(or universities) located nearby. Their rationale for this is that
patent citations can be treated as a paper trail left by spillovers.
Obviously it is an incomplete paper trail, since not all spillovers
can be tracked in this way. But nevertheless it is an imaginative
approach to studying this difficult question. Their results find a
strong localization of spillovers. Most spillovers occur within a
metropolitan area, but some occur within the same state. Locali-
zation appears more significant for patents registered by univer-
sities and by medium and small firms than for the patents filed by
top corporations.

Feldman ( 1994) found an important correlation between the
location of innovative output and of manufacturing value-added.
Her results indicate that regional innovative output increases
in the presence of high private and academic research ex-
penditure within the state. However, it is the presence of
related industries that has the most significant effect, which
seems to demonstrate the importance of regional networks of
innovators, at least in highly innovative industries. Specialized
business services in the region also have a significant role. In later
work, Audretsch and Feldman ( 1996) also found evidence for
the clustering of innovative output. Baptista and Swann ( 1996)
found that firms located in strong clusters, where employment in
their own (two-digit) industry is high, are significantly more
likely to innovate than firms located in more sparsely populated
clusters.

This chapter has surveyed some of the main theoretical and em-
pirical work on industrial clusters and the geographical concen-
tration of innovative activity. We have seen that while this
phenomenon is pervasive, the forces leading to agglomeration
(especially in high-technology industries) differ between Europe
and the USA. As we have seen, the spatial organization of pro-
duction and innovation can be approached from several perspec-
tives. Nevertheless, a recurrent theme in the survey has been the
effects of agglomeration externalities in general, and of knowl-
edge spillovers in particular.

Industrial clusters arise from the pervasiveness of agglomera-

-50-