![Sample our Geography journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5937/TOC-Subject-Sample-2021-Geography.jpg"})
ABSTRACT
Recently, radical innovations have received increasing attention in order to achieve long-term economic success. Regional clusters, being frequently used as an innovation policy instrument, have been shown to have the potential to support innovations in general. However, it remains unclear whether clusters are really a beneficial environment for the generation of radical innovations. This study aims to shed light on the specific role clusters can play in radical innovation processes. In order to do this, we apply a quantitative approach on the firm-level and combine several data sources (e.g. AMADEUS, PATSTAT, German subsidy catalogue). Our results show that clusters indeed provide a suitable environment for radical innovations. Furthermore, we find that radical innovations rather occur in the periphery of the cluster, where actors tend to be more open to the exchange of external knowledge. This happens in general through linkages with other actors, which we also find to be beneficial for the emergence of radical innovations up to a certain degree. Our findings implicate that policy makers should continue to support clusters and further develop funding schemes. Moreover, managers should be open to collaborations with other actors for the cross-fertilization of knowledge to promote radical innovations.
Acknowledgements
The authors would like to thank two anonymous referees for comments on an earlier draft and the participants of our presentation at the 1st ‘Rethinking Clusters’ workshop in Florence. The usual disclaimer applies. The data that support the findings of this study are available from the corresponding author, [NG], upon reasonable request.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1. Although the analysis of our study focuses on invention processes, the paper uses the terms ‘innovation’ and ‘invention’ interchangeably.
2. Based on the results of the comparative empirical approach applied in Grashof and Fornahl (Citation2017), highlighting that the spatial connection, the thematic connection and interdependencies are regarded within the literature as the core elements of cluster definitions, industrial districts stressing particularly informal relationships, social capital and trust, are only seen as one specific form of a cluster and hence are not taken explicitly into account here.
3. In line with our cluster definition, we do not consider cities as clusters in this study.
4. A full list of the NACE codes can be found at Eurostat e.g.: https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Glossary:Statistical_classification_of_economic_activities_in_the_European_Community_(NACE).
5. In accordance with the literature, 45 min are here perceived to be an adequate limit for close geographical distance (Brenner Citation2017; Scholl & Brenner Citation2016).
6. By using this standard threshold, we avoid to choose arbitrarily a threshold, which constitutes a limitation to several studies dealing with the relationship between clusters and firm performance (e.g. Hervás-Oliver et al. Citation2018b).
7. Even though patents are commonly used in empirical studies, we still want to acknowledge its flaws. For example, not all innovations are patented and some innovations cannot be patented. For a discussion on shortcomings of patent data, see e.g. Griliches (Citation1990). Nevertheless, patents offer extensive and detailed information on the inventory process such as the date, applicant and technology and over a long time. Hence, it very well fits our empirical approach.
8. A Token algorithm with a log-based weight function has been utilized. It belongs to the group of vectorial decomposition algorithms and compares the elements of two text strings by separating them by their blank spaces (for more information, see e.g.: Raffo Citation2017; Raffo & Lhuillery Citation2009).
9. For a detailed overview of the number of radical innovations and the pace of technology evolution by industry, please see Appendix 3.
10. A geographical distribution of the firms in our sample and in total Germany can be found in Appendix 2.
11. The results concerning the application of the cluster index and the interaction term as well as the results of the subsample can be provided by the authors upon request.
12. As indicated by Ai and Norton (Citation2003) problems may raise regarding the interpretation of such an interaction term. However, by using log-odds, we argue that the interpretation problems raised by Ai and Norton (Citation2003) are not that relevant in our case, as the logit model is a linear model in the log odds metric (logit-scale) whereas transformed to the probability scale it indeed becomes nonlinear (Kohler & Kreuter Citation2008; MacKenzie et al. Citation2018; UCLA Citation2018).
13. The cluster index-specific marginal effects of being located in a cluster are illustrated in Appendix 4.
14. Results can be provided by the authors upon request.
15. Particularly referring to the calculation of the cluster index.
16. Since causality is hard to determine with cross-sectional data, in line with Hervás-Oliver et al. (Citation2018b) we claim correlation rather than cause and effect.