judged from the reference level of full integration (no bias at all as indicated by T=0), but
from a (unknown) reference level of bias that would occur when institutional and language
barriers between countries were to be fully removed. 11 This reference level of bias that is
expected to remain in homogeneous geographical territories is expected to be substantial
following from the theory of geography of innovation as discussed in the second section.
Even if scientific knowledge does not share the characteristics of industrial innovation for
what concerns its degree of specificity, tacitness, and appropriability, some degree of spatial
concentration is expected to remain.
4.2 Intra-national versus international collaboration
The next question regarding the integration process at the EU level holds whether the
integration process is an effect of a decreasing bias of countries to collaborate nationally or an
effect of a decreasing bias with regard to the choice of EU partners, or a combination between
the two. Fig. 2 plots the Ti=j-values of intranational collaborations within the EU countries for
each year (formula 4). The results indicate that there is a stable positive bias to collaborate
nationally. Over the years, the average value is about +2.04 with means that the probability of
national collaboration is e2.04 = 7.7 times higher than when partner selection would have been
at random. There is no real trend in the national bias over the years. The results at least
suggest that the bias to collaborate nationally has not decreased during the period 1993-2000.
FIGURE 2 AROUND HERE
Fig. 3 plots the Ti j-values of international collaborations in the EU for each year (formula 5).
From the results two important observations can be made. First, the bias to collaborate with
other EU countries has been negative over the whole period considered. Over the years, the
average value is about -1.62 with means that the probability to collaborate with another
member state is on average e-1.62 = 0.20 times lower than when partner selection would have
been at random. Second, the negative bias towards European collaboration has become less
and less over the years. The integration process as indicated by the trend in Fig. 1 can thus be
understood as the result of a decreasing bias in the selection of a European partner while the
11 The application of the mutual information indicators to regions within a homogeneous territory such
as the United States would give an indication of the degree of spatial concentration that is expected to
remain. This analysis falls outside the scope of this study.
12