Comparing knowledge bases: on the geography and organization of innovation Jerker Moodysson CIRCLE, Lund University Lecture at the Norwegian Research School in Innovation; Program in Innovation and Growth; Course on Innovation Systems, Clusters and Innovation Policy, Kristiansand, October 25, 2012 Background • Theoretical development, specification and application of the ”knowledge base” approach/typology • Publications 2004-2012, today’s presentation will focus particularly on three: – Moodysson, Coenen, Asheim 2008, Environment and Planning A – Moodysson 2008, Economic Geography – Martin & Moodysson 2012, European Urban and Regional Studies • Collective work, influenced by many (e.g. Gertler, Isaksen, Tödtling, Boschma, Manniche etc) • Roman Martin’s dissertation: Knowledge Bases and the Geography of Innovation (successfully defended Oct 2) Ambition • Better understand innovation processes in different types of economic activities • Specify when geography matters for interactive learning/innovation, in what respect, and why • Move beyond dichotomies of local/global, tacit/codified, high-tech/low-tech etc • Transcend sector classifications – less relevant for many (emerging and transforming) industries (c.f. life science, cleantech, ICT, new media etc). Low explanatory value for heterogeneity of innovation practices (also in traditional/established industries). • Combine qualitative and quantitative approaches Basic assumptions • Proximity contributes to reduced transaction costs and more efficient knowledge exchange. Spatial and relational proximity • Compatibility of knowledge (either through similarity or through relatedness) is one key aspect of relational proximity • Firms conduct routinized behaviour → they search in proximity to their existing knowledge → transcending cognitive domains requires absorptive capacity • More effective to exchange knowledge with others who share knowledge space, but only to a certain degree – optimal cognitive scope, related variety Basic assumptions Applicability of knowledge Novelty Effectiveness = novelty x communicability (non-redundant cognition) Communicability Cognitive distance Basic assumptions • Knowledge is important for innovation in all sectors, high-tech as well as low-tech. Most innovations are not ”high-tech” or ”science-based” (but still knowledge based) • Knowledge is composed by two intertwined dimensions – Codified knowledge – information. Easy to transfer over spatial distance – Tacit knowledge – we know more than we can tell. Embedded in people and organizations. Impossible to transfer over spatial distance – Knowledge always has a tacit dimension (you need tacit knowledge to interpret information) Basic assumptions Research Development Production Marketing Basic assumptions Research Knowledge Invent Potential and/or Market produce analytic design Detailed design and test Redesign and produce Distribute and market Heterogeneity • Innovation processes differ in many respects according to the economic sector, field of knowledge, type of innovation, historical period and country concerned. They also vary with the size of the firm, its corporate strategy or strategies, and its prior experience with innovation. In other words, innovation processes are ”contingent” (Pavitt, 2005, p. 87). Basis for heterogeneity • Majority of research on innovation up till the mid 2000s based explanations on two main dimensions – Sector specificities (e.g. the SIS approach) – National context (e.g. the NIS approach) • Among the most famous explanatory devices has been – the ”Pavitt taxonomy”, ultimately building on and further aggregating traditional sector classifications (Standard Industrial Classification) – the ”Varieties of Capitalism” approach, taking national institutional specificities into account (main categories LME vs CME etc) Pavitt’s taxonomy • Describe and explain similarities and differences among sectors in the sources, nature and impact of innovations • Focus on industry level – firms grouped together into an industry on the basis of their main output. Builds on traditional sector classification system (SIC/NACE etc) • Two step classification: firms firstly attributed to an industry according to their main product, and subsequently the whole industry is attributed to a class of the taxonomy (see next slide) • Empirically based (inductive) classification based on 2000 innovations in the UK 1945-1979 Pavitt’s taxonomy • Supplier dominated firms – Manufacturing, agriculture, housebuilding, financial/commercial services. Inhouse R&D/engineering capabilities weak, most innovation from suppliers • Production-intensive firms – (1) Mass production industries. Technological lead maintained through knowhow and secrecy – (2) Small-scale equipment and instrument suppiers. Firm specific skills, ability to respond sensitively to users’ needs • Science-based firms – Industries aiming to exploit scientific discoveries. R&D activities of firms in sector, underlying sciences at universities. Patents, secrecy, technical lags, firm-specific skills • Differences explained by sectoral characteristics: sources of technology (inside firms, R&D labs), users’ needs (price, performance, reliability), and means of appropriating benefits (secrets, technical lags, patents) Problems with Pavitt/sectors • The existence of multi-product and multi-technology firms • Platform technogies and emerging sectors – new ”sectors” continuously born (e.g. ICT, life science, new media etc) • Modes of innovation differ substantially between firms within sectors (Leiponen & Drejer, 2007) • Large categories of firms with very similar modes across countries and sectors (Srholec & Verspagen, 2012) • Most varience (83-95%) given by heterogeneity at firm level. Sectoral specificities explain 3-10%, national specificities 2-11% Study based on 13 035 innovating firms covering 26 sectors in 13 European countries (Srholec & Verspagen, 2012). • Alternative explanations? Knowledge bases? • (How) can the KB approach help us better understand the relation between knowledge content, modes of innovation, interaction, and relative importance of spatial and relational proximity between firms, universities and other actors in an innovation system context? • (How) can the KB approach help us better understand innovation processes carried out by firms and related actors working with different types of economic activity? • (How) can we better specify firms/activities according to the KB approach? Better than sector taxonomies? Better than the VoC-approach? The KB typology Analytical Synthetic Symbolic Understand and explain features of the (natural) world by application of scientific principles Construct solution to functional problems/ practical needs by combining knowledge and skills in new ways Trigger reactions (desire, affect etc) in minds of beholders by use of symbols and images Focus on the process rather than the outcome • Dimensions represent theoretically derived concepts rather than empirical cases • Deliberately accentuates certain characteristics (not necessarily found clear cut in reality) • Heuristics aimed to provide a systematic basis for comparison The KB typology Analytical (science based) Synthetic (engineering based) Symbolic (artistic based) Developing new knowledge about natural systems by applying scientific laws Applying or combining existing knowledge in new ways Creating meaning, desire, aesthetic qualities, affect Scientific knowledge, models, deductive Problem-solving, custom production, inductive Creative process, communication Collaboration within and between research units Interactive learning with customers and suppliers Experimentation, in studio, project teams Strong codified knowledge content, highly abstract, universal Partially codified knowledge, strong tacit component, more context-specific Interpretation, creativity, cultural knowledge, sign values, strong context specificity Meaning relatively constant between places Meaning varies substantially between places Meaning highly variable between e.g. place, class, gender Disclaimer • We are fully aware that all real cases (firms, industries, activities) draw on combinations of all three knowledge bases • Nevertheless it is possible to specify the crucial KB of a firm (or activity) i.e. the KB upon which those actors ultimately build their competitiveness (through innovation), the KB which they cannot do (innovate) without (and neither outsource) Illustration: The Astonishing Tribe Empirical illustrations • Processes and activities • Firms and industries • Discussion: next steps Application: processes and activities • Aim: Decompose innovation processes, identify and understand modes of innovation. Address the dichotomy of ‘proximate’ and ‘distant’ knowledge sourcing by looking specifically at the characteristics of the knowledge creation process • Approach: ‘innovation biographies’. Combining insights from studies of clusters and innovation systems with an activity-oriented focus • Objects of study: innovation processes in life science (pharmaceutical and functional food applications) Initial observation • Strong concentration in a few nodes. Agglomeration of (seemingly) similar firms in close proximity to Lund University • Global network connections are indispensable for novel knowledge creation among those firms • After mapping the spatial patterns of innovation (measured through formal partnerships, copatents and co-publications) we applied an intensive research design with particular focus on the actual content of the knowledge generation and collaboration Approach • Combination of theoretical reasoning, readings of the innovation literature, in-depth studies of innovation projects • Used both for theory development (i.e. further specifications of the KB approach) and for empirical analysis (i.e. explaining different spatial and organizational patterns observed) • First step of this project focused exclusively on analytical and synthetic KB Modes of knowledge creation Analytical Synthetic Understand and explain features of the (natural) world Design or construct a solution to human problems/practical needs Discovery and application of scientific laws Apply or (re)combine existing knowledge in a novel way Deconstruct natural systems Construct functional systems Know-why Know-how Formalized, scientific, standardized experimentation and abstraction Less formalized, practical experimentation and trial-and-error Analysis You start with theory. You create theoretical models with a reasonable potential to succeed in practice […] or put differently, you believe it will succeed. You then take it to the lab to test if it works. If it doesn’t work, theory is useless. Synthesis We construct and operate […] systems based on prior experiences, and we innovate in them by open loop feedback. That is, we look at the system and ask ourselves ‘How can we do it better?’ We then make some change, and see if our expectation of ‘better’ is fulfilled. The life science value chain/ problem sequence 2-4 years I 2-4 years II III 4-6 years IV V 1-3 years VI DBFs Academia Pharma I: Identification and validation of target structure (cause of disease) II: Identification and validation of biotech application (possible treatment) III: Pre-clinical tests IV: Clinical tests, phase 1 V: Clinical tests, phase 2 VI: Clinical tests, phase 3 VII: Registration and commercialisation VII Example 1 Project phase Research to understand human antibodies Development of antibody library (platform technology) Research to discover antibody based HIV drug Pre-clinical and clinical trials Dominant mode of knowledge creation Analytical Synthetic Analytical / Synthetic Analytical Actors involved Local: researchers at university department Local: University and spinn-off DBF Local: DBF Global: DBF Local: DBF Global: PRO Reveal the mechanisms of antibodies. Formalised, rational, scientific process. time Example 1 Project phase Research to understand human antibodies Development of antibody library (platform technology) Research to discover antibody based HIV drug Pre-clinical and clinical trials Dominant mode of knowledge creation Analytical Synthetic Analytical / Synthetic Analytical Actors involved Local: researchers at university department Local: University and spin-off DBF Local: DBF Global: DBF Local: DBF Global: PRO Learn how to control, select, and reproduce antibodies. Experimentation in the lab, trial and error. time Example 1 Project phase Research to understand human antibodies Development of antibody library (platform technology) Research to discover antibody based HIV drug Pre-clinical and clinical trials Dominant mode of knowledge creation Analytical Synthetic Analytical / Synthetic Analytical Actors involved Local: researchers at university department Local: University and spinn-off DBF Local: DBF Global: DBF Local: DBF Global: PRO Create a medical treatment of this tool. HIV was the selected application. A combination of analytical and synthetic mode of knowledge creation. The antigens causing HIV had to be understood; the antibodies that could block these antigens had to be defined; then they had to be selected from the ’library’. time Example 1 Project phase Research to understand human antibodies Development of antibody library (platform technology) Research to discover antibody based HIV drug Pre-clinical and clinical trials Dominant mode of knowledge creation Analytical Synthetic Analytical / Synthetic Analytical Actors involved Local: researchers at university department Local: University and spinn-off DBF Local: DBF Global: DBF Local: DBF Global: PRO Create a medical treatment of this tool. HIV was the selected application. A combination of analytical and synthetic mode of knowledge creation. The antigens causing HIV had to be understood; the antibodies that could block these antigens had to be defined; then they had to be selected from the library. Understandingtime and defining (analytical): DBF in collaboration with New Jersey firm. Selection (synthetic): spinn-off DBF in collaboration with old univ dept in Lund Example 1 Project phase Research to understand human antibodies Development of antibody library (platform technology) Research to discover antibody based HIV drug Pre-clinical and clinical trials Dominant mode of knowledge creation Analytical Synthetic Analytical / Synthetic Analytical Actors involved Local: researchers at university department Local: University and spinn-off DBF Local: DBF Global: DBF Local: DBF Global: PRO time Highly formalised. DBF in collaboration with hospitals and research institutes in Stockholm and Great Britain. Example 2 Project phase Development of probiotic in medical context Development of probiotic in commercial food context Pre-clinical and clinical trials Dominant mode of knowledge creation Synthetic Synthetic Analytical Actors involved Local: various departments at university Local: DBF and food company Local: DBF Global: PRO Medical problem: how to cure a leaking gut after surgery. Three reserchers from different disciplines (surgery, food technology, applied microbiology). Combined their skills and developed a ferment nutrient solution that could be administered by tube. time Example 2 Project phase Development of probiotic in medical context Development of probiotic in commercial food context Pre-clinical and clinical trials Dominant mode of knowledge creation Synthetic Synthetic Analytical Actors involved Local: various departments at university Local: DBF and food company Local: DBF Global: PRO A related application on the commercial market: functional food. Combine knowledge about function with knowledge about food production time Example 2 Project phase Development of probiotic in medical context Development of probiotic in commercial food context Pre-clinical and clinical trials Dominant mode of knowledge creation Synthetic Synthetic Analytical Actors involved Local: various departments at university Local: DBF and food company Local: DBF Global: PRO A related application on the commercial market: functional food. Combine knowledge about function with knowledge about food production The functional part: time a local DBF. The food part: a local food company. Very much trial and error to make these systems work togehter. Example 2 Project phase Development of probiotic in medical context Development of probiotic in commercial food context Pre-clinical and clinical trials Dominant mode of knowledge creation Synthetic Synthetic Analytical Actors involved Local: various departments at university Local: DBF and food company Local: DBF Global: PRO Highly formalised. Primarily a mattertime of getting scientific certification and support by researchers and physicians. DBF in collaboration with research institutes globally. Findings • Innovation processes involve elements of both analytical and synthetic knowledge • The characteristics of ”the core of the matter” in terms of KB differ (not only between firms and industries, but also within those) • Dominant KB (in quantitative terms) ≠ crucial KB (what the activity cannot do without) • A number of case studies in different sectors used as preliminary classification basis (this could be further developed and maybe used for more accurate “sector” classifications? Will come back to this) Application: firms and industries • Aim: Examine the geographical and organizational patterns of knowledge sourcing among firms with different crucial KB (classification of firms based on sample of case studies similar to those described above) • Research questions – What is the role of regional/global knowledge sources (for firms drawing on different crucial KB)? – What is the role of less/more formalized knowledge sources (for firms drawing on different crucial KB)? • (parts of) life science, (parts of) food, (parts of) moving media in Skåne. NB. Selection of cases not based on sector statistics. Expected patterns of knowledge sourcing (based on theoretical reasoning) global Analytical Synthetic Symbolic regional less formalized Source: own draft. more formalized 38 Expected patterns of knowledge sourcing • Knowledge sources in geographical proximity are particularly important for synthetic or symbolic firms, whereas analytical firms tend to be less sensitive to geographical distance • Formalized (scientific, codified, abstract and universal) knowledge sources are more important for analytical firms, whereas synthetic and symbolic firms rely on less formalized knowledge sources Knowledge sourcing through… • Monitoring refers to search for knowledge outside the firm, but without direct interaction with these external sources • Mobility refers to retrieving knowledge input through recruitment of key employees from other organizations (e.g. firms, universities) • Collaboration refers to exchange of knowledge through direct interaction with other actors • Network analysis based on data generated through structured interviews 40 Monitoring Mean fairs magazines surveys journals moving media food life science moving media food life science moving media food life science moving media food life science 3.00 3.11 2.72 3.19 3.07 2.83 2.44 2.86 3.31 2.31 1.86 3.31 Std. Deviation 1.29 1.40 1.39 1.39 1.27 1.34 1.25 1.30 1.51 1.21 1.08 1.31 N 36 28 29 36 28 29 36 28 29 36 28 29 Table: relative importance of various sources for gathering market knowledge through monitoring. Source: own survey. Analytical firms rely more on formalized knowledge sources than symbolic and synthetic firms. 41 Mobility Mean university technical college same industry other industries moving media food life science moving media food life science moving media food life science moving media food life science 2.94 2.11 3.93 2.26 1.89 1.90 4.36 3.96 3.87 2.61 2.93 1.77 Std. Deviation 1.45 1.23 1.55 1.15 1.20 1.40 .93 1.04 1.41 1.13 1.30 1.04 N 35 28 30 35 28 30 36 28 30 36 28 30 Table: relative importance of various sources for recruitment of highly skilled labour. Source: own survey. Analytical firms recruit primarily from universities and other firms in the same industry; synthetic and symbolic firms recruit primarily from other firms. 42 Figure: Knowledge sourcing through collaboration in media Source: own survey. Graphical illustration inspired by Plum and Hassink (2010). Figure: Knowledge sourcing through collaboration in media Source: own survey. Graphical illustration inspired by Plum and Hassink (2010). Figure: Knowledge sourcing through collaboration in media Source: own survey. Graphical illustration inspired by Plum and Hassink (2010). Figure: Knowledge sourcing through collaboration in media Source: own survey. Graphical illustration inspired by Plum and Hassink (2010). Figure: Knowledge sourcing through collaboration in food Source: own survey. Graphical illustration inspired by Plum and Hassink (2010). Figure: Knowledge sourcing through collaboration in life science Source: own survey. Graphical illustration inspired by Plum and Hassink (2010). Knowledge sourcing through collaboration 20.7% 24.5% 46.8% 24.4% 33.3% international 23.9% 54.8% moving media national regional 42.2% food 29.4% life science Table: share of regional, national and international linkages between actors Source: own survey. 49 Conclusions • Symbolic firms retrieve knowledge from less formalized sources and recruit primarily from other firms of similar type. Knowledge exchange through collaboration takes place in localized networks • Synthetic firms retrieve knowledge from less formalized sources and recruit primarily from other firms. Intentional knowledge exchange takes place on the regional and national level • Analytical firms rely on knowledge stemming from scientific research and recruitment from higher education sector. Knowledge flows and networks are very much globally configured • Findings support theoretically derived expectations Discussion: next steps • The KB approach/typology helps us do alternative and better industry classifications(?) – Compare similar industries with different KB in same regional setting (e.g. traditional vs functional food, forestry, specialty chemicals, ICT etc) – Compare different industries drawing on same KB, for verification of the robustness of the KB approach (this is partly what we have done, but could take this further) – Ultimately skip industry classifications based on characteristics on the output side (e.g. producs) and instead focus on the process side (knowledge base) – Better understaning of related variety (e.g. Asheim, Boschma, Cooke 2011) Discussion: next steps • The KB approach will benefit from cleaning the typology, avoiding circular arguments(?) – Mode and rationale of knowledge creation is the core of the matter – Spatial and social configuration of networks are expectations/empirical questions • How to deal with the challenge moving beyond qualitative approach and work with larger datasets? – Occupation data? – Professional background of entrepreneurs? – Other ideas? Contact details jerker.moodysson@circle.lu.se www.circle.lu.se