Highlights :

• We show that the negative effect of geographical distance on knowledge flows stems from how firms gain sources of knowledge through their existing network.
• In aggregate, the distance elasticity of patent citations flows has remained constant since the 1980s, despite the rise of the internet.
• At the micro level, firms disproportionately cite existing knowledge sources, and patents cited by their sources.
• We introduce a model featuring the latter phenomenon, and generating a negative distance elasticity in aggregate.
• While the distance effect should have decreased with constant country composition, the rise of East Asian economies, associated to large distance elasticities, compensated lower frictions in other countries.

 Abstract :
This paper shows that the negative effect of geographical distance on knowledge flows stems from how firms gain sources of knowledge through their existing network. We start by documenting two stylized facts. First, in aggregate, the distance elasticity of patent citations flows has remained constant since the 1980s, despite the rise of the internet. Second, at the micro level, firms disproportionately cite existing knowledge sources, and patents cited by their sources. We introduce a framework featuring the latter phenomenon, and generating a negative distance elasticity in aggregate. The model predicts Pareto-distributed innovator sizes, and citation distances increasing with innovator size. These predictions hold well empirically. We investigate changes of the underlying parameters and geographical composition effects over the period. While the distance effect should have decreased with constant country composition, the rise of East Asian economies, associated to large distance elasticities, compensated lower frictions in other countries.

 Keywords : Knowledge Diffusion | Innovation Networks | Spatial Frictions | Patent Citation

 JEL : L14, O33, R12