Academic knowledge externalities: spatial proximity and networks Roderik Ponds, Frank van Oort & Koen Frenken
University: a regional booster? Background and motivation
University: a regional booster? Many studies suggest existence of localised knowledge externalities (or spillovers) from academic research Importance of scientific research for innovation differs between industries impact academic knowledge externalities as well Background and motivation
Pieken organiseren zich (in RIS) Scientific and technological knowledge Innovation and valorisation Economic growth Academic institutions Firms Non-profit & Governmental agencies Regional Innovation Systems
Knowledge externalities are as localized as their mechanism are: 1.Spin-off & start-up dynamics 2.Labour mobility 3.Networks of knowledge exchange Mechanism of knowledge externalities
Informal knowledge exchange through social networks, which are mostly localized (Breschi & Lissoni 2003, 2006) Besides this, formal knowledge exchange through research collaboration: Strong growth of collaboration in processes of knowledge creation (see for example Wagner- Doebler 2001) University-industry collaboration key feature of science-based industries (eg. Pavitt 1984, Cockburn & Hendersson 1998) Networks of knowledge exchange
University-industry research collaboration not limited to regional scale (see eg. McKelvey et al. 2003) Given the importance of this mechanism in science-based technologies: network (of research collaboration) and spatial dimension necessary to analyze relation between academic knowledge externalities and regional innovation Mechanism of knowledge externalities
Collaboration: a growing phenomenon? -Share of co-publications over time-
Knowledge production function approach: regional innovation is a function of regional private and academic R&D expenditures Academic R&D can also come from other regions In two ways: a.Through localized mechanisms (from nearby regions) b.Through networks of research collaboration (from 'connected' regions) Spatial cross-regressive model: Research design
Data Focus on science-based technologies (7) in the Netherlands: 4 physical science- based and 3 life-sciences-based industries Regional innovation measured by patent intensity (EPO, ) Technology specific private and university R&D ( )
Biotechnology, /- 70% abroad
Semiconductors, /- 80% abroad
Spatial weight matrix: inverse travel time between regions i and j (cut-off point 90 minutes) Network weight matrix: intensity of research collaboration between university in region i and firms in region j Weight matrices
Research collaboration measured by co- publications between firms and universities in the relevant scientific fields ( ) Relevant scientific fields defined by analysis of citations of patents per technology to scientific journals (classified in scientific subfields) Assumption: co-publication reflects (formal) research collaboration and knowledge exchange between organisations involved. Specification of network weight matrix
1 Region 1Region 2 Region 4Region 3
Specification of network weight matrix Sending/ Receiving Region 1Region 2 Region 4Region 3 Sending/ Receiving /20 -
Number of patents – lifesciences- Negative Binominal regression (robust standard errors between parentheses) 1234 University R&D 0.287** (0.046) 0.334** (0.044) 0.313** (0.043) 0.350** (0.039) Private R&D 0.629** (0.103) 0.559** (0.097) 0.380** (0.113) 0.318** (0.110) W space 0.677** (0.157) 0.642** (0.153) W networks 0.163** (0.065) 0.155** (0.056) Dummy Agriculture & food chemistry (0.281) ( (0.240) (0.227) Dummy Biotechnology (0.297) ( (0.261) (0.220) Constant (0.292) ** ( (0.290) * (0.269) Alpha 0.867** (0.189) 0.737** ( ** (0.151) 0.597** (0.119) Cragg & Uhler's R
Knowledge production function approach (KPF) with (column standardized) spatial and relational weight matrices for academic R&D to explain regional patent intensity Pooled technologies: 3 x 40 observations life- sciences based technologies, 4 x 40 observations physical science-based technologies Technology dummies Empirical model
Number of patents – physical sciences- Negative Binominal regression (robust standard errors between parentheses) 1234 University R&D 0.234** (0.068) 0.228** (0.073) 0.183** (0.052) 0.158** (0.055) Private R&D 0.989** (0.112) 0.993** (0.115) 0.645** (0.111) 0.497** (0.101) W space (0.258) (0.374) W networks 0.188** (0.030) 0.200** (0.028) Dummy Optics ** (0.383) ** (0.383) ** (0.335) ** (0.371) Dummy Information technology ** (0.329) ** (0.333) ** (0.284) ** (0.302) Dummy semiconductors ** (0.340) ** (0.337) ** (0.295) ** (0.290) Constant (0.230) (0.338) 0.642** (0.226) (0.325) Alpha 1.189** (0.155) 1.187** (0.156) 0.919** (0.158) 0.843** (0.160) Cragg & Uhler's R
The results suggest the presence of network knowledge externalities in both life-sciences and physical sciences based technologies. Localized academic knowledge externalities seem to occur - in both technologies - within the regions where the university is located, so at a very local scale. Interregional localized externalities seem only to take place within life-sciences based technologies. Conclusions
These outcomes suggest that, within the Netherlands, academic knowledge externalities within science-based technologies cannot be easily attached to a specific spatial scale (global-local paradox). It seems that policy measures focussing on an increase of academic knowledge externalities (if necessary at all) should not be focussed on specific regions. Given the wide spatial range of these externalities, the national scale seems more appropriate. Conclusions