Page 1 of 11 A Research and Institutional Size Based Model for National University Web Site Interlinking Mike Thelwall 1 School of Computing and Information Technology, University of Wolverhampton, 35/49 Lichfield Street, Wolverhampton, WV1 1EQ, UK. Email: cm1993@wlv.ac.uk Web links are a phenomenon of interest to bibliometricians by analogy with citations, and to others because of their use in web navigation and search engines. It is known that very few links on university web sites are targeted at scholarly expositions and yet, at least in the UK and Australia, a correlation has been established between link count metrics for universities and measures of institutional research. This paper operates on a finer-grained level of detail, focussing on counts of links between pairs of universities. It provides evidence of an underlying linear relationship with the quadruple product of the size and research quality of both source and target institution. This simple model is proposed as applying generally to national university systems, subject to a series of constraints to identify cases where it is unlikely to be applicable. It is hoped that the model, if confirmed by studies of other countries, will open the door to deeper mining of academic web link data. Introduction Web links are a fascinating object of study because of the implicit information that they contain: the premise that a page with many inlinks is likely to be useful drives the search engine Google, for example (Brin & Page, 1998). Bibliometric interest has recently concentrated upon counts of links to entire web sites or national domains. Ingwersen (1998) introduced the key link count metric, the Web Impact Factor (WIF), which has since been used in various guises to drive research into this phenomenon. Academic web sites have been the most intensively studied and it is now known that aggregates of links to an institution correlate with measures of its research output, once size is taken into account (Thelwall, 2001d, Thelwall, 2002b). This paper analyses links at a finer-grained level of detail, counting not all links to an institution but instead all links between pairs of institutions. This allows an exploration into patterns of inter-university linking to investigate in particular whether source and target institutional size and research are useful indicators and, if so, the most effective model that can be derived from using these as predictor variables. The majority of previous research into web link metrics has focussed on sites with academic content: university web sites (Smith, 1999; Darmoni et al., 2000; Thelwall, 2000; Thomas & Willet, 2000; Vreeland, 2000; Thelwall, 2001d; Thelwall, 2002b); e-journals (Harter & Ford, 2000; Kim, 2000); journal web sites (Soualmia et al., 2002); and subject-specific web sites (Larson, 1996; Hernández-Borges et al, 1999), perhaps motivated by the analogy with citations (Davenport & Cronin, 2000). Entire countries have also been subjected to web link analysis, however (Ingwersen, 1998). In order to develop meaningful metrics for the academic web, a theory of 1 Journal of Documentation, 2002: 58(6), 683-694.