This is the latest article in a series concerned with journal citations that has appeared in Communications of the ACM (CACM) over the past decade. As with the previous articles [1,2,3], the range of journals selected by Nerur and his co-authors is broad indeed, but somewhat biased toward management information systems. The end result is a rather eclectic selection that spans management, information systems (IS), computer science (CS), databases, artificial intelligence (AI), decision support (DS), software engineering (SE), and more. I personally cannot think of a single academic colleague who would regularly peruse the 27 journals listed here. To be fair, the authors have been guided by the previous CACM studies. Nevertheless, I would be much more interested in results obtained on a subset of this journal list.
That said, let us now turn our attention to the findings of Nerur et al. Their focus is on relative journal influence, which they attempt to quantify by identifying knowledge sources and storers in a citation network. They further consider the reciprocity (mutual influence) of citation flows. Given the inherent limitations of raw citation measures, the authors choose a log-multiplicative measure to control for such biases. For the nominated journals, Nerur et al. generate a 27-by-27 matrix from articles referenced between 1998 and 2002 in the science and social science citation indexes, using the ISI Web of Knowledge database. As with the previous studies, CACM was found to be the most influential source; in other words, it was the publication most cited by other journals (after controlling for self-citations and the number of citations sent by CACM to other journals in the network).
The authors use a log-multiplicative model with multiple dimensions of association, similar to one used in a previous study of economics journals [4]. They are able to identify primary knowledge sources (cited journals) and storers (citing journals). Moreover, an application of Ward’s hierarchical clustering technique first extracted two broad categories (sociotechnical and technical), and the following five dimensions (sub-clusters): European, North American, computational intelligence, techno-centric, and computer science.
This article focuses less on the journal rankings themselves, and more on their relative influence, using the measures outlined above. While the authors’ findings possess a certain inherent interest, their usefulness is somewhat restricted, due to their aggregated nature. Indeed, it would be very interesting to see a follow-up study performed that maps journals related to a specific model curriculum, in particular those jointly developed by the Association for Computing Machinery (ACM) and the Institute of Electrical and Electronics Engineers Computer Society (IEEE-CS).