Collaboration Intensity and Regional Innovation Efficiency in Germany—A Conditional Efficiency Approach

Abstract

Many case studies highlight a positive relationship between regions' innovation performance and the intensity of collaboration among regional organizations. However, few efforts have been made to analyze this relation with quantitative approaches. In addition to a theoretical discussion, the paper presents an empirical investigation on this issue utilizing conditional efficiency analysis and patent co-application data for the Electrics & Electronics industry in 270 German labor market regions. The results show that the relationship between regions' innovation performance and the intensities of regional as well as inter-regional collaboration take the form of an inverted-U shape. Regions with average regional and inter-regional collaboration intensities are found to outperform those characterized by extremely low, high or unbalanced collaboration behavior.

Keywords:

• Regional innovation efficiency
• collaboration intensity
• patent co-application
• conditional efficiency analysis

Notes

¹ Boschma (2005) names also organizational, cognitive, institutional and social proximity.

² Regional innovation performance refers to the aggregated innovation performance of all regional organizations.

³ A similar but more general definition can be found in Fritsch (2000: 415).

⁴ Other advantages include the simultaneous consideration of multiple output variables and that no distributions have to be specified for the error term as well as for the efficiency estimate. A more detailed discussion can be found in Coelli et al., (1998) and Daraio and Simar (2007).

⁵ This corresponds to an output-oriented version, which has been argued to be more appropriate in this context than the input-oriented version (Broekel and Brenner, 2007).

⁶ Here m is a trimming parameter defining the sensibility of the estimation with respect to outliers in the data.

⁷ It is acknowledged that patents rather capture inventions than innovations. However, in order to stay consistent with the literature, the term “innovation” is used in the paper.

⁸ Nomenclature statistique des activités économiques dans la Communauté européenne, Rev. 1.1.

⁹ The existence of innovation generators is a prerequisite for the generation of innovation implying that regions with zero R&D employment are expected to have zero innovations. Efficiency cannot be meaningfully defined in this case.

¹⁰ As before, patents are assigned to regions according to the inventor principle.

¹¹ Accordingly, I have to assume that inter-regional and inter-national collaboration intensities are positively correlated.

¹² ***, ** and * denote a significance level of 0.01, 0.05 and 0.1, respectively.

¹³ Significance is based on Wilcoxon rank sum test.

¹⁴ All estimations are also conducted in two alternative setups. In the first, no control variables are considered and in the second all control variables are taken into account that are significantly correlated to the innovation output. The results are very similar to the reported ones. They can be obtained upon request from the author.

¹⁵ The ranking of regions according to their efficiency and collaboration intensity values is very stable over time. There are also no indications of significant temporal trends in the variables.

¹⁶ A constant of 0.001 is added to all regions' output to ensure proper estimations.

¹⁷ While the efficiency estimate is spatially autocorrelated, the correlation is low in magnitude (Moran's I: 0.06**, Geary C: 1.05).

¹⁸ The primary reason for the little change in the ranking is the heavily skewed regional distribution of patents. It implies a scarcity of comparison observations for regions with large patent numbers resulting in these being automatically deemed efficient.

¹⁹ The analysis has been repeated for regions with at least 5, 10 and 20 patents as well as under consideration of a dummy for East Germany (p-values: REGION, 0.018; INTER, 0.003; EAST, 0.001). The results remain more or less the same as reported above. An exception being the negative trend for INTER, which is becoming less pronounced when small regions are not considered or it is controlled for regions being located in East Germany (which tend to be small in terms of patent output). See Figures 11 and 12.