Geography Reference
In-Depth Information
footing. Highly intricate recipes resist dif usion to even nearby actors. But for recipes of
intermediate interdependence, nearby actors receive enough guidance from the template
that local search delivers them an ef ective replica of the original knowledge on which
they can build, while distant actors begin their search processes from such l awed start-
ing points that subsequent ef orts to receive and build on the interdependent recipe tend
to fail.
4. Empiricalcorroboration
To test our hypothesis, we analyzed prior art citations to all US utility patents granted
in May and June of 1990 (
n
= 17,264).
5
The data came from the Micro Patent data-
base and NBER public access data on patents (Hall et al., 2001). Following much
previous research, we view a prior art citation as evidence of knowledge dif usion: the
applicant has successfully assimilated the knowledge underlying the original patent to
a new setting and built on it. Our statistical approach is to estimate the likelihood that
a focal patent receives a citation from a future patent as a function of several factors:
the interdependence of the knowledge underlying the focal patent, the proximity of the
inventors of the focal and citing patent in a social network, the interaction of interde-
pendence and social proximity, and a set of control variables. The results of the estima-
tion allow us to examine how the likelihood of citation by a socially proximate inventor
compares to the likelihood of citation by a distant inventor as a function of knowledge
interdependence. The crucial test of our hypothesis is whether the gap between the
two probabilities peaks when the focal patent embodies moderately interdependent
knowledge.
Patents and the meaning of citations
Patents and their citation patterns provide an attractive test bed for our hypothesis for
several reasons. First, these citations have been carefully assigned. The US Patent Oi ce
requires all applicants to demonstrate awareness of their invention's precedents by
citing similar 'prior art' patents. Patent examiners in each technological domain review
and supplement the prior art references to ensure accurate and comprehensive cita-
tions. Second, consistent with our ontology of knowledge, technology historians have
demonstrated that one can conceptualize patented inventions as combinations of pre-
existing technological components (Basalla, 1988). The process of invention therefore
involves both the replication of prior discoveries and the extension of those discoveries
to new applications and in new combinations. When a citation to prior art emerges on
a new patent, it suggests that the inventor has both successfully received and built on
the knowledge underlying the earlier patent. Third, Fleming and Sorenson (2001) have
developed a technique for measuring the interdependence among the components of
an invention. The technique draws on information uniquely available for patents and
potentially dii cult to duplicate in other settings.
This setting nevertheless also has its limitations. First, our analysis rests on the
assumption that some potential knowledge recipients have better access to the template
than others. If every patent fully revealed the inventor's underlying knowledge of the
invention, this assumption would not hold. Inventor's incentives, however, minimize
the likelihood of this problem. Patent applicants prefer to disclose as little as possible
to limit their competitors' ability to benei t from their disclosure (Lim, 2001). Indeed,
