We divide the empirical analysis into three sections relating to the three elements of
the model defined in section 2. We focus initially on plants' knowledge sourcing
activities, examining the determinants of R&D intensity and the extent of plants
supply chain and non supply chain knowledge sourcing activities. Section 4.2 then
considers the determinants of plants' innovation success and section 4.3 looks at the
link between innovation and business performance.
4.1 Knowledge Sourcing
R&D intensity and the two indicators of the extent of plants' innovation collaboration
are defined as percentages and Tobit estimation is therefore the relevant technique.
Table 3 gives some preliminary estimates of these relationships across the six regions
where the explanatory variables are included in the model if - in at least one of the
three models - they have a t-statistic in excess of one. The first point of interest with
the models is the sign and significance of the other knowledge sourcing activities.
This gives an indication of the substitute (negative) or complementary (positive)
relationship between the different types of knowledge sourcing activity. Our results
emphasise complementarity although the strength of this relationship is not uniform
between the three knowledge sourcing activities (Table 4). In particular, we find that
plants engaged in non-supply chain collaboration were also much more likely to be
engaged in supply-chain collaboration. Both were only weakly linked to R&D
intensity. R&D intensity, however, was more likely among plants undertaking supply-
chain collaboration.
Other aspects of these initial knowledge sourcing equations are less well defined with
little consistency evident between the indicators which proved important. In terms of
the impact of market position on R&D intensity, for example, older plants and those
identifying significant barriers to innovation because of low rates of return were more
likely to be undertaking R&D. Neither effect proved important for either supply chain
or non supply chain collaboration. Perhaps most surprising, however, was the
weakness of variables linked to both relative and absolute plant size. Regional factors
also proved difficult to identify with any precision with little clear link between levels
of regional R&D (in any sector) and plants' knowledge sourcing activities.
At this stage perhaps the clearest result from the knowledge sourcing equations is
therefore the relatively strong complementarity between knowledge sourcing
activities. The implication being the type of positive externalities envisaged in much
of the endogenous growth theory literature.
4.2 Innovation Success
In the conceptual model defined above (section 2), innovations (new or improved
products) are the result of an innovation production function having the knowledge
sourcing activities as the inputs (i.e. equation 4). In the multi-regional database
innovation success is defined in terms of the percentage of sales derived from
products newly introduced or updated over the previous 3 years. Tobit is therefore
again the relevant estimator and two initial estimates of the innovation production
function are given in Table 5. Key interest in the relationship focuses on the
significance of the knowledge sourcing activities for innovation success and whether
11