Geography Reference
In-Depth Information
ences the dynamics of dif usion. Specii cally, a socially proximate actor's advantage over
a distant actor in obtaining and building on knowledge peaks when the components
underlying the knowledge display intermediate interdependence. Though our empirical
results come from patent data alone, the basic logic of our hypotheses applies to knowl-
edge in general, not just the knowledge underlying inventions. Hence, future research
might usefully examine these dynamics across a wide range of applications - including
organizational learning, the dif usion of management practices, knowledge manage-
ment, and the sustainability of knowledge-based competitive advantage.
Notes
* All parties contributed equally to this research. We thank Jasjit Singh for generously sharing his inventor
collaboration network data. We also appreciate the helpful comments of George Baker, Matt Bothner,
Koen Frenken, Bob Gibbons, Jerry Green, Rebecca Henderson, Bruce Kogut, Dan Levinthal, Woody
Powell, Rosemarie Ham Ziedonis, three anonymous reviewers, and seminar participants at Harvard, the
University of Michigan, MIT, New York University, Ohio State University, the University of Toronto,
Washington University, and Wharton. All errors remain our own. Harvard Business School's Division of
Research provided i nancial support. This chapter was originally published in
Research Policy
,
35
(2006),
994-1017. Subsequent versions of this chapter have been presented at the annual conferences of the
Academy of Management, the European Group on Organizational Studies, and the European Meeting
on Applied Evolutionary Economics, as well as at the ESF Exploratory Workshop on Evolutionary
Economic Geography.
1. Hansen (1999) also focuses on the interplay between social relations and knowledge l ow. His research
dif ers from ours in three respects: (1) it does not explore the issues related to recipient search as a mecha-
nism for the interplay, (2) it focuses on the strength of the connection between inventors rather than social
proximity in a network; and (3) it analyzes the ef ects of a
portfolio
of relations rather than the characteris-
tics of a connection in a dyad. We nonetheless see close parallels that we revisit in the discussion section.
2. In this game, one child whispers a message into the ear of another, who then whispers what she heard into
the ear of a third child and so forth. At the end, the i nal person announces the message he heard and the
i rst person reveals the message that she originally whispered; the two usually dif er dramatically.
3. This assumption limits the applicability of our theory to innovations that involve multiple components.
This restriction should not severely constrain its scope, however; few innovations do not involve the
combination of multiple physical components or processes. For example, even the synthesis of nylon, a
polymer, involved the integration of several distinct processes (Smith and Hounshell, 1985).
4. Though not considered here, one might also consider the importance of tie 'strength'. Weak ties have long
reach but low bandwidth; thus, they operate most prominently in the dif usion process when transferring
only short, simple messages (Hansen, 1999).
5. We constructed this dataset in the course of prior research. For details on its construction, see Fleming
and Sorenson (2001).
6. We chose four patents for the 'control' group so that the sample would have a roughly equal proportion
of realized and unrealized dyads. Although some feel that conditioning on important factors improves
the statistical power of a case-control sample (e.g. Jaf e et al., 1993, implicitly make such an argument
in drawing controls from the same classes as the citing patents), the ideal method of selecting controls
remains an open debate. Matching controls to cases on one or more dimensions can lead to two problems
in particular that concern us. First, correcting the logit for over-sampling on the dependent variable
requires that one knows the sampling probabilities (King and Zeng, 2001); matching controls to cases
precludes the possibility of calculating this information. Second, matching on an endogenously deter-
mined factor risks generating biased results (e.g. when investigating dif usion processes, one would not
want to consider the geographic distribution of activity exogenous). Given these concerns, we sample
future patents at random and control for heterogeneity in the estimation.
7. Including the foreign inventors does not change the results qualitatively.
8. Our measure
k
is related to but distinct from the parameter K in the NK simulation models that have
become popular in theoretical work on complex systems (Kauf man, 1993). In NK simulations, the con-
tribution of each element in a system to overall system i tness depends on the states of K other elements.
K is set by the modeler and, like our empirically measured
k
, rel ects the degree of interdependence among
components in a system. Despite the conceptual linkage between our measure
k
and Kauf man's K, we
do not purport to have measured his K in a literal sense. For instance, our
k
does not equal the number
of elements that af ect the contribution of each focal element.
