Do Companies Benefit from Public Research Organizations Do Companies Benefit from Public Research Organizations? The impact of the Fraunhofer Society in Germany Diego Comin (CEPR, Dartmouth) Georg Licht (ZEW) Maikel Pellens (ZEW, KU LEUVEN) Torben Schubert (U Lund, FhG ISI)
MOTIVATION One way: investment in applied research organizazions R&D policy has typically focused on compensating for private underinvestment Lower private cost of R&D: subsidies, tax credits Invest directly in scientific knowledge Capibility to apply R&D towards productivity growth matters too (Ortega-Argiles et al. 2014; Castellani et al. 2016) Mostly overlooked in policy (but gaining attention: EC 2017; ESIR 2017) Call for „learning-oriented“ innovation policy One way: investment in applied research organizazions Leading countries with strong APRO: NL (TNO), FI (VTT), DE (FhG), SE (RISE) But anecdotal - little empirical work studying this
RELATED LITERATURE Large body on positive effects of fiscal R&D incentives (Berger, 1993; Bloom et al. 2002; Bronzini and Piselli, 2016, van Cappelen et al. 2012, Czarnitzki et al, 2011; David et al. 2000; Hall, 2003; Knoll et al. 2014, Cowling 2016) Research on investment in public research as policy lever largely focused of universities & basic knowledge (Belderbos, 2004; Cardamone et al. 2015; Darby et al. 2004; Lööf & Broström, 2008; Maietta 2015; Monjon and Waelbroeck 2003; Mizzio and Derwick, 2002) Idea: generate spillovers to broader economy (e.g. collaboration) Well-documented positive effects Studies on extra-universitary public research organizations remain sparse Robin & Schubert (2013); Kaiser and Kuhn (2012) analyze effect on firm performance, but cannot distinguish U and other PRO
(APPLIED) PUBLIC RESEARCH Why would extra-universitary (A)PRO be different? Focus on applied research & technological co-development Closer to market, easier to appropriate Need less AC to absorb than with U (Toole et al. 2014) U and APRO collaboration with firms might have different effects Stronger direct impact Higher dissemination to smaller, less R&D intense firms Different policy rationale for funding U versus (A)PRO U: overcome public good market failure of R&D (A)PRO: overcome knowledge exploitation bottlenecks
INSTITUTIONAL CONTEXT Germany: strong extra-university research landscape Helmholtz Association, Max Planck society, Leibniz Association, Fraunhofer Society Employs 35% of PRO R&D employees Fraunhofer Gesellschaft Est. 1949 Focus on applied research & knowledge transfer 24.500 employees, budget > 2bn Contract research with industry large part of funding (30%, ~ 600 m) 67 research institutes in 7 thematic clusters: ICT, life sciences, light & surfaces, microelectronics, production, defense & security, materials
Data FhG Projects Confidential FhG contract database 131.158 contracts 1997-2014 Duration, volume, institute, description German C.I.S. Panel Information on innovation and performance Representative annual sample of German firms with at least 5 employees Select firms observed at least three times Match on name and address Exclude foreign clients as well as public institutes
FhG Projects: overview Average project lasts 8 months and costs 37k EUR (FhG revenue) 42% of collaborating firms appear once; 31% 3+ times Tendency towards applied and well-defined projects, but cover full research spectrum Keyword analysis: technology generation (75%) technology implementation (25%)
Empirical strategy RQ: how does collaboration with FhG affect firm performance? Collaboration: FhG expenditures Performance: Turnover and productivity growth rate [avg: 6.7% & 6.6%] (Productivity: added value per employee) Additionally: does firm innovation strategy change? Change in innovative sales [avg: -0.01%pt] Change in high-skilled employees [avg: +0.2%pt] Specification:
DESCRIPTIVE STATISTICS
Empirical strategy Identification challenge: unobserved selection into FhG Our approach: identification through scale heteroscedasticity Lewbel (2012) This case: first-stage heteroscedasticity driven by firm size, therefore: 𝑧 𝑖,𝑡−1 = 𝑣 𝑖,𝑡−1 ∗ ln 𝐸𝑀𝑃 𝑖, 𝑡−1 − ln 𝐸𝑀𝑃 𝑡 where 𝑣 is estimated first-stage error term.
Empirical strategy
Empirical strategy
Results Notes: * p <0.10, ** p<0.05, *** p<0.01. Regression additionally includes controls for R&D intensity, ln(age), and ln(employees), as well as exporter, group, East-German, industry, and year dummies.
Results Notes: * p <0.10, ** p<0.05, *** p<0.01. Regression additionally includes controls for R&D intensity, ln(age), and ln(employees), as well as exporter, group, East-German, industry, and year dummies.
Additional results: firm and project characteristics Firm characteristics: Small firms little benefit, but young firms profit more than older Manufacturing and services firms both benefit, but stronger productivity gains for services (less turnover gains) Also effects for low-R&D intensity firms (but not for non-R&D performers) Project characteristics: Innovation success & turnover growth driven by smaller projects focusing on new technology creation & repeated interaction Efficiency gains driven by large technology implementation projects – no additional benefits from further interations
Conclusions Empirical evidence of effects of Fraunhofer research on firm performance Unique panel data of German firms with Fraunhofer contracts Identification through scale heteroscedasticity Policy implications Support R&D diffusion alongside input Investing in APRO such as FhG seems effective channel Macroeconomic extrapolation: FhG contracts yield 2.2 bn EUR of value added compared to annual budget of 2.1 bn (upper bound estimate) Future research Further robustness testing – dynamic selection? Simulate long-term effects