Brigitte Unger and Martin Zagler* Working Paper No. 74 December 2000

Transcription

1 Vienna Universiy of Economics & B.A. Deparmen of Economics Working Paper Series ORGANIZATIONAL VERSUS TECHNOLOGICAL DETERMINANTS OF INNOVATION Brigie Unger and Marin Zagler* Working Paper No. 74 December 2000 Absrac In his paper, we analyze, esimae and compare five prooypic models of innovaion, from a simple learning-by-doing model o an elaboraed model ha includes financial, organizaional and echnological deerminans of innovaion. From he comparison of he five models we conclude ha insiuions in paricular he naional sysem of innovaion and organizaions maer. The imporance of neworks and he imporance of informaion flows from differen economic acors add o he explanaory power and gives he bes model fi among all models specified. Keywords: Deerminans of Innovaion, Organizaion, Neworks, Innovaion and Knowledge, Naional Sysems of Innovaion, Endogenous Growh. JEL-Code: O41, L80, O14, O31, O33. Adress of he auhors: Vienna Universiy of Economics & B.A. Augasse 2-6, 1090 Vienna, Ausria Unger@isis.wu-wien.ac.a, Zagler@isis.wu-wien.ac.a hp:// * We are graeful o Jerald Hage, Bar Noeboom, Philippe Schmier, Henk Thomasson, Marcelo Fernandes, and Ulrike Mühlberger for inspiraion, helpful commens and suggesions.

2 1. Inroducion 1.a. Moivaion Technological facors have been he key explanaory variable for innovaions wihin he economic profession (see e.g. Freeman, 1994 for a survey). Iniially, aggregae producion funcions have been he focus of research. Exernal echnological progress, i.e. echnological innovaions, could change he oherwise fixed relaion beween inpus and maximal oupu. This seems, if a all, appropriae for process innovaions in manufacuring only. In he service secor, inpus and oupus ofen canno nealy be separaed. Furhermore, boh in he manufacuring and service secor, produc innovaions gained in imporance. In elecommunicaions, for example, mos innovaions oday are in finding new applicaions for he new echnology, in developing new sofware and services, raher han exploring new digial sysems. In bioechnology, new producs, such as red endives or new pharmaceuical drugs are developed wih he same echnology (see Tser 2000). Produc innovaions and is deerminans, herefore, should be included in he analysis of deerminans of innovaion. In line wih inerdisciplinary research on innovaion (Hage, 1998, and Hage and Hollingsworh 1999), we favor he view ha i is variey ha accouns for innovaions oday much more han echnological change. Furhermore, we argue ha i is no a genial idea alone ha accouns for an innovaion and ha i is no he sruggle of he fies ha accouns for is diffusion, bu ha ideas, innovaive and markeing aciviies ake place in a specific environmen. The lieraure on Naional Sysem of Innovaions (see e.g. Dosi, Pavi and Soee 1990, Lundvall 1992, Edquis 1997) was a breakhrough in bringing his facor ino he innovaion lieraure by focusing on he naional insiuional environmen. I argues ha naional insiuions, such as he financial sysem, he educaional sysem and governmen suppor for research influence he innovaive aciviies of firms and secors. The lieraure on business sysems added ha organizaional variables are relevan for firms o be innovaive (see Whiley 1999). I argues ha he reducion o purely echnological explanaions and individualisic compeiion fails o ake ino accoun organizaional variables, such as cooperaion beween firms, research insiues and cliens. Informaion flows beween hese economic acors can ake place in formal and informal neworks ha improve he innovaive oucome bu can also hamper i. Furhermore, i argues ha a echnological approach canno adequaely deal wih he mos expanding secor of he economy, namely, services. The following approach ries o handle his problem by linking endogenous growh

3 heory wih he innovaion lieraure on variey. We sill use producion funcions bu in a very modified way: inpus are all kinds of insiuional and organizaional facors ha influence knowledge and innovaive aciviies, oupu is variey. This allows us o ake ino accoun boh insiuional and organizaional variables sressed by he naional sysem of innovaion and business sysems lieraure. Furhermore, wih innovaion being defined as variey, i can also be applied o produc innovaions in he service secor. And, by including he sock of knowledge, i can deal wih he fac ha radiional economic facors such as physical labour are being surpassed by informaion and communicaion echnology in a modern knowledge economy (Kessels 2000) and learning sociey (Soue 2000). Innovaion in he following, herefore, will be defined as an increase in he variey of goods and services, raher han a purely echnological advance. And i can be caused by all kinds of insiuional and organizaional facors. The paper proceeds as follows. Afer a discussion of he relaed lieraure and a presenaion of he daa, we firs inroduce he model. The innovaion secor consiss of profi maximizing firms 1 under perfec compeiion, which employ researchers 2, who produce innovaions, which in our case are new produc varieies. Wages of he researchers in he innovaion secor have o be prefinanced, in his respec an innovaion can be inerpreed as an invesmen. The financial funds can eiher be provided by capial markes or from profis generaed in he goods and service secor, which operaes under imperfec compeiion. 3 In line wih he endogenous growh heory, we assume consan reurns o scale for a single firm, resuling in increasing reurns a he secor level, due o an increase in variey. Whils he assumpion of perfec compeiion in he innovaion secor is made for mere convenience and ease of exposiion, he imperfec compeiion in he goods and service secor is crucial, as profis are a necessary condiion o enable innovaions, given ha perfec compeiion would generae zero profis. The assumpion of imperfec compeiion is in line 1 Innovaions can be eiher produced by a specialized innovaion firm (e.g. Silicon Valley), or by a profi oriened deparmen wihin a firm. Innovaions here are creaed on purpose, and herefore our model canno deal wih eiher innovaions by coincidence, or he philanrophical ineres of firm owners. 2 In his model researchers are applied scieniss, working on specific innovaion asks. In his respec, he assumpion ha more researchers produce more innovaions seems plausible. I cerainly holds less for basic research. 3 Alhough we do no show i in his paper, our heoreical resuls can be derived from profi- and uiliy maximizing behavior of households and firms. These microfoundaions are elaboraed in Zagler and Unger (2000) and Zagler (1999).

4 wih empirical facs. Apar from disinguishing wo specific modes of financing of innovaion projecs, from profis versus from financial markes, we disinguish wo generic models of innovaion: echnological versus organizaional. The echnological model is based on he assumpion of a producion funcion of innovaions, where labor inpus produce new produc varieies. I is assumed ha labor produciviy increases wih he sock of knowledge, which we se proporional o he exising variey of producs. Second, we enhance he echnological model o include organizaional explanaory variables, focusing on informaion and innovaion neworks, as well as exracion raes from innovaive effor. While he innovaion secor in he firs wo models exhibis a sock of knowledge exernaliy, in he las wo models i exhibis an organizaion (neworking) exernaliy. The assumpion of scieniss devoed o research and o organize informaion flows wihin he neworks replaces he saic assumpion of scieniss producing on a given sock of knowledge. The las par of he paper empirically discriminae beween hose four explanaions of innovaion raes using economeric model selecion crieria. The four models for Innovaion Raes: Innovaions financed from Profis from Capial Markes Produced by Sock of knowledge echnological heerodox echnological orhodox Neworks and informaion organizaional heerodox organizaional orhodox 1.b. Relaed Lieraure: Insiuions and Innovaion Since he 1960s he convenional neoclassical model for echnological change was criicized for is implausible micro-foundaion. Empirical sudies on he behavior of he firm and on secors suggesed ha innovaive behavior and echnological change should be sudied wihin a naional insiuional conex (Penrose 1959, Freeman, Young and Fuller 1962). Paul David (1975) inroduced he concep of a variey of producion processes ha could be designed wih he exising sae of knowledge and he concep of localized learning which direcs echnical change. He differeniaed his approach from neoclassical producion heory by suggesing ha subsiuion in he producion funcion may involve an elemen of innovaion. This opened he door o wha has since become known as he pah dependen model of

5 innovaion (see Barzokas 2000 for a shor hisorical survey and Nelson 1995 for an exended overview). The 1980s were devoed o he caching-up debae (Abramoviz 1984). Reasons for differences in growh among counries were seen in differences in insiuions. Solow once remarked ha economiss had o become amaeur sociologiss if hey waned o undersand growh. In he 1990s insiuions challenged he orhodox paradigm from several sides. Firs, endogenous growh heory had developed a framework ha allowed o qui he orhodox framework by sressing 1. exernaliies resuling from capial and knowledge accumulaion 2. accumulaion of human capial and 3. coninuing long-erm growh generaed by he sock of codified knowledge. Wih his, insiuions became endogenous facors in he economic sysem. Furhermore, he perfec compeiive marke was replaced by he assumpion of marke failures. For a criical assessmen of endogenous growh heory see Ben Fine (2000). Bu i allowed o impor insiuions back ino economics. The heory of endogenous growh was expored ino inernaional rade heory, where i led o a highly simulaing debae and o serious modificaions of he orhodox facor equalizaion heorem and comparaive advanage heorem (see e.g. Ben David 1997). Also indusrial economics focused on iner-linkages beween firms, marke condiions and performance raher han on he orhodox neoclassical model. Research shifed from sudying deerminans of growh raes for counry groups o sudying single counries (see Temple 2000). Organizaions and insiuions have, hence, also become imporan in hese sub-fields of economics. Bu growh is no direcly linked wih innovaion. Growh does no necessarily need echnological innovaion. I can occur hrough imiaion, a fac for which he Japanese became famous. (We migh call i organizaional innovaion hough). Also, innovaion does no necessarily imply growh. Successful innovaors can be unable o reap he economic benefis of i. Fagerberg e al (2000) righly argue ha Europe needs o adap o innovaion based growh. For compuer elecronics i can be shown ha very ofen he follower reaps he benefis by improving he design, whereas he leader had o bear he coss of developing he echnology (Harmann and Teece 1990). Teece (1984) has sressed his poin a he firm level and has elaboraed firm sraegies for srong and weak appropriabiliy regimes. In srong appropriabiliy regimes innovaions could easily be kep secre, paened or proeced wih copyrighs. In weak appropriabiliy regimes conracual soluions or inegraion were necessary. Technology, an innovaing firm s complemenary asses, firm sraegic

6 managemen, neworks and insiuions (e.g. paen laws) are he deerminans of economic success idenified by Teece (1984), Teece (1986) and Teece, Pisano, Shuen (1997). The sruggle of he fies has o be replaced by he righ balance beween compeiion and cooperaion (Teece and Jorde 1989) and public policy has o focus on he firm s boundaries more han on purely echnological R&D issues. While his srand of he lieraure sresses mainly managerial and organizaional issues (see also Teece and Grindley 1997) he lieraure on naional innovaion sysems sresses insiuions. The lieraure on Naional Innovaion Sysems (see e.g. Dosi, Pavi and Soee 1990, Lundvall 1992, Edquis 1997) direcly addressed he quesion of how insiuions influenced innovaion a he firm and secor level. In paricular he educaion sysem, he financial sysem and public policy for Research and Developmen were sressed as imporan insiuions. Financial markes play an imporan role for innovaions because hey influence ownership and he exen o which here is finance for high risk invesmens. Capial marke based financing sysems allow for riskier innovaions han credi based sysems. Credi based sysems wih capial channeled largely hrough banks, end o produce more paien and growh seeking financiers, wih long erm orienaion bu also a cerain risk aversion, han do equiy financed sysems (Soskice 1991, Whiley 2000). Anoher source of financing is sill from own profis. This mode of financing is ypical for small and medium firms. Firms ha wan o keep privae propery and o avoid shareholders, ry o finance long erm invesmens from own profis, whereas loans are aken for shor erm invesmens. In he following, we disinguish financing of innovaions from own profis versus from exernal sources, wihou specifying wheher from banks or sock markes. We call financing from own profis he heerodox model, because i is assumed by Neo-Marxiss and Pos-Keynesians, financing from exernal sources he orhodox (neoclassical) model and wan o es, which model of financing explains innovaions beer. Beside financing, Guerrieri and Tylecoe (1997, p.109) find hree furher insiuions imporan for innovaion: he educaion sysem, he role of universiies and governmen suppor. Empirical work by Crouch, Finegold and Sako (1999) on he educaion sysems revealed ha i is sill differen in European counries. The vocaional raining sysem in many counries, such as Germany or Ausria, guaranees good sable workers a he medium skill level, as opposed o more exremes in Anglo-Saxon counries. In a challenging urn around of European percepions of he US, James Galbraih (2000) sresses he superioriy of he US universiy sysem over Europe for employmen and growh. He also claims ha he US has as higher governmen suppor for boh research and universiies.

7 We deal wih hese insiuional variables en bloc wihou disinguishing beween hem by using counry dummies. We call he firs wo models echnological, because hey use he sock of exising knowledge as exernally, ad hoc given echnology, in order o produce new variey of goods and services. However, by means of he counry dummies, we ake ino accoun ha he available sock of knowledge canno be exraced wihou naional insiuions. Secor dummies allow for he fac ha echnological exracion migh differ beween secors. Anoher srand of he lieraure dealing wih innovaion is organizaional sociology (Hage 1999, Whiley 2000) and inpu oupu analysis (Debresson 1998), boh poining ou a he imporance of neworks and informaion flows for innovaion. In paricular he relaion beween firms (cooperaion, conracual relaions, compeiion), beween supplier and firms, cusomer and firms, universiies and firms, he governmen and firms are considered imporan for innovaions. The lieraure on business sysems (Whiley 1999) shows ha innovaions depend boh on he naional insiuional insiuions (NIS) and on he organizaional srucure ha develops in such an insiuional environmen. Lundvall (1999) shows he parallels beween boh srands of he lieraure. In order o capure his debae we modeled wo organizaional models based on neworks and informaion flows. Again we disinguish beween he heerodox and orhodox mode of financing and marke srucure. We consider our work as a bridge beween endogenous growh heory and oher heerodox models and he non-echnical debae in socioeconomics. Clearly, we mus make some compromise in order o build such a bridge. On he one hand, we assume ha workers can move freely beween he innovaion secor and he res of he economy, hence abiliy is no a crierion o become a researcher. We have o assume ha innovaion is a predicable process under risk bu no uncerainy. Wih our simplifying assumpion on he generaion of new varieies, we absrac from large pars of he innovaion lieraure. Whils our model could in principle be exended o include he informaion flow process, i is beyond he scope of his paper. On he oher hand we have o deal wih far more variables and a complex world han neoclassical economics is used o deal wih. This shows in he variey of possible inerpreaions of our resuls. They are less deermined han orhodox and heerodox economiss would like hem o be. The esing of he models concenraes on organizaional and informaion variables. We have ried o inegrae he naional sysem of innovaion lieraure by including financial variables. Bu we have no aken oher insiuional variables explicily ino accoun, which gives rise for furher research.

8 1.c. The Daa Innovaions can be approached from he inpu side (e.g. R&D expendiures, oal innovaion expendiure), from he oupu side (e.g. paen applicaions or paens graned, lieraure based oupu indicaors idenified from scienific and rade journals, expors and sales of innovaive producs) and from survey sudies. Each indicaor has is advanage and disadvanage (see Kleinknech 1996 and Unger 2000). Secor and firm size biases can occur. For example he food indusry is high in adverisemen innovaions bu low in R&D as opposed o high ech indusries such as bioechnology. Some secors, such as compuer elecronics, do no paen bu have copyrighs. Ohers avoid paening for secrecy reasons or because hey are small firms ha have o overcome a hreshold in order o paen. Laely he OECD (1998, p.58ff) has provided useful daa on flows of echnical informaion on innovaive aciviy and on neworks of innovaion. The mos recen publicaion, upon which we base our analysis, comes from he European Union who compleed he second Communiy Innovaion Survey Sudies (CIS II) in 1996 (see Eurosa 1999). This survey sudy was done in 12 European counries (he Unied Kingdom, Belgium, France, Neherlands, Finland, Sweden, Ausria, Germany, Ireland, Luxembourg, Norway and Spain) and consised of abou 30 quesions. Spain was excluded in our analysis due o a lack of repored daa. The sample populaion was 133,284 enerprises in manufacuring and 162,778 in services. Secors were aggregaed according o he NACE classificaion. The manufacuring secor consised of six subsecors: (1) food, beverages, exile and leaher (da-dc); (2) wood, wood producs, pulp, paper, paper producs, publishing and prining (dd-de); (3) coke, refined peroleum, nuclear fuel, chemicals, chemical producs, fibers, rubber, plasic and mineral producs (df-di); (4) meal and fabricaed meal producs (dj); (5) machinery, equipmen n.e.c., elecrical and opical equipmen (dk-dl); (6) ranspor equipmen (dm-dn). The four service subsecors were (7) wholesale and commission rade, excluding moorcycles (g51); (8) financial inermediaion (j); (9) compuers and relaed aciviies, archiecural and engineering aciviies, relaed echnical consulance (k72-k742); (10) land, air and waer ranspor, ranspor by pipelines, aciviies of ravel agencies and elecommunicaions (i60-i642). Daa on profi shares and ineres raes for he firs wo models were aken from OECD publicaions. The advanage of survey daa is ha innovaions can be direcly measured and do no have o be approximaed. However, several biases can occur (see Debresson 1996). The firs

9 is he ime bias, which is due o he fac ha innovaion is a specific ac ha only occurs once and can only be laer judged as innovaive. Exper respondens will end o idenify many more innovaions wihin he mos recen decade han in he wo or hree preceding decades. In he CIS II daa, he ime bias should be reduced since firms were asked abou heir innovaive aciviies and projecs raher han abou heir percepion of innovaion. Second, all surveys of innovaion find an unexpeced high number of innovaions. Debresson (1996, p.47) quoes surveys, in which 20% o 40% of naionally mos innovaive firms also claimed ha hey were he mos innovaive in he world. One can be skepical abou respondens view regarding heir own accomplishmens and be reasonably sure ha hey are biased in claiming ha hey have done wha is good, righ and fashionable o do - innovae (Debresson 1996, p.47). Third, here ends o be a culural bias in he daa. In some counries, such as he Neherlands, respondens are more modes when crossing off iems (see Unger 1999). To give an example, when asked o indicae, which objecives were very imporan for innovaion, he Duch ended o cross off less iems han did respondens in oher counries. Since he survey daa do no correc for his, here will always be a lower percenage of Duch firms han in oher counries. The Duch also end o cross off in he middle caegory, while he Ausrians end o selec exremes. Fourh, here migh also be a secor bias in he daa. Agens in a secor which is expeced o be innovaive, such as elecommunicaion, could have a srong endency o exaggerae heir innovaive capaciy compared o hose in a secor such as food and exile ha does no have his repuaion. We ake accoun of he naional and secor bias, by using counry and secor dummy variables. Togeher, hese dummy variables can be considered as represening he naional sysem of innovaion and secor specializaion. 2. Technological Deerminans of Innovaion 2.a. The Technological Heerodox Model (Profi Finance) An economically successful produc innovaion is he ransformaion of a new business idea ino producion a a larger scale and finally he increase in he variey of available goods and services. Wih no sysemaic serious hampering of his idea-innovaion-diffusion chain and wih all firms being equally likely o innovae, he number of service and manufacuring firms will be proporional o he number of innovaions n. As exising innovaions also form he basis of knowledge, he number of innovaions n is also an index of (exploiable, available sock of) knowledge in he sociey.

10 The Innovaion Secor Innovaions require ime and effor, where we assume effor corresponds o labor inpu of scieniss in he innovaion secor, s. In accordance wih much of he sociological lieraure on innovaion, we assume ha he exising number of innovaions, capured by he index n here, has a posiive, and for he sake of simpliciy, linear impac on he creaion of new varieies of goods and services. In echnical erms, his discussion implies ha new varieies of goods and services are creaed according o, n& = φs n. (1) New varieies ( n& ) are produced by researchers (s ), building on he exising sock of knowledge (n ), wih produciviy φ. Given ha i is uncerain wheher a single innovaion will be successful, φ also measures he probabiliy of success in innovaion, when he number of aemps o innovae is large. I is very likely ha φ will differ across secors and counries. Differences beween secors such as he food and exile indusry versus elecommunicaions can be echnologically deermined. Then φ summarizes he echnological deerminans of innovaion. If i is always possible o replicae he innovaion echnology of ohers 1:1, we should no observe naional differences in φ. In his case innovaions are purely echnology driven. Oherwise, differences among counries will reflec differences in naional sysems of innovaion (cf. Dosi, Pavi, and Soee, 1990 and, Lundvall, 1992), or differences in secor composiion. Bu secor and counry variaions migh also be due o survey biases. The inclusion of he exising sock of knowledge (n ) can be moivaed in wo ways. Firs, i accouns for he knowledge economy (cf. Kessels, 2000, and Soue, 2000), saing ha knowledge becomas a more and more imporan facor of producion. Second, i enables pah dependenices (Barzokas, 2000, and Nelson, 1995), as he given sock of knowledge has a impac on he evoluion of he fuure sock of knowledge. Whils labor (scieniss) in he innovaion secor ypically have o be financially rewarded immediaely, he reurns from innovaions will only occur afer ime. Innovaive aciviies herefore inrinsically require financing. Since innovaions can be highly risky, i is no always eviden how much of is financing came o depend upon he developmen of sock markes and financial markes, or wheher financing sems from reinvesed profis wihin he firm. The laer is in line wih heerodox economiss in he Neo-Marxis or Pos-Keynesian radiion, who assumed ha firm owners reinves all profis. This assumpion sill holds oday for counries where capial markes are less developed and (junk) bond markes are absen.

11 The Goods and Service Secor Firms in he goods and service secor engage in mark-up pricing, seing prices by a consan facor over coss. Then he profi share in he goods and service secor is consan and defined o equal k, given consan reurns o scale. We assume for simpliciy ha each produc line is produced by a single firm. In his case, n represens he number of produc varieies as well as he number of firms 4. Each firm employs e workers a he wage rae w, profis d in he manufacuring and service secor are proporional o he wage sum in he secor, d k = w ne. (2) 1 k As scieniss are he only privae inpu ino he innovaion process, and assuming ha labor is homogenous and can move freely beween he hree secors, manufacuring, services, and innovaion, in he absence of financial markes all revenues of he manufacuring and service secor, d, mus equal coss, w s. 5 This implies he following accouning equaion for he innovaion secor, 1 k k n w e = w s. (3) Assuming ha workers are eiher employed in innovaion aciviies or oher aciviies, he labor marke clearing condiion reads, s n e = 1, (4) + where he labor force has been normalized o uniy. Togeher wih equaion (3), his implies ha he share of scieniss in he labor force equals he profi share in services, s = k. (5) Subsiuion ino innovaion echnology (1) gives he growh rae of innovaions, n& n = φ k. (6) Therefore, he innovaion rae depends only on he success probabiliy of innovaions and on he profi share. Evidenly, as success in innovaion is higher, he oupu of innovaion is higher. Second if more funds are available o finance researchers, i.e. a higher profi share, 4 Wihou loss of generaliy, we could assume ha each produc is produced by x firms, or ha a single firm produces a series y of produc lines, in which case we would have o muliply he righ hand side of equaion (2) by a facor x/y. 5 The assumpion ha researchers earn he same income han employees in he goods and service secor migh be unrealisic. However, we can assume ha wages are paid per producive uni, and higher skilled workers, which could all be allocaed in he innovaion secor, possess more producive unis.

12 more innovaions will occur. 2.b. Technological Orhodox Model (Full Access o Capial Markes) The inroducion of capial markes has wo effecs on he model economy. Firs, invesor can eiher inves ino a secure bond a ineres r, or ino an innovaion ha yields profis and poenial gains in resale value. Second, he exisence of a sock marke raises compeiion amongs innovaors over poenial funds. Therefore, he innovaion secor is populaed by perfecly compeiive firms, ha sell innovaions in order o maximize profis. This is equivalen o he assumpion ha firms mainain research labs as independen profi ceners. We coninue o assume, however, ha profi shares in manufacuring and he service secor remain consan. Given perfec compeiion, marginal revenues of an innovaive firm equals he sales price of a novel innovaion q o a manufacuring or service firm in a paricular produc line. Marginal coss will equal produciviy adjused wages, or w /φn. The innovaion will be sold on a bond marke, where marke efficiency leads o he following no-arbirage condiion (see Zagler/Unger 2000 for he derivaion), k we 1 k + q& = r q, (7) saing ha an invesor would be indifferen beween invesing ino a secure bond a ineres r, or ino an innovaion which yields profis and poenial gains in resale value. Aggregaing over all n service secor firms, and dividing boh sides of equaion (7) by sock marke capializaion v = n q, he no-arbirage condiion (7) reduces o, n& n φ k = ne 1 k + ( ω r ), (8) where ω is he growh rae of sock marke capializaion and he expression in parenhesis is he poenial speculaive gain on capial markes. Subsiuion of he labor marke clearing condiion (5) and innovaion echnology yields he growh rae of innovaions, n& n = φ k + ( 1 k)( ω r ). (9) The firs par of equaion (9) is idenical o equaion (6), where an increase in he success rae of innovaions and higher profis increase he innovaion rae. Hence, he firs par describes he self-financing of innovaions, whereas he second par describes he capial marke financing of innovaions. The difference beween he growh rae of sock marke capializaion (ω ) and he ineres rae indicaes he araciveness of equiy versus bond financing. Whenever ineres raes go up, he innovaion rae will decline, whils sock marke

13 booms should be beneficial for physical invesmen. The dependence of exernal finance (1 k) declines wih increasing profi shares. So far, we have presened wo very exreme cases of financial deerminans of innovaion. I is enirely feasible ha whils some innovaors face financial consrains, ohers don. Given ha equaion (6) is a special case of equaion (9), we canno disinguish beween full and parial access o financial markes in he reduced form of equaion (9). 2.c. Empirical Resuls For he esimaion, we ook he proporion of innovaive enerprises in percen of all enerprises o approximae he rae of innovaion. This is valid under he assumpion ha changes in he share of innovaive enerprises leaves he probabiliy of success for any given innovaion unaffeced. The profi share ( operaional profis ) was aken from he OECD inernaional secoral daa base. For he missing counries, we compued profi shares from he OECD indusrial srucure survey according o he formula (1 - labor coss/value added). We included daa on he long erm governmen bills and bond rae from he OECD Financial Saisics for he 11 counries in Furher research should include he growh rae of sock marke indices for all counries and secors, which we did no consider due o lack of daa. For he esimaion, we linearized equaions (6) and (9) around he seady-sae. A logisic ransformaion was applied o all variables ha were expressed as percenage shares. As we had o rejec ess for homoscedasiciy, we adoped a general leas square esimaion. Dummy variables for all counries and secors were inroduced in order o accoun for he above menioned secoral and naional differences. The Meal and Meal Produc Indusry in he UK was chosen arbirarily as he poin of counry and secoral reference. Table 1 presens he resuls. The firs column liss he independen variables, he second and hird column presen he esimaion resuls for wo versions of he profi finance model, equaion (6), whils he forh and fifh column presen he esimaion resuls for wo versions of he capial marke finance model, equaion (9). Whils he firs version of each model conains all he variables, we eliminaed sep-by-sep he leas significan variables, unil he lowes value of he Akaike informaion crierion was reached, following he general-o-specific modeling echnique (Hendry/Richard, 1983). Due o he esimaion mehod, he resuls have o be inerpreed carefully. Each field in able 1 indicaes he logi ransformed impac, or he loglikelihood, of an independen variable on he rae of innovaion, showing he -saisic in parenhesis below. Several ransformaions

14 have o be made in order o obain he effec of a change in an independen variable on he innovaion rae. Moreover, as he logi esimaion is nonlinear, we can only compue scenarios, bu canno make generalized saemens on he impac of an independen variable on he innovaion rae. The compuaion of a scenario follows a sepwise procedure. Firs, we compue he loglikelihood of a scenario according o L = β 0 + β 1 X 1 + β 2 X β n X n, where he β s are he coefficiens presened in he following ables, and he X s presen he value of he independen variables for a paricular scenario. Given he loglikelihood of a scenario, we can compue he probabiliy of innovaion P according o P = 1/(1 + e -L ). Noe ha as L goes o minus infiniy, he probabiliy goes o zero. As L goes o plus infiniy, he probabiliy goes o uniy. Wih L equal o zero, he probabiliy equals one half. Moreover, he impac of a given change in an independen variable is he greaer he closer he baseline scenario (e.g. he consan only) is o zero, due o he well-known S-shape of he logi funcion. P represens he share of firms in a given secor which have been innovaive. If hey produce on average innovaions of size m, hen he innovaion rae equals n& = Pm + (1 P) 0 = mp, hence he innovaion rae is proporional by a facor m o he innovaion probabiliy. Neiher profi shares nor ineres raes where of significance in he proceeding regressions. The laer poins o he fac ha financial consrains were no significan impedimen o innovaion. The prior can eiher be due o he qualiy of he profi share daa, a model misspecificaion, or due o he fac ha he linearizaion of equaion (9) has capured he profi share a an exremal poin, which gives suppor o he capial marke model. The consan in he regression corresponds o he innovaion produciviy φ in he meal and meal produc indusry in he UK. In order o obain he parameer for anoher indusry or counry, he value of he respecive dummy variable has o be added o he consan. To give an example, in he firs model, innovaion produciviy in he meal and meal produc indusry is in he UK, whereas i is 30 % higher, namely 1.46 in Germany. Alogeher we can conclude ha he echnological model based on he sock of knowledge does no give a good explanaion for innovaion growh and variey. The mode of financing does also no play an imporan role. Boh he profi share and ineres raes are insignifican. However, he oher non echnological facors menioned above, educaion, universiy sysem, public finance, secor specializaion and pah dependency play an imporan role for innovaion. One can see his from he las wo lines of able 1.

15 Table 1: Technological Models of Innovaion EQUATION (6) - A EQUATION (6) - B EQUATION (9) - A EQUATION (9) - B Consan (1.77) (1.22) (0.67) (0.95) Profi Share (0.12) (-0.35) (0.08) (-0.27) Ineres Rae (0.65) (-0.33) Food, Texile, Leaher (-5.53) (-5.24) (-5.66) (-5.60) Wood. Paper. Prining (-11.74) (-11.00) (-5.42) (-5.37) Chemicals (-2.91) (-2.68) (-2.31) (-2.19) Machinery and Equipmen (-7.86) (-7.64) (-4.62) (-4.53) Transpor Equipmen (-4.58) (-4.56) (-3.22) (-3.10) Wholesale and Commission Trade (-13.54) (-13.12) (-8.31) (-8.26) Financial Inermediaion (-2.85) (-2.81) (-4.75) (-3.00) Compuer Services (-8.31) (-8.34) (-3.10) (-9.66) Transporaion, Telecommunicaion (-6.15) (-6.13) (-9.69) (-4.67) Belgium (-9.72) (-9.49) (-7.14) (-3.29) Germany (2.89) (3.47) (2.44) (4.99) France (-5.07) (-4.27) (-3.00) (-7.92) Ireland (7.90) (8.44) (4.02) (2.45) Luxembourg (-0.41) * Neherlands (0.20) (0.66) Ausria (1.81) (2.22) (2.87) (3.06) Finland (-7.01) (-6.21) (-4.25) (-4.92) Sweden (-2.15) (-1.07) Norway (-3.37) (-2.88) (-2.97) (-3.17) R 2 (in %) join F-es (Counry Dummies (0.00) (0.00) (0.00) (0.00) insignifican) Akaike info crierion Schwarz crierion Coefficiens are he ransformed logi GLS coefficiens of he linearized model. -saisics (for he f-es he p-value) are in parenheses. Counry and Secor labels indicae dummies for a paricular counry or secor respecively. The reference poin is he Meal and Meal Producs Indusry in he UK. Bold numbers correspond o significance a he 5 % level, numbers in ialics a he 10 % significance level. * A weighed average of all counry dummies and he consan reproduces he ineres rae, we herefore had o forgo anoher counry. We have chosen Luxembourg for his purpose.

16 The hird o he las line indicaes he f-es of he join null hypohesis for all counry dummies. If we could no rejec his hypohesis, he echnology parameer φ would be idenical across counries, implying ha innovaion would be echnology driven. However, we have o rejec he hypohesis for all model specificaions, indicaing ha non-echnological facors play an imporan role in he deerminaion of innovaions. This gives suppor for he hypohesis of naional sysems of innovaion (Dosi, Pavi, and Soee, 1990, Lundvall, 1992). F-ess reveal ha we have o disinguish several disinc sysems of innovaion. TEST. We will elaborae his issue in he following chaper. 3. An Organizaional Deerminans of Innovaion 3.a. Innovaion Neworks We aler he previous model in wo disinc ways in order o capure wo organizaional elemens in he sociological debae on innovaion (Hage, 1998) which only recenly enered economics (see OECD 1998 and Debresson 1999). The firs novel elemen in our analysis is he idea ha he sock of knowledge canno be ranslaed ino innovaive echnology wihou cos. Innovaors mus engage in cosly aciviy o acquire knowledge by forming inernal or exernal neworks, hence, equaion (1) changes o, n& = φ s α n, η, (10) where η represens neworking capial, s n, is eiher he amoun of ime a paricular researcher devoes o he innovaion of new producs, or he number of scieniss (or science mangers) engaged in neworking aciviies, wih a declining marginal produc, given 0 < α < 1. By assumpion, innovaive aciviies have a diminishing marginal produc. Nework capial is acquired according o he following process, η 1 α η, = ψn s. (11) The ime devoed o neworking aciviies, s η, = s - s n,, also exhibis a diminishing marginal produc in erms of innovaion. The number of exising innovaions faciliaes neworking, as he number of coaliions increases wih he number of firms n. Therefore, his represens a neworking exernaliy, in conras o he previous model, where i represened a raher dubious innovaion exernaliy and in which he posiive correlaion beween new and exising innovaions was raher ad-hoc. There are wo dimensions o efficiency in innovaion. Firs, here is a radeoff beween neworking and acual innovaion aciviies. By he specificaion of our model, i is of course opimal o devoe a share 1 - α of he working ime o neworking aciviies, and he remains

17 o acual research and developmen. However, as social ineracions are harder o replicae han echnological producion schemes, his gives a firs source of innovaion inefficiency. Second, neworks may in iself be inefficien, in paricular due o a false composiion of paricipans, which we aim o capure by an informaion se ψ, which depends on he composiion of neworks in a paricular secor or counry. We will describe he composiion of his informaion se in more deail in chaper 3.c. 3.c. Exracion and Innovaion Financing The second novel feaure of he model in his chaper is exracion, i.e. he degree o which innovaive effors are ransmied ino new producs or services. We assume ha only a fracion δ of all innovaions can be markeed. The iceberg loss of exracion, (1 - δ) corresponds eiher o a ax on innovaion, or o coss generaed by organizaional inefficiency. Assuming once again ha in he absence of financial markes all profis from he manufacuring and service secor are reinvesed in he innovaion secor, he presence of iceberg coss of innovaion alers equaion (3) o, 1 k k n w e = (1 δ ) w s. (12) Togeher wih he labor marke clearing condiion (4), we can deermine he number of workers in he innovaion secor, s = k 1 δ (1 k). (13) This allows us o derive he rae of innovaion of he economy, given ha we assume an organizaional srucure in he innovaion secor, in which a fracion β of he workforce engages in innovaive aciviies, while he remaining ime is devoed o neworking aciviies, 6 n& n A = 1 δ (1 φ k. (14) k) As in equaion (6), he innovaion rae depends posiively on he success rae of innovaions and he profi share. In addiion o he echnological success of innovaion, now he appropriabiliy, or he success of markeing new producs, now plays a role in innovaion. Finally he parameer A, A = ψβ α (1 - β) 1-α, is an organizaional characerisic based on he degree a which innovaive firms allow or induce heir employees o engage in neworking aciviies eiher wihin or ouside he firm. 6 The following chaper shall idenify ha i is opimal o se β = α.

18 In analogy wih chaper 2.b., he organizaional model also is enriched by access o he financial markes. The value of he innovaion will equal marginal coss, or aggregaed over all manufacuring and service secor firms, v = w / φψ A. (15) Taking accoun of echnological exracion in he no-arbirage condiion, his allows us o solve for he innovaion rae of he manufacuring and service economy, n& n = A (1 δ )(1 k) φ k + ( 1 δ (1 k) 1 δ (1 k) ω r ). (17) The firs par of equaion (17) is idenical o equaion (14), represening he selffinancing par of innovaions. Now, as in equaion (9), he exernal financing plays a role in he deerminaion of innovaions. The difference beween he growh rae of sock marke capializaion (ω ) and he ineres rae indicaes, once again, he araciveness of equiy versus bond financing. 3.d. Empirical Resuls The coefficien β was approximaed by he fracion of innovaive enerprises ha engage in cooperaion. Wheher his was wih universiies, oher firms, cliens or suppliers was no idenified in he CIS survey. The iceberg coss δ were approximaed by he share of enerprises wih unsuccessful innovaion projecs as a percenage of all enerprises. We ried o measure neworking efficiency ψ wih a se of informaion flow variables. In he survey, innovaing enerprises were asked where he relevan informaion for heir projecs came from. Given ha differen sources of informaion will have a differen impac on he rae of innovaion, he informaion se can be insrumenalized o describe nework efficiency. In he CIS II survey, invenors were asked wheher suppliers, cliens, persons wihin he enerprise iself, or he enerprise group, compeiors, fairs and exhibiions, journals and conferences, paen disclosure and consulan firms were imporan sources of informaion for innovaion. We ran a linearized logi GLS esimaion of equaions (14) and (17), using he same poin of reference as above. As he informaion variables were only available for manufacuring and services, bu no for each secor, we had o eliminae all counry dummies, since hey were a linear combinaion of all oher independen variables. Table 2 presens he resuls. We once again approximaed he bes model fi by he general o specific mehod.

19 Table 2: Organizaional Models of Innovaion EQUATION (14) - A EQUATION (14) - B EQUATION (17) - A EQUATION (17) - B Consan (5.51) (6.22) (-3.33) (-3.34) Profi Share (0.43) (0.71) (0.39) (0.39) Ineres Rae (3.41) (3.42) Cooperaion Rae (-1.39) (-1.81) (-1.25) (-1.48) Exracion Rae (3.82) (6.83) (2.74) (2.90) Manufacuring: wihin Firm (5.31) (5.39) (3.29) (3.30) wihin Group (-7.22) (-9.34) (3.21) (3.23) from Compeiors (-0.25) (-3.25) (-3.29) from Cliens (0.71) (-3.41) (-3.42) from Consulancies (-1.57) (-1.33) (3.39) (3.40) from Suppliers (3.16) (4.79) (-3.24) (-3.25) from Paen Disclosures (3.34) (3.31) (3.38) (3.39) from Conferen-ces and Journals (7.97) 2.11 (9.19) (-3.28) (-3.29) from Fairs and Exhibiions (-1.44) (-1.87) (3.39) (3.40) Services: wihin Firm (-0.40) (2.87) (3.33) wihin Group (2.36) (2.69) (4.08) (4.11) from Compeior (0.72) (0.47) from Cliens (0.00) (1.48) (3.50) (3.84) from Consulancies (-1.25) (-0.93) (-3.43) (-3.45) from Suppliers (1.29) (0.99) (2.62) (3.29) from Paen Disclosures (-1.92) (-1.85) (1.14) (1.31) from Conferen-ces and Journals (-2.31) (-2.38) (3.23) (3.25) from Fairs and Exhibiions (2.50) 1.14 (3.93) (-3.28) (-3.29) R 2 (in %) Akaike info cri Schwarz cri Coefficiens are he ransformed logi GLS coefficiens of he linearized model. -saisics (for he f-es he p-value) are in parenheses. The reference poin is he Meal and Meal Producs Indusry. We have omied o presen dummy variables in his able. Resuls are available in he appendix Bold numbers correspond o significance a he 5 % level, numbers in ialics a he 10 % significance level.

20 Firs, we noe ha he consan is significan in all model specificaion, implying ha in his seing on average he sock of knowledge n exhibis a posiive impac on he new innovaions. However, he exremely large value for he las specificaion (equaion 17) implies ha he innovaion probabiliy is close o zero for he baseline scenario, where all independen variables are se o zero. This implies ha echnology, which we have capured by he consan hroughou his paper, canno explain he innovaion rae any more, bu ha organizaional deerminans drive he growh process (Whiely, 1999, and Soskice, 1991). Moreover, i also relaivaes he high parameer values in our las model. Due o he S-shape of he logi funcion, i requires a sharp increase in he likelihood L o obain he same probabiliy P, if he consan is negaive and large. To give an example, consider a scenario where a manufacuring firm considers o include only wihin firm informaion channels. This would increase heir innovaion rae by 1 % in model 14B. In model 17B, we would see he idenical change in he innovaion rae for an ineres rae of 10,2 %, and a correspondingly higher impac if he ineres rae is lower, and vice versa. Parameer values indicaed in equaion herefore only appear larger han in previous models, bu hey are no, which is due o he nonlineariy of he model. Second, we noe ha profi shares are insignifican in all model specificaions, for he same reasons as discussed above. Third, we find ha he cooperaion rae is insignifican in all specificaions. This suppors he view ha firms on average allocae neworking and research aciviies in an efficien manner. Forh, we find ha exracion raes are highly significan, implying ha failures o marke innovaions will foser he rae of innovaions, as i akes more aemps o succeed. Bu his of course implies ha counries or indusries wih high exracion raes will be less engaged in innovaion, or ha economic success and he poenial o innovae are only purely relaed. The laer would suppor Fagerberg e al (2000) who regre ha Europe has no enough innovaion based growh. Finally, we find ha informaion variables are imporan, more so for manufacuring han for services. 3.e. Informaion Channels Boh versions of he organizaional models included a variey of informaion flow variables, in order o picure neworks. In models A all available informaion daa have been used excep informaion from universiies which was only relevan for less han 0.5 % of innovaive enerprises. In he manufacuring secor informaion from wihin he firm and is group, from suppliers, paen disclosures and conferences is imporan for innovaion in all model specificaions. In he service secor informaion wihin he enerprise group, from conferences

21 and from fairs and exhibiions are imporan and significan in all specificaions. In addiion, we find ha in manufacuring informaion from compeiors, cliens, consulans, and fairs, and in services informaion from cliens, consulans, and suppliers urn imporan only in he specificaion of model 17. Bu he signs of he coefficiens vary. A posiive impac on innovaion in manufacuring is only provided by compeiors, in services by cliens, fairs and exhibiions. In models B only saisically significan informaion channels have been seleced. In he heerodox model 14 B (wihou ineres raes) we excluded some of he informaion channels, which were leas significan. These were in he manufacuring secor informaion from compeiors and from cliens, in he service secors informaion from compeiors and from wihin he firm. In he orhodox (capial financed) model 17B we reduced he variables o he saisically significan ones by excluding informaion from compeiors in he service secor. A comparison of he four esimaions of he organizaional model reveals ha informaion from wihin he firm and wihin he enerprise group is significan for innovaion. I is surprising ha enerprise group informaion exhibis a negaive impac in manufacuring innovaiveness in he heerodox model. This migh be due o he fac ha i is family enerprises ha rely on informaion wihin he enerprise and finance from own profis, whereas oher enerprises are more dependen on informaion from wihin he enerprise group. Moreover, inernal informaion has a negaive impac on innovaion in he service secor according o he orhodox model. Informaion from compeiors does no seem o play a posiive role for innovaiveness if so a all. Informaion from cliens is imporan for he service secor. In manufacuring i has eiher no imporance (heerodox model) or a negaive significan one (orhodox model). Informaion from consulancies are imporan for manufacuring in he orhodox model, for services hey have a negaive impac on innovaiveness. Informaion from suppliers have a posiive effec excep for manufacuring in he orhodox model. Fairs and exhibiions are very imporan for innovaiveness in manufacuring bu no for he service secor. Learning from fairs and exhibiions and from conferences sill seem o be he dominan informaion channel in boh manufacuring and services. In addiion, we found ha compeiors and informaion wihin he enerprise group are significan informaion sources in manufacuring, whereas fairs and exhibiions are imporan for he service secor in almos all models. Alogeher we can conclude ha informaion channels play quie a diverse role (cf. Hage, 1999, Debresson, 1998), depending wheher hey ake place in he service secor or in

22 manufacuring. There is almos a sign reversal in he boh secors of he role of informaion on innovaiveness. Informaion flows are also differen beween he heerodox and he orhodox model. Overall, informaion seems more imporan in he orhodox model, where innovaors have o compee and o collec informaion in order o know how o finance heir innovaions. 4. Model Selecion Given ha all models can accoun for a large par of he variaion in he dependen variable, wih R 2 beween 79 and 90 %, we now urn o selec he model, which can bes describe he daa. The las wo lines of able 1 and 2 presen he Akaike informaion crierion and he Schwarz crierion, which are boh widely used o disinguish beween differen model seups. They are boh based on he likelihood of a regression, bu penalize for he inclusion of addiional variables, he Schwarz crierion o a higher degree. The smaller boh he Akaike and Schwarz crierion are, he beer is he fi of he model. Boh he Schwarz and he Akaike informaion crierion allow us o rule ou he basic echnological models of innovaion, as presened in able 1. No only is his is in line wih he f-es ha we conduced on his class of models, bu i also confirms ha he approach o invesigae organizaional deerminans of innovaion can indeed improve our undersanding of he daa. However, we canno sraighforwardly discriminae beween he wo organizaional models of innovaion. Whils he Schwarz crierion poins owards he profi finance version of he organizaional model of innovaion, he Akaike crierion sugges organizaional model wih access o financial markes. There are wo argumens in favor of he laer model. Firs, ineres raes, which do no appear in he formulaion of he profi finance model (14), bu are an explanaory variable in he financial marke version (17), are highly significan. Second, we have used he Akaike crierion o omi regressors for he second version of each model. This implies ha we have found he minimal value of he Akaike crierion for boh models. By conras, we canno be sure ha he opimal value of he Schwarz crierion could no be lower for he financial marke version of he organizaional model of innovaions as well. Our empirical findings suppor he organizaional lieraure on innovaions and leave room for naional sysem of innovaion inerpreaions. The mode of financing should sill be enlarged by a banking secor, he educaional sysem and skill differences should be modeled as well.