Modelling Science

Contemporary science is a social activity localised in such places as classrooms, laboratories, university offices and conferences. Though localised in specific places, part of science explicitly aims at developing knowledge that is globally applicable. As such, scientific practice is thus
likely to be affected both by the global disciplinary context of scholars as well as the way individual scholars relate to one another in a local context.
This paper focuses on one particular kind of scientific practice, namely citation, and assesses whether geographical proximity among scholars affects the likelihood of citation. The question we address thus concerns the role of geographical proximity in citation: how and to what extent does geographical proximity between a citing text and (potentially) cited texts impact on a text citing one particular text rather than another text? As such, we derive the following hypothesis: the probability of a citing text being geographically proximate to a cited text conditional on the citing text citing the cited text is higher than the probability of a citing text being less geographically proximate to a potentially cited text conditional on the citing text not citing the potentially cited text.
To address this issue empirically we use single-authored publication data as extracted from Scopus Elsevier 1996-2008. Though focusing on the citing instead of the cited text methodologically we follow Jaffe et al (1994) and Breschi and Lissoni (2003) in that:
1. We construct a set of citing texts and from their reference list a set of cited texts. In order to construct a control group of potentially cited texts that match the group of cited texts we search for texts that (i) are published in the same journal and year (ii) have the largest amount of overlap in references with the cited text (iii) are not cited by the citing text. We thus assume that the control group of non-cited texts that match the cited texts are of equal relevance to the citing text as the text that is cited.
2. We compute the geographical proximity between the citing texts and both the cited and non-cited text as the kilometric distance separating the author affiliations.
3. We control for social proximity in order to assess the geographically localized nature of citation practices in science. Here, we construct social networks of scholars as can be derived from past co-publications.
In all we thus propose to assess the probability of a citing text being geographically proximate to a cited text conditional on the citing text citing the cited text and controlled for social proximity being higher than the probability of a citing text being less geographically proximate to a potentially cited text conditional on the citing text not citing the potentially cited text. As such, we determine the extent to which science citation are localized.

References
Breschi, S., & Lissoni, F. (2003). Mobility and social networks: localised knowledge spillovers revisited. Centro di Ricerca sui Processi di Innovazione e Internazionalizzazione (CESPRI) WP: 142.
Jaffe, A. B., Trajtenberg, M., & Henderson, R. (1993). Geographic localization of knowledge spillovers as evidenced by patent citations. The Quarterly Journal of Economics , 108 (3), 577- 598.