Geography Reference
In-Depth Information
to which actors within the i rm interact with each other. Managers can inl uence who
likely interacts with whom through the assignment of individuals to facilities, the design
of laboratories and factories, and the structure of reporting relationships (Allen, 1977).
To distribute knowledge ef ectively, a i rm might usefully expend resources to foment
close and dense social connections between sources and intended recipients of complex
knowledge, while letting networks remain sparse elsewhere. Indeed, leaders might fruit-
fully construe the task of knowledge management
not
as the construction of central
databases of information (as sometimes presented today), but rather as an ef ort to build
social networks that match the nature and intended l ow of knowledge. Ef ective organi-
zational design, however, surely requires a deeper understanding of how social structure
af ects knowledge dif usion than considered here; networks have subtle features and
nuances that doubtlessly inl uence their ability to convey knowledge, both simple and
complex (Hansen, 1999).
To this point, our argument has assumed that the degree of interdependence between
combinations of components remains i xed. In the long term, however, the ef ective
interdependence of knowledge may change. Firms and inventors can invest in R&D to
specify interfaces and embed knowledge within physical components, thereby reducing
the dii culty of combining a particular combination of components with other elements
in the future (Baldwin and Clark, 2000). In structuring knowledge, managers must
perform a delicate balancing act. Isolating interdependencies within substructures has
important attractions, including the ability to perform a greater number of independent
experiments (Baldwin and Clark, 2000) and the capacity to adjust more readily to envi-
ronmental shifts (Levinthal, 1997). Engineering curricula support this preference with a
strong emphasis on reliability, black box design techniques, and the re-use of previously
combined components (e.g. Mead and Conway, 1980). Such modularization, however,
also entails frequently overlooked costs. Designing and implementing an architecture
that isolates interdependencies within substructures involves considerable engineering
costs (O'Sullivan, 2001). But those direct costs potentially pale in comparison to the
indirect costs - the opportunities that the lack of complexity opens for new entrants
(Rivkin, 2000), the reduction in variety from which developers can select (Christensen et
al., 2002), and the constraints on potential performance (Fleming and Sorenson, 2001).
Managers who manipulate interdependencies should recognize that they simultaneously
alter the propensity of knowledge to l ow to actors near
and
far.
Despite the costs of modularizing, a secular trend towards modularization may inl u-
ence the evolution of industries, creating a distinctive pattern. Direct costs likely strike
i rms as more tangible than indirect costs as they decide where to direct R&D ef ort.
Thus, i rms may over-invest in less complex technology as they seek to maximize ei -
ciency. As this process reduces the ef ective interdependence of the knowledge being dif-
fused, knowledge should l ow more easily, generating two industry-level patterns. First,
an industry that begins its life in a concentrated region should become less concentrated
geographically as the advantage of preferential access to the template declines (for related
ideas, see Audretsch and Feldman, 1996; Stuart and Sorenson, 2003).
19
Second, the move
towards less complex knowledge likely reduces dif erentiation across i rms' products over
time, leading to more intense price competition and ef orts to control standard interfaces
and key modules - a pattern identii ed in the product lifecycle literature.
To reiterate, our results demonstrate that knowledge complexity importantly inl u-
