Improving the accuracy of institute for scientific information's journal impact factors

Abstract

The Institute for Scientific Information (ISI) publishes annually listings of impact factors of scientific journals, based upon data extracted from the Science Citation Index (SCI). The impact factor of a journal is defined as the average number of citations given in a specific year to documents published in that journal in the two preceding years, divided by the number of “citable” documents published in that journal in those 2 years. This article presents evidence that for a considerable number of journals the values of the impact factors published in ISI's Journal Citation Reports (JCR) are inaccurate, particularly for several journals having a high impact factor. The inaccuracies are due to an inappropriate definition of citable documents. Document types not defined by ISI as citable (particularly letters and editorials) are actually cited and do contribute to the citation counts of a journal. We present empirical data in order to assess the degree of inaccuracy due to this phenomenon. For several journals the results are striking. We propose to calculate for a journal impact factors per type of document rather than one single impact factor as given currently in the JCR. © 1995 John Wiley & Sons, Inc.