Spatial spillovers and innovation activity in European regions

Transcription

1 Spatial spillovers and innovation activity in European regions Rosina Moreno-Serrano University of Barcelona Raffaele Paci University of Cagliari and CRENoS Stefano Usai University of Cagliari and CRENoS Aims of the paper Explore the evolution of technological activity across European regions and the determinants of innovation diffusion and creation Estimate a Knowledge Production Function (KPF) where innovation output is function of R&D, and other internal and external factors. Use the spatial econometric methodology to define the correct specification of the KPF and to assess the role of several types of spillovers. Analyse the importance of geographical proximity (both within and across national borders) in the diffusion of knowledge. Explore the role played by similarity of the technological composition between two regions in the diffusion of knowledge. 1

2 The literature behind us/1 From a theoretical point of view: knowledge and technological progress are engines of economic dynamics in most endogenous growth models (since Romer, 1986). In the spatial context this implies that local growth depends on the amount of technological activity which is carried out locally and on the ability to exploit external technological achievements known as technological spillovers (Martin and Ottaviano, 001; Coe and Helpman, 1995). Importance of geographical proximity for sharing innovations and knowledge and other local advantages (Glaeser et al, 199; Henderson, 1997, Paci and Usai, 000, Breschi and Lissoni, 000) Importance of other types of proximity and links for technological spillovers to operate (Boschma, 005, Coe, Helpman and Hoffmaister, 1997, Keller, 000, Verspagen, 000) The literature behind us/ A useful approach to analyse the process of innovation creation is the estimation of a KNOWLEDGE PRODUCTION FUNCTION (KPF) originally formalised by Griliches, 1979, and mainly applied at the firm level. It has been refocused by Jaffe, 1989, to study the geographic scope of knowledge spillovers through citations. Empirical estimations of KPF have been carried out for different levels of aggregation: For the US case (Acs et al, 1994; Audretsch and Feldman, 1996; Anselin et al, 1997, Feldman and Audretsch, 1999) For the EU case: Maurseth and Verspagen, 1999; Bottazzi and Peri, 00).

3 CRENoS database on innovation activity Updated statistical databank on regional patenting at the European Patent Office Patent applications countries in Europe (the 15 members of the pre-004 EU plus Switzerland and Norway) 175 regions (inventors residence) Up to 100 (digit ISIC) manufacturing sectors European Regions CRENoS (ID-CRENoS; Region; Nuts level) 1 AT11 Burgenland 44 A Koeln 88 FR6 Bourgogne 11 IT71 Abruzzo AT1 Niederoesterreich 45 A Muenster 89 FR Nord-Pas-De-Calais 1 IT7 Molise AT1 Wien 46 A4 Detmold 90 FR41 Lorraine 1 IT8 Campania 4 AT1 Kaernten 47 A5 Arnsberg 91 FR4 Alsace 14 IT91 Puglia 5 AT Steiermark 48 B1 Koblenz 9 FR4 Franche-Comte 15 IT9 Basilicata 6 AT1 Oberoesterreich 49 B Trier 9 FR51 Pays De La Loire 16 IT9 Calabria 7 AT Salzburg 50 B Rheinhessen-Pfalz 94 FR5 Bretagne 17 ITA Sicilia 8 AT Tirol 51 C Saarland 95 FR5 Poitou-Charentes 18 ITB Sardegna 9 AT4 Vorarlberg 5 D1 Chemnitz 96 FR61 Aquitaine 19 LU Luxembourg 0 10 BE1 Reg.Bruxelles-Cap 1 5 D Dresden 97 FR6 Midi-Pyrenees 140 NL1 Noord-Nederland 1 11 BE Vlaams Gewest 1 54 D Leipzig 98 FR6 Limousin 141 NL Oost-Nederland 1 1 BE Region Wallonne 1 55 E1 Dessau 99 FR71 Rhone-Alpes 14 NL West-Nederland 1 1 CH01 Région Lémanique 56 E Halle 100 FR7 Auvergne 14 NL4 Zuid-Nederland 1 14 CH0 Espace Mittelland 57 E Magdeburg 101 FR81 Languedoc-Roussillon 144 NO01 Oslo Og Akershus 15 CH0 Nordwestschweiz 58 F Schleswig-Holstein 10 FR8 Provence-Alpes-Cote D'Azur 145 NO0 Hedmark Og Oppland 16 CH04 Zürich 59 G Thueringen 10 FR8 Corse 146 NO0 Sør-Østlandet 17 CH05 Ostschweiz 60 DK Denmark GR11 Anatoliki Makedonia, Thraki 147 NO04 Agder Og Rogaland 18 CH06 Zentralschweiz 61 ES11 Galicia 105 GR1 Kentriki Makedonia 148 NO05 Vestlandet 19 CH07 Ticino 6 ES1 Asturias 106 GR1 Dytiki Makedonia 149 NO06 Trøndelag 0 11 Stuttgart 6 ES1 Cantabria 107 GR14 Thessalia 150 NO07 Nord-Norge 1 1 Karlsruhe 64 ES1 Pais Vasco 108 GR1 Ipeiros 151 PT11 Norte 1 Freiburg 65 ES Navarra 109 GR Ionia Nisia 15 PT1 Centro (P) 14 Tuebingen 66 ES Rioja 110 GR Dytiki Ellada 15 PT1 Lisboa E Vale Do Tejo 4 1 Oberbayern 67 ES4 Aragon 111 GR4 Sterea Ellada 154 PT14 Alentejo 5 Niederbayern 68 ES Madrid 11 GR5 Peloponnisos 155 PT15 Algarve 6 Oberpfalz 69 ES41 Castilla-Leon 11 GR Attiki 156 SE01 Stockholm 7 4 Oberfranken 70 ES4 Castilla-La Mancha 114 GR41 Voreio Aigaio 157 SE0 Oestra Mellansverige 8 5 Mittelfranken 71 ES4 Extremadura 115 GR4 Notio Aigaio 158 SE04 Sydsverige 9 6 Unterfranken 7 ES51 Cataluna 116 GR4 Kriti 159 SE06 Norra Mellansverige 0 7 Schwaben 7 ES5 Comunidad Valenciana 117 IE01 Border, Midland And Western 160 SE07 Mellersta Norrland 1 Berlin 74 ES61 Andalucia 118 IE0 Southern And Eastern 161 SE08 Oevre Norrland 4 Brandenburg 75 ES6 Murcia 119 IT11 Piemonte 16 SE09 Smaaland Med Oearna 5 Bremen 76 FI1 Ita-Suomi 10 IT1 Valle D'Aosta 16 SE0A Vaestsverige 4 6 Hamburg 77 FI14 Vali-Suomi 11 IT1 Liguria 164 UKC North East Darmstadt 78 FI15 Pohjois-Suomi 1 IT Lombardia 165 UKD North West Giessen 79 FI16 Uusimaa (Suuralue) 1 IT1 Trentino-Alto Adige 166 UKE Yorkshire And The Humber Kassel 80 FI17 Etela-Suomi 14 IT Veneto 167 UKF East Midlands Mecklenburg-Vorpommern 81 FI Aland 15 IT Friuli-Venezia Giulia 168 UKG West Midlands Braunschweig 8 FR1 Ile De France 16 IT4 Emilia-Romagna 169 UKH Eastern Hannover 8 FR1 Champagne-Ardenne 17 IT51 Toscana 170 UKJ South East Lueneburg 84 FR Picardie 18 IT5 Umbria 171 UKI London Weser-Ems 85 FR Haute-Normandie 19 IT5 Marche 17 UKK South West 1 4 A1 Duesseldorf 86 FR4 Centre 10 IT6 Lazio 17 UKL Wales 1 87 FR5 Basse-Normandie 174 UKM Scotland UKN Northern Ireland 1

4 Spatial distribution of innovation Early eighties: strong centre-periphery distribution of innovation activity Late nineties: the intensity to innovate has increased considerably over the two decades in all countries the innovations have been spreading to some more regions in the South of Europe (Spain, Southern Italy and Finland) the degree of disparities in the regional distribution of innovative activities has decreased Geography of innovative activity/ countries patents per inhabitants (annual average) variation (annual average) Country no of regions 81-8 rank rank rank rank Switzerland 7 17,6 1 6,1 1 6, 1 4,4 1 5,6 0,1 6,7 Germany 40 10,6 19, 16,8 5, 8,5-1,9 10,1 Finland 6 1,9 11 6,7 7 14,5 4 4,7 17,8 11,1 1, Sweden 8 8,5 4 10,8 17,7 4,1 4,4 7,1 7,8 Netherlands 4 5,6 5 10,6 4 11,6 5 18,4 5 9,1 1, 11,5 Luxembourg 1 9, 6, 9 8,8 8 15,6 6-5,8 5,0 14, Denmark 1,0 9 5, ,0 6 15, 7 8,9 8, 10,6 Austria 9 4, 7 8,4 6 8,8 9 1,0 8 9,7 0,7 9,8 France 4,9 6 8,4 5 9,1 7 11,9 9 7,6 1,1 6,8 Belgium,8 10 5,6 11 8, ,9 10 9,9 6,0 8,6 United Kingdom 1 4, 8 6,7 8 6,6 11 9,0 11 6,6-0,1 7,7 Norway 7 1,5 1,5 1 5, 1 7,8 1 11,4 6,1 9,5 Italy 0 1,4 1,7 1 4,6 1 6, 1 14,5,9 7,9 Ireland 0,6 14 1,6 14,8 14 5, 14 14, 7,4 15,9 Spain 15 0,1 15 0,5 15 1,1 15 1, ,0 9, 1, Greece 1 0,1 16 0, 16 0, 16 0,5 16 1,4 6,0 1,1 Portugal 5 0,0 17 0,1 17 0, 17 0, ,1 9,8 19,0 4

5 Geography of innovative activity /regions Region Nation Oberbayern Stuttgart Nordwestschweiz Zuid-Nederland Zurich Uusimaa Rheinhessen-Pfalz Darmstadt Mittelfranken Stockholm Karlsruhe Freiburg Tubingen Vorarlberg Ostschweiz Koln Zentralschweiz Unterfranken Braunschweig Sydsverige CH NL CH FI SE AT CH CH SE Pat pc ranking 18,08 9,50 1 4,1 1 8, ,40, , ,90 5 9, ,8 11 1,09 8 8, ,9 5,5 64 9, ,97 6 7,9 19 5,4 9,1 55 9,1 16 Period Pat pc ranking Pat pc ranking 9,1 6,9 1,8 7,9 6 8,9 1,8 1 18,4 1 15,5 18 7,4 4,7 5 9,4 0 19,5 10 4,9 5 6,0 4 6,8 4 6,0, 6 17,8 1 1,0 1 0, 7 1,0 8 19,9 8 16, 15 18, , , , 5 1, 19, , 14 19, ,4 1 17,4 1 19,9 9 10,7 6 14,7 1 8,0 9 6,6 57 9, 1, Pat pc ranking 50,4 1 4, 4,4 6,8 4 6,6 5 5,5 6,5 7, 8 1,0 9 0,8 10 9,6 11 9, 1 8,6 1 5,8 14 5,5 15 4,8 16 4,5 17, 18, 19,1 0 Map 1. Distribution of innovative activity in European regions, (patents per capita, annual average)

6 Spatial dependence of innovative activity Presence of a strong and positive spatial autocorrelation process in the innovative activity among contiguous areas Tab. Spatial autocorrelation in the innovative activity (Moran's I test, normal approximation) contiguity Z-value Prob Z-value Prob Z-value Prob Z-value Prob Total manufacturing 1st- order nd-order rd-order Map. Scatter for innovative activity in the European regions, (patents per capita, annual average; number of regions in parenthesis ) 6

7 The determinants of innovative activity the main determinants of the local process of innovative activity are on a blend of: internal factors - production factors (R&D) - externalities within the region (agglomeration economies, knowledge not codified, institutions) external regional factors (spillovers through trade across sectors, common markets for skilled labour, ). I RD 1 Z Z i i 1i i e i Estimation strategy OLS to check for the presence of spatial dependence (R&D, control variables) If any, ML estimation of spatial dynamic models (spatially lagged dependent variable, spatially lagged R&D) Check for different weight matrices (contiguity, distance) 7

8 log I Basic regression i, t 1 log RDi, tq log NS i, ts log MAN i, ts I = patents per capita (annual average ) 17 c1 NAT c ic i, t RD = share of GDP devoted to R&D ( ) NS = density of population ( ) MAN = share of manufacturing employment ( ) NAT = national dummies Extensions: Spatial lag dependent variable Spatial lag R&D with distance decay effects Spatial spillovers within and across national borders Spatial spillovers and technological similarities Estimation of innovative activity. Dependent variable: Log (I). OLS estimation (equation ) ML estimation (equation 4) Variables Wbin Wdist Wbin Wdist Log (RD) (0.000) (0.000) (0.000) Log (NS) (0.000) (0.000) (0.000) Log (MAN) (0.000) (0.007) (0.010) W Log (I) (0.000) (0.000) NAT dummies Yes Yes Yes R -adj AIC LM-ERR (0.001) (0.019) LM-LAG (0.000) (0.000) LR Test (0.000) (0.000) 8

9 Innovative activity with distance decay effect Dependent variable: Log (I). OLS estimation (equation 6) Log (RD) (0.000) (0.000) (0.000) (0.000) (0.000) Log (NS) (0.000) (0.000) (0.000) (0.000) (0.000) Log (MAN) (0.000) (0.000) (0.000) (0.000) (0.000) W1 Log (RD) (0.005) (0.011) (0.01) W Log (RD) (0.047) (0.056) W Log (RD) (0.489) W0-50 Log (RD) (0.000) (0.000) W Log (RD) (0.516) NAT dummies Yes Yes Yes Yes Yes R -adj AIC LM-ERR (0.85) (0.448) (0.596) (0.4) (0.477) LM-LAG (0.001) (0.006) (0.00) (0.00) (0.00) Innovative activity within and across countries Dependent variable: Log (I) OLS estimation ML estimation OLS estimation Variables Wwithin Wacross Wwithin Wacross Log (RD) Log (NS) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Log (MAN) (0.000) (0.000) (0.000) (0.000) W Log (I) 0.04 (0.007) Spatial error term () (0.000) W within Log (RD) (0.01) W across Log (RD) (0.40) NAT dummies YES YES YES YES R -adj AIC LM-ERR

10 Technological proximity WTech: matrix calculated by means of patent application data ( ) disaggregated into 101 sectors (energy and manufacturing, -digit Ateco91) for each region, where fik is the share of a particular patent class k in the total of patents of region i. P ij K f ik k1 K K fik k1 k1 f f jk 1/ jk Pij equal to unity for regions whose technological characteristics are identical Pij is zero for regions whose vector of characteristics are orthogonal The closer to unity the greater the degree of similarity of the two regions technological structure is. The spatial lag of patents constructed with this weight matrix, would imply a weighted sum of other regions patents with weights proportional to the proximity of the firms in technological space Innovative activity with tech-distance matrices Dependent variable: Log (I). OLS estimation ML estimation Variables Wtech Wtech-contiguity Wtech-distance Wtech Wtech-contiguity Wtech-distance Log (RD) (0.000) (0.000) (0.000) (0.000) Wtech 0.05 (0.000) Wtech-contiguity Log (I) 0.08 (0.000) Wtech-distance Log (I) (0.000) Controls Log(NS) (0.000) (0.000) (0.000) (0.000) Log(MAN) (0.000) (0.000) (0.005) (0.011) NAT dummies YES YES YES YES R -adj AIC LM-ERR (0.16) (0.14) (0.18) LM-LAG (0.001) (0.000) (0.000) LR Test (0.000) (0.000) (0.000) 10

11 Summary of main results/1 Importance of internal R&D expenditure on innovation activity, with a highly significant and stable coefficient. Important role played by other internal factors reflecting agglomeration economies and national characteristics. Presence of spatial association and necessity to take into account innovation spillovers arising from other regions both through patenting activity and R&D efforts. Such spatial spillovers decay with distance: strong impact of the 1st and nd order neighbours, or, using distance, up to 50 kilometres. Summary of main results/ Knowledge spillovers occur mainly between regions of the same country: national innovation systems seem to dominate the European one. Technological similarity between two regions positively influence innovation spillovers with a magnitude comparable to geographical proximity. Thus, interregional knowledge spillovers exist between regions which are close both geographically and technologically. 11