Mapping potentials for input–output-based innovation flows in industrial clusters – an application to Germany

ABSTRACT

Our paper pursues two aims: first, it presents an approach based on input–output innovation flow matrices to study intersectoral innovation flows within industrial clusters. Second, we apply this approach to the identification of structural weaknesses in East Germany relative to the western part of the country. The case of East Germany forms an interesting subject because while its convergence process after unification began promisingly in the first half of the 1990s, convergence has since slowed down. The existing gap can now be traced mainly to structural weaknesses in the East German economy, such as the absence of strong industrial cluster structures. With this in mind, we investigate whether East Germany does in fact reveal the abovementioned structural weaknesses. Does East Germany possess fewer industrial clusters? Are they less connected? Does East Germany lack specific clusters that are also important for the non-clustered part of the economy?

KEYWORDS:

• Industrial clusters
• qualitative input–output analysis
• intersectoral innovation flows

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 The concept of product-embodied R&D flows has been applied and extended in many input-output studies. See, for example, Leoncini et al. (1996), Sakurai et al. (1997), Papaconstantinou et al. (1998), Amable and Palombarini (1998) and Hauknes and Knell (2009).

2 We do our analysis at the level of NUTS3 regions. At this level of geographical aggregation, Germany is divided into 429 regions.

3 An obvious limitation of QIOA is the loss of information from the input-output table through binarisation. Equation 2 shows that the value of the filter rate F influences the number of relevant input-output flows. Schnabl (1994, 2000) developed a multi-stage iterative procedure to determine this filter value endogenously. The algorithm aims to minimise the loss of information through the binarisation – in other words, the procedure aims to maximise the information content of the binary input-output table. In order to determine the optimal filter rate, Schnabl (1994) applied two different measures: an entropy measure and the average value of an element of the resulting connectivity matrix. The optimal filter rate can be derived from the average of these two measures. As we aim to find information on whether two important production locations in the region under consideration are connected, this limitation might be regarded as negligible.

4 The elements on the main diagonal are not considered.

5 There are other approaches to incorporate innovation or knowledge into interdependence studies (Drejer, 2003). The application of R&D employees is in line with that of Leoncini et al. (1996). Verspagen (1997) and Los and Verspagen (2000) use patent data, while Schnabl (1995) and Duering and Schnabl (2000) apply sectoral R&D expenditures.

6 Applying the regional perspective to the clusters in the city of Nuremberg, for example, leads to 11 additional linkages (see last paragraph in section 4.3).

7 Appendix 2 contains a table presenting in-depth results for specific sectors.