CITATIONS TO SCIENTIFIC PAPERS

This paper presents some preliminary evidence about the existence of power laws representing the upper tail of the citation distributions among 221 scientific sub-fields or Web of Science categories distinguished by Thomson Scientific within the natural and the social sciences. The main finding is that, in a sample consisting of 767,828 articles published in 1998 with a 5-year citation window, in 181 out of 221 sub-fields (representing approximately 77 % of the sample of articles) the existence of a power law cannot be rejected. In most sub-fields, the upper tail that can be represented by a power law is small but captures a considerable proportion of the total citations received. The value of the scale parameter is between 2 and 3 for only 21 sub-fields or 6 % of the total, greater than 3 for 114 sub-fields or 67%, and greater than 4 for the remaining 46 sub-fields that represent 27 % of the total. The estimation of the parameters of the power laws has been done with a novel procedure for citation distributions that is shown to perform better than the standard maximum likelihood methods in the presence of extreme observations