Distance (Still) Hampers Diffusion of Innovations

Abstract

This paper explores a newly emerging source of data on the occurrence and spread of innovations
drawn from descriptions of goods and services in the US trademark register. Using these data we
describe patterns of regional innovation in the United States and study the effect of distance on the
early diffusion of innovations. To identify innovations and their locations we tokenize trademark
descriptions and identify novel, fast spreading tokens (words). While trademarks appear to capture
many innovations missed in patents, among tokens that co-occur in patent documents the diffusion
dynamics for patents and trademarks are similar. We also find that fast growing new tokens are
frequently new to English, and their use in language co-evolves with the frequency of linked patents
and trademarks. Finally, we show that regional incidence of new tokens parallels patterns of inventive activity reflected in patent data. We exploit occurrence of new trademark tokens to re-examine how spatial distance affects the diffusion of innovations in the U.S. economy. Aggregating innovations at the year and census tract level we estimate Poisson models of diffusion intensity between locations, revealing persistent, strong and negative effects of distance on the intensity of diffusion between locations within the US.

Keywords: Innovation, Diffusion, Rate of Diffusion, Distance, Innovation Index, Trademarks, Patents

JEL Classification: O3, O51, R1, R32