these reasons, one should expect scientific knowledge production to be less localised than
industrial innovation.
A number of studies have addressed the characteristics of tacitness, specificity and
appropriability of knowledge as variables that explain the degree of geographical localisation
of knowledge production (Feldman, 1999). For example, a U.S. patent citation study has
found that specificity and appropriability of knowledge as documented in patents contributed
significantly to the extent that citation originated from the same region (Jaffe et al., 1993). If
one accepts that scientific knowledge production is typically characterised by a low degree of
specificity and appropriability, this finding suggests that the degree of localisation of
scientific knowledge production is indeed lower compared to industrial R&D. Concluding,
both theory and evidence suggests that scientific knowledge production differs from industrial
knowledge production in that the latter may primarily be accumulating at a regional scale,
while the latter is expected to internationalise over time.
The recent rise in international collaborations in scientific research can relatively easily be
indicated by computing the share of international co-authorships in all publications. It has
been estimated that the share of international collaborations has doubled during the period
1987-1997 to account for 15 percent of world publications (Wagner, 2002). To assess the
benefits of international collaboration is a somewhat more difficult exercise. Empirical studies
that addressed the benefits of international collaboration have focused on scientific impact
and productivity (Katz and Martin, 1997). The impact of scientific output resulting from
collaboration as measured by citation rates is substantially higher than average. The
difference in citation impact is even higher for international collaboration. Furthermore, it has
been found that the productivity of scientists is positively dependent on the frequency of
collaboration. Collaboration tends to increase the level of personal productivity as measured
by the number of publications produced per year.
A large part of European science policy can be considered as an attempt to capitalise on the
potential of scientific collaboration among member states. Not surprisingly, research
collaboration and mobility of researchers is at the core of its policies. Given the evidence on
the positive effects of research collaboration, and the policy importance attached to it, an
important question holds whether empirical data show that the European science system has
indeed become more integrated or not. Within the European context, the number of European
collaborations has undeniably increased over the last few decades. However, the number of
collaborations is in itself no indication of integration (Leydesdorff, 1992). For example, the
number of collaborations can double in a period, but at the same time the distribution of