Does Cluster Policy Trigger R&D Activity? Evidence from German Biotech Contests

Abstract

This article evaluates the research and development (R&D) enhancing effects of two large public grant schemes for the German biotechnology industry (BioRegio and BioProfile). Both grant schemes are organized in the form of contents for cooperation and aimed at fostering the performance of innovative firms by their organization in research clusters. We apply a difference-in-differences estimation technique in a generalized linear model framework, which allows us to control for different initial regional conditions in R&D activity of the biotech sector. Our econometric findings support the view that winners generally outperform non-winning participants during the treatment period, thus indicating that exclusive funding as well as the stimulating effect of being a “winner” have positive effects on R&D activity in the short-term. Apart from this direct winner effect, for the non-winning participants no beneficial indirect effect due to a mobilization of local actors during the application phase could be detected. Finally, first attempts in estimating the long-term effects of the contests for cooperation approach on the winner regions' R&D activity in the post-treatment period show ambiguous results.

Acknowledgements

Previous versions of this article have been presented at the GfR Summer Conference 2010 in Hannover, the RWI Therapy Summer Seminar 2010 and the IWH/University Jena Joint International Workshop on “Which regions benefit from emerging new industries? Evidence from photovoltaic and other high-tech industries”. The authors thank Uwe Cantner, Dirk Fornahl, Joel Stiebale, Christoph M. Schmidt and participants of the above events for helpful comments and advices. Also special thanks to Björn Alecke for providing some of the data used for this research, as well as Karl-Heinz Herlitschke for his efforts to extract and prepare information from the ESPACE database.

Notes

1. Using data on patenting activity, we explicitly account for the fact that patent applications are typically made with a time lag relative to the received funding.

2. However, Dohse (2000, p. 1111) points out that the implementation of regionalized technology policy in Germany was not purely intended to be a “…carbon-copy of the ideas proposed in the theoretical literature”.

3. In fact, the Bavarian State Government provided an amount of €1.45 billion for R&D projects to Bavarian firms and research institutes, which comes close to typical funding schemes at the Federal government level.

4. We thank Mr Günter Krauss from the Federal Ministry of Education and Research—Department Z 22 “Information technology”—for his effort in extracting related information from the PROFI database. In order to minimize the potential endogeneity problem stemming from the fact that winning the contests is directly associated with financial benefits for the respective regions, we use the number of raised research projects rather than the financial volume. The correlation of both indicators is reasonably high (Pearson's correlation coefficient is 0.87 for all projects and 0.91 for cooperative R&D projects), so that the number of projects serves as a good substitute for the financial volume. This strategy also avoids putting strong assumptions on the annual streams of funding over the project period.

5. The index measures industrial geographic concentration, see Ellison and Glaeser (1997) for further details. The threshold level of 0.005 for a significant geographical concentration was chosen in line with the empirical literature (see, e.g. Alecke et al. 2006).

6. A total of 55 NUTS3 regions participated in the BRC (see Table A1 in the appendix for details). Although the BRC was officially set up on a 5-year basis, between 1997 and 2001, we include an additional year as the treatment period in order to account for the usual funding practice according to an N+1 period, where N stands for the nominal time span of a specific project.

7. MINT employees are defined as employees trained in mathematics, informatics, natural sciences and technology.

8. For our estimation approach, we explicitly test for the appropriateness of the ZIP specification versus the standard Poisson model by means of standard post-estimation tests.

9. We do not include the DiD terms in the non-linear Probit part of the model since both the BRC and BPC aim to improve the track record of promising biotech regions rather than initiating a regime switch from non-innovators to innovators.

10. Additionally, we have to keep in mind that the group of non-winning participants is defined as net of the winning regions from the BPC contest, and thus has been subject to a dual selection mechanism, leaving only poor candidates within this group.

11. While highly profitable exit opportunities are offered to investors in non-listed firms, venture capital investments in biotechnology went up by a factor of six between 1997 and 2001 (see OECD 2006, p. 119). According to the “selection of the fittest” hypothesis, firms and scientists in BRC winning regions are more stimulated by the rapid growth of the venture capital market. As a result, inventions could be better protected by patent applications to secure a unique selling proposition in the commercialization process of innovative ideas.

12. Detailed results can be obtained from the authors upon request and are reported in an earlier version of this paper (“Does the Support of Innovative Clusters Sustainably Foster R&D Activity? Evidence from the German BioRegio and BioProfile Contests”, Quaderni della facoltà di Scienze economiche dell'Università di Lugano, No. 1105-2010).