“The Diffusion of Scientific Knowledge Across Time and Space,” P. Azoulay, J.S. Graff Zivin, B.N. Sanpat (2011)

How important is it to be near “star” scientists? Theories in urban geography regularly suggest that face-to-face contact is critical for the spread of ideas – dense cities tend to be richer partly for this reason. Even in the age of the internet, physical proximity may still be important for knowledge spillovers. But how important? Azoulay et al attempt to answer this question in a paper that might be called “the last, best gasp of citation-based knowledge studies”.

The authors gather article (common) and patent (less common) data for a sample of almost 10,000 academic life scientists over a decades-long period. In what was surely a tedious process – have pity on their poor RAs – they gather CVs with affiliation information for each of these scientists over their academic lifetime, and note when academics shift to a new university. They then examine citation rates on papers written before the move in the original and new location.

Mean article citations rates from the new location rise by about one citation per article per year, and stay roughly steady at the old location. The rise in citations at the new location begins two years before the actual move, which is unsurprising if universities deliberately try to hire researchers working in a similar area as their current staff. The rise is compared to baseline “author-article dyads” that are, in theory, matched quite nicely in the pre-move period; that is, matched articles have similar numbers of authors, a similar age, the same journal, and similar numbers of pre-move citations. The usual caveats about using matching in an attempt to gain identification apply, of course. Patent citations look a little different: citations to patents held by academics fall sharply after the move, which indicates that face-to-face time may be particularly important for translating academic research into industry. This result should be treated with caution, however, due to the low number of authors with a patent, normal patent thicket issues, etc.

You may know these authors from their very nice “Superstar Extinction” paper in QJE last year – the identification there seems quite a bit cleaner, so that paper probably remains the gold standard in this thread of literature. There is a major worry in this whole line of research, however, which is best described in MacRoberts and MacRoberts’ highly-cited (hah!) “Problems in Citation Analysis”. Basically, it is not clear at all that citations should be read as signifying “knowledge spillover”. For instance, citations serve a ceremonial role, so, as Azoulay et al note, the lack of decline in citations for a academic that left his original university may simply reflect a “ceremonial” namecheck rather than any deeper connotation of transferred knowledge.

http://pazoulay.scripts.mit.edu/docs/diffusion_rdie.pdf (NBER WP from last fall’s innovation conference)

%d bloggers like this: