explicit way, their choices of journals. In a time when the whole publishing system is changing, this comes as a real difficulty. Moreover, both the WoS and Scopus have some strong biases. It is not very difficult to identify some of these biases that are well known, such as the preference for English, the very poor representation of journals produced outside the United States and the very US-centric set of journals in the social sciences. It is more difficult to measure the effect of these biases and that comes as a real difficulty when doing statistical analyses. The WoS, since it is older, created by the promoters of bibliometrics, is supposedly more stable than Scopus, although it has been shown that both deliver similar results. This holds true for large countries only, and for statistical analyses of fields with a large number of publications (because of the statistical size of the sample of publications). It is probably not true for the analysis of country-specific data, when these countries have small rates of scholarly production; that is, practically all Arabic countries. Another issue with the databases is their coverage of certain fields. It appears that they underestimate certain domains, while tending to overestimate others: for instance, the biomedical sciences are better covered by the WoS. Books and other forms of publication, more frequent in the social sciences and humanities, are poorly covered by both databases (although they attempt to at least partially cover edited books). Every small journal in the United States is covered in the Social Sciences Citation Index (in WoS), but many widely circulated journals outside the United States are missing; this is true for publications from Europe
as well as from other countries and regions. The underlying model of WoS and Scopus is a commercial world centered in the United States, a model that has been challenged even for so-called “hard” sciences.1
