The abstract reads:
We report a puzzling pair of facts concerning the organization of science. The concentration of research output is declining at the department level but increasing at the individual level. For example, in evolutionary biology, over the period 1980 to 2000, the fraction of citation-weighted publications produced by the top 20% of departments falls from approximately 75% to 60% but over the same period rises for the top 20% of individual scientists from 70% to 80%. We speculate that this may be due to changing patterns of collaboration, perhaps caused by the rising burden of knowledge and the falling cost of communication, both of which increase the returns to collaboration. Indeed, we report evidence that the propensity to collaborate is rising over time. Furthermore, the nature of collaboration is also changing. For example, the geographic distance as well as the difference in institution rank between collaborators is increasing over time. Moreover, the relative size of the pool of potential distant collaborators for star versus non-star scientists is rising over time. We develop a simple model based on star advantage in terms of the opportunities for collaboration that provides a unified explanation for these facts. Finally, considering the effect of individual location decisions of stars on the overall distribution of human capital, we speculate on the efficiency of the emerging distribution of scientific activity, given the localized externalities generated by stars on the one hand and the increasing returns to distant collaboration on the other.Now this makes sense. As we become more specialised, as the division of labour becomes greater, we each know more about less and thus collaboration offers increasing returns. Also if we think of the falling cost of communication as falling transaction costs then greater inter-departmental collaboration makes sense. This is because there is an asymmetry in the way that falling communication costs affects intra-departmental and inter-departmental transaction costs. Things like the internet, skype and email will not affect intra-departmental transaction costs much, if at all, but they will lower the transaction costs of inter-departmental collaboration. And thus we should expect to see greater use of the relatively cheaper option. Also one could see the increasing difference in institutional rank as being, in part, due to the now lower costs for up-and-comers, who may be starting their careers at lower ranked institutions, to collaborate with "stars".