Geography Reference
In-Depth Information
together by means of fairs, conferences, business meetings, et cetera), instead of perma-
nent co-location (Torre, 2008; Torre and Rallet, 2005). In sum, for analytical purposes,
it is essential to dei ne geographical proximity in such a restricted manner, and to isolate
it from the other dimensions of proximity.
As proximity is an analytical concept, it of ers many advantages in theoretical and
empirical work explaining the (spatial) structure of networks.
First, one can extend the list of relevant proximity dimensions with any other dimen-
sion without changing the meaning of each dimension. For example, linguistic or ethnic
proximity can be introduced. Thus, the proximity dimensions are analytically orthogo-
nal even though many dimensions of proximity may empirically turn out to be corre-
lated. Just to give one example, social proximity between two organizations is generally
higher for geographically proximate organizations, because friendships are more easily
established and maintained over short distances.
Second, by incorporating multiple proximity dimensions in an explanatory frame-
work, one can test what forms of proximity are best able to explain the formation of net-
works. For example, many networks are geographically localized. One is then tempted
to argue that this is the case because of transportation costs. However, if one explains the
presence or absence of links using indicators of both geographical proximity and social
proximity, one may i nd that networks are actually based on social proximity and not on
geographical proximity. Yet, if i rms with a high social proximity are often co-located,
it can seem that geographical proximity is underlying the formation of networks. Thus,
ideally, one takes into account as many proximity dimensions as possible as to control
for all possible reasons that may underlie network formation between organizations.
Third, the analytical nature of proximity concept allows one to understand the inter-
play between dif erent dimensions. In particular, one can expect proximity dimensions in
innovation networks to be substitutes rather than complements (Boschma, 2005). That
is to say, to establish a (successful) relation, one is in need of proximity in at least one
dimension to manage the uncertainty involved. Being proximate in a second dimension,
then, adds relatively little to the probability a link is formed, or the probability that the
relation is successful. Making use of patent data, Singh (2005) found that geographical
proximity is especially important in the establishment of interdisciplinary research collab-
oration (when cognitive proximity is low), while inventors working in the same i eld (i.e.
cognitive proximity is high) collaborate on average over longer geographical distances.
Making use of publication data, Ponds et al. (2007) found that geographical proximity
is especially important in the establishment of university-industry-government relation-
ships (i.e. institutional proximity is low) and less important in university-university col-
laboration where actors operate under the same institutions (i.e. institutional proximity
is high). Agrawal et al. (2006) found that knowledge is transferred between i rms in dif-
ferent locations (so geographical proximity is low) by employees that are socially linked
because of a shared past. Breschi et al. (Chapter 16 in this volume) found similar results
when analysing the social networks of US inventors who are mobile in space. Although
inter-regional mobility of inventors is very low, the few inventors who did move between
regions often maintained their ties with former co-inventors, providing a channel of
knowledge dif usion to their prior location.
