Innovation and absorptive capacity: What is the role of technological frontier?

Transcription

1 Innovaion and absorpive capaciy: Wha is he role of echnological fronier? Verònica Gombau, Agusí Segarra Universia Rovira i Virgil Deparamen d Economia, CREIP, XREAP, Grup de Recerca d Indúsria i Terriori Av. Universia, 1, Reus, Spain ABSTRACT This paper explores how absorpive capaciy affecs he innovaive performance and produciviy dynamics of Spanish firms. A firm s efficiency levels are measured using wo variables: he labour produciviy and he Toal Facor Produciviy (TFP). The heoreical framework is based on he seminal conribuions of Cohen and Levinhal (1989, 1990) regarding absorpive capaciy; and he applied framework is based on he four-sage srucural model proposed by Crépon, Dugue and Mairesse (1998) for seing he deerminans of R&D, he effecs of R&D aciviies on innovaion oupus, and he impacs of innovaion on firm produciviy. The presen sudy uses a wosage srucural model. In he firs sage, a probi esimaion is used o invesigae how he sources of R&D, he absorpive capaciy and a vecor of he firm s individual feaures influence he firm s likelihood of developing innovaions in producs or processes. In he second phase, a quanile regression is used o analyze he effec of R&D sources, absorpive capaciy and firm characerisics on produciviy. This mehod shows he elasiciy of each exogenous variable on produciviy according o he firms levels of efficiency, and hus allows us o disinguish beween firms ha are close o he echnological fronier and hose ha are furher away from i. We used exensive firm-level panel daa from 5,575 firms for he period. The resuls show ha he inernal absorpive capaciy has a srong impac on he produciviy of firms, whereas he role of exernal absorpive capaciy differs according o naure of he each indusry and according he disance of firms from he echnological fronier. Key words: R&D sources, innovaion sraegies, absorpive capaciy, echnological disance, quanile regression Conac: veronica.gombau@urv.ca, agusi.segarra@urv.ca, Acknowledgemens This research was parially funded by he Spanish Minisry of Innovaion and Science (projec ECO ) and he Consolidaed Group of Research 2009-SGR-907. An earlier version of his paper was presened a he XIV Encuenro de Economía Aplicada held in Huelva, June, We are graeful o paricipans for heir useful commens. Any errors are, of course, our own. 1

2 1. Inroducion Innovaion has been a major opic in he lieraures on growh, produciviy and indusrial organizaion for many decades. The abiliy o creae economic value by inroducing new producs o he marke, redesigning producion processes, or creaing new organizaional pracices is criical o compeiive advanage and growh for firms, indusries and counries. Joseph Schumpeer (1934) carried ou he firs rigorous sudies of he deerminans of innovaion performance a firm level, and hese sudies were followed by conribuions from Schmookler (1962), Scherer (1982), and Griliches (1979, 1994), among ohers, who prepared he ground for subsequen heoreical and applied advances. These approaches have provided he mos innovaive sudies ino imporan aspecs of hese new analyical perspecives. In he las hiry years, he sudies by Cohen and Levinhal (1989, 1990) on absorpive capaciy and he complemenariy of R&D sources a firm level have grealy simulaed research ineres in he naure and complexiy of hese issues. I is well known ha Cohen and Levinhal (1989) define absorpive capaciy as a firm s abiliy o recognize and assimilae exernal knowledge. The absorpive capaciy of he firm does no ake on a saic dimension, bu raher is dynamic and depends on he knowledge previously amassed by he firm, which in urn becomes a key facor in fuure innovaion. These auhors also highlighed he complex naure of he sources and processes ha characerize any innovaive process, which means ha he complemenariy of innovaion manifess iself in various ways beween he inernal and exernal knowledge of he firm, beween he absorpive capaciy of individuals and organizaions, and beween inernal and exernal R&D, o menion only he mos imporan aspecs. A firm s incenives o innovae are also condiioned by he naure of new knowledge, as was shown by Nelson (1959) and Arrow (1962). Nelson (1959) demonsraed ha scienific knowledge was a public asse (especially from he perspecive of basic research) and ha his characerisic limis he knowledge generaor s abiliy o appropriae is own knowledge and generaes posiive exernaliies (or knowledge spillovers) ha benefi ohers. Meanwhile, Arrow (1962) showed ha he knowledge resuling from R&D aciviies is a non-rival and non-exclusive good wih hree characerisics ha limi privae markes capaciy o opimally allocae resources in accordance wih social demands; hese characerisics are: he indivisibiliy of he informaion, he knowledge generaor s problems of appropriabiliy, and uncerainy. In his scenario, he incenives of he innovaing firm, in erms of supply, are limied by knowledge spillovers and he increased risk ha he firm mus ake on; whereas in erms 2

3 of demand, he firm s acquisiion of R&D, paricipaion in cooperaive projecs and abiliy o benefi exclusively from inernal R&D aciviies are closely relaed o is absorpive capaciy. Despie he limiaions ha firms face when rying o ensure ha hey benefi exclusively from heir R&D aciviies, many inves in large sums of money in scienific and echnological developmen. In his regard, when Rosenberg (1990) posed his quesion Why do firms do basic research (wih heir own money)? he had few doubs abou he answer. As far as he was concerned, despie he low levels of appropriabiliy regarding heir own basic research, firms sill inves in R&D, he principal reason being ha i generaes skills and abiliies ha faciliae he absorpion of exernal knowledge. Tha is, raher han being merely an asse ha realizes is value wihin he confines of he firm alongside he oher agens, inernal R&D is, for he direcors of innovaive firms, one of he essenial ways of srenghening he firm s absorpive capaciy and benefiing from he knowledge generaed by ohers. The complex naure of firms R&D and innovaion aciviies is increasingly aracing he aenion of researchers, heoriss and empiriciss. Furhermore, governmens are grealy ineresed in gradually reducing he barriers o R&D (making, buying and cooperaing) and in developing he skills in firms and insiuions o encourage creaive and innovaive environmens, all of which means ha he various effecs of R&D, innovaion and produciviy need o be sudied. In his regard, he aim of his paper is o sudy he links beween firms absorpive capaciy, heir poenial o innovae and heir levels of produciviy. Specifically, he paper will look a wo basic quesions: Wha are he main ways in which Spanish firms amass skills and improve heir capaciy o gaher and incorporae exernal knowledge?; and how does a firm s disance from he echnological fronier affec he differences beween is sources of innovaion and he improvemens in is producion? The paper will use a wo-sage srucural model. Firs i will use a probi esimaion o invesigae how R&D sources, absorpive capaciy and a vecor of a firm s individual feaures influence he firm s abiliy o carry ou produc or process innovaions. Second, a quanile regression is used o analyze he effec of R&D sources, absorpive capaciy and firm characerisics on produciviy. This mehod allows us o observe he elasiciy of each deerminan of produciviy according o a given firm s produciviy. This paper uses an exhausive daa source from he Spanish Technological Innovaion Panel (henceforh PITEC), which brings ogeher in a collaboraive 3

4 venure he Spanish Naional Insiue of Saisics (INE) and he Foundaion for Technological Innovaion (COTEC). The panel obains firm-level daa and uses a collecion mehodology ha is relaively consisen over a good number of periods. The daa are aken from he Communiy Innovaion Survey (CIS) and include informaion regarding innovaion aciviies ha is comparable wih microdaa on innovaion from many oher European counries. PITEC covers a broad range of secors, and includes he aciviies of boh manufacuring and services firms. Is principal advanage is ha i allows longiudinal daa o be obained for more han 12,000 firms for he period. This paper makes five main conribuions. Firs, we look a empirical developmen in four groups of secors in erms of he echnological inensiy of heir manufacuring processes (high and low echnological inensiy) and he knowledge inensiy of services (knowledge-inense services and oher services). Second, we analyze he labour produciviy, in accordance wih mos oher sudies on his area, bu also go a sep furher and analyze he Toal Facor Produciviy (TFP). Third, we adop a dual economeric mehodology where we sar by applying an OLS followed by quanile regression in order o observe in deail he elasiciy of he R&D sources and he deerminans of absorpive capaciy according o he firms differen produciviy levels. In his way, we can deal explicily wih he subjec of he firms disance from he echnological fronier, somehing which has hardly been ouched upon in he micro-economics lieraure. Fourh, we analyze absorpive capaciy from various perspecives, and disinguish beween hose firms ha are close o he echnological fronier and hose ha are lagging behind a a disance, anoher area ha o dae has no been explicily deal wih in he lieraure. Finally, we use a panel of daa ha allows us o make a more in-deph analysis of cross secion daa han hose carried ou hus far. The resuls show ha alhough all firms inernal absorpive capaciy behaves in he same way, heir exernal absorpive capaciy is condiioned by he characerisics of he secor o which each firm belongs and by is disance from he echnological fronier. The remainder of his paper is organized as follows. In secion 2 we review he lieraure regarding he role of innovaion sources, he facors relaed o absorpive capaciy and he disance from he echnological fronier. In secion 3 we presen a wo-sage model which uses a probi and a quanile regression o sudy how he deerminan facors of absorpive capaciy and a vecor of a firm s individual characerisics affec he likelihood of he firm carrying ou innovaions of is producs, processes or produciviy. Secion 4 describes he seps followed during he daa cleansing process and he mos imporan 4

5 feaures of he 5,575 firms ha make up he definiive sample of innovaing firms for he period. Finally, secion 5 deails he empirical resuls and secion 6 summarizes he main conclusions and discusses innovaion policies. 2. Concepual framework: he effec of R&D on produciviy Despie he proliferaion of sudies on he naure and effecs of innovaion aciviies a boh an individual and aggregae level, we sill do no have a general framework ha can provide a comprehensive answer o various quesions raised by economiss, agens and policy makers. Consequenly, his paper aims o show he main advances made in he analyical inerpreaion of he naure of R&D aciviies and he effec of hese aciviies on he efficiency levels of innovaive firms. The naure of R&D If we are o undersand he role played by a given economic facor, we need o idenify he paricular feaures ha disinguish i from oher facors. When invesigaing research aciviy and echnological developmen, sudying in deph he naure and peculiariies of knowledge is no jus advisable, i is essenial. This is he main conribuion made by Richard Nelson and Kenneh Arrow. Nelson (1959) poined ou ha science, and in paricular basic research, was a public asse, which causes considerable problems for he knowledge generaor regarding appropriabiliy and gives rise o posiive exernaliies, or spillovers, ha benefi ohers. In addiion, Arrow (1962) sudied in deail how he marke assigns resources in he ambi of R&D and innovaion. For Arrow, he resuling oupu of R&D aciviies is a non-rival and non-exclusive asse wih hree characerisics ha limi he marke s capaciy o opimally allocae resources in accordance wih social demands, hese being: he indivisibiliy of he informaion, he knowledge generaor s problems of appropriabiliy, and uncerainy, which only serves o increase he risks assumed by he acors involved. The indivisibiliy of informaion generaes economies of scale in he producion of knowledge; he problems of appropriabiliy reduce he incenives of innovaive firms o inves in R&D, resuling in subopimum R&D invesmen; and high levels of uncerainy increase he risks and expenses associaed wih R&D aciviies and cause problems in he relaions beween agens paricipaing in he same cooperaive projec. 5

6 The presence of knowledge spillovers reduces he incenives of innovaive firms o inves and cooperae in R&D projecs, and generae social exernaliies ha affec broad groups of agens (Heijs, 2001). Romer (1986, 1987, 1990) and Lucas (1988) showed ha spillovers occur because of he problems ha he innovaor encouners when rying o appropriae his own invenion. The effecs of echnological exernaliies were analyzed by Zvi Griliches in his sudy in He highlighed he magniude of he exernal economies linked o R&D, and he lack of privae invesmen from innovaive firms. Innovaive firms compensae for he lack of incenives o carry ou inernal R&D by adoping exernal sources ha range from he acquisiion of exernal knowledge o collaboraion agreemens. Alhough hese exernal sources (acquisiion and cooperaion) may be a good opion for miigaing he failings of he marke and making up for he shorfall in inernal R&D, hey also bring a series of relaional problems. According o Nooeboom (2002), he exernal sources generae some relaional problems (adverse selecion, moral risk, informaion flows and revenue exracion) because of he informaion asymmeries and he difficulies in deermining he value of he invenion and he effor invesed by he agens involved. Sources of innovaion in firms Firms have differen ways of engaging in R&D aciviies. The firs is o do i inernally, a sraegy known as he decision o make. The second is o acquire exernal echnology hrough ransacions on he open marke. In his insance, firms basically have wo opions: o acquire echnological knowledge incorporaed ino goods or asses, or o acquire echnological knowledge ha is no incorporaed ino any angible maerial, wheher his is by subconracing R&D aciviies, purchasing paens, ec. In he lieraure, his sraegy is known as he decision o buy (Veugelers and Cassiman, 1999). The hird sraegy consiss of esablishing agreemens o collaborae in R&D projecs wih oher agens (universiies, firms, suppliers, cusomers or even compeiors, ec.). In cooperaive R&D projecs, firms share resources and risks wih he aim of reducing coss and effor (Mowery e al., 1996). 1 Since he las decade of he wenieh cenury, new developmens have made clear he dual naure of R&D (i.e. i generaes knowledge and improves he abiliy of organizaions o absorb exernal knowledge) and he inerrelaions beween he various sources of R&D. For his reason, in addiion o he hree 1 The fundamenal difference beween buying and cooperaing is ha in buying here is a unilaeral relaionship (money is exchanged for R&D resuls), whereas in cooperaion each parner conribues a proporional par o he projec (Croisier, 1998). 6

7 R&D sources of making, buying and cooperaing, his ex will look a a fourh dimension linked o he dual naure of R&D and o he capaciy of firms o absorb exernal echnological knowledge (Veugelers y Cassiman, 1999). Mowery (1983) was he firs auhor o sugges he exisence of possible relaions beween he sraegies of making, buying and cooperaing. His sudy of he facors ha influence he exernalizaion of R&D aciviies shows ha he greaer a firm s echnological capaciy (which is derived from is inernal R&D aciviies), he greaer is likelihood of acquiring echnology exernally. In a subsequen sudy, Mowery and Rosenberg (1989) posied again a similar relaion beween cooperaion and he inernal developmen of R&D aciviies; on his occasion, hey demonsraed ha firms acquire more knowledge as he complemenariy beween sources increases. This area of research received was boosed by he seminal work of Cohen and Levinhal (1990) on absorpive capaciy. These auhors poined ou ha inernal R&D aciviies play a double role: on he one hand, hey are an imporan source of knowledge ha generaes innovaive ideas; and on he oher hand, hey increase a firm s capaciy o idenify, assimilae and exploi exernally available knowledge; ha is, hey increase is absorpive capaciy. On he basis of his concep, various sudies have analyzed he relaions beween inernal and exernal know-how, or in sraegic erms, he relaions beween he decisions o make, buy and cooperae. For example, Arora and Gambardella (1990, 1994) sae ha large firms wih a larger knowledge base are more acive in searching for and acquiring exernal echnology. In his regard, Lowe and Taylor (1998) found a similar relaion beween inernal R&D and he acquisiion of echnology hrough licences, ec., whereas Freeman (1991) indicaed ha firms ha possess an R&D deparmen end o make more inensive use of exernal knowledge sources. However, innovaion sraegies are no only relaed in he manner ha has jus been oulined. To efficienly exploi he acquired echnology, a firm mus assimilae and inegrae i ino is producive processes, and his requires he firm o be in possession of a good knowledge and echnological skills base. In his regard, he acquisiion of exernal know-how may incenivize raher han subsiue inernal R&D aciviy. Veugelers (1997) offers empirical evidence supporing his relaionship. He found ha boh buying echnology and cooperaing encourage he firm o carry ou inernal R&D, especially when i possesses is own echnological infrasrucure (i.e. an R&D deparmen wih full ime saff). Kaiser (2002) and Becker and Diez (2004) also arrive a similar resuls, and find ha firms ha cooperae inves more in R&D aciviies han hose ha do no. 7

8 Usually, here is empirical evidence demonsraing he role played by inernal knowledge in idenifying and acquiring exernal knowledge and, conversely, he role played by he exernal acquisiion of echnology in simulaing he developing of inernal R&D aciviies. These findings corroborae he likelihood of a firm joinly adoping various mechanisms aimed a obaining echnology; in oher words, hey confirm ha he sraegies of making, buying and cooperaing coexis in he overall business sraegy. The previous resuls led o a new research line focused on he analysis of complemenariies beween he differen innovaion sraegies. Neverheless i is imporan o poin ou ha research in his area is sill in is infancy and faces considerable mehodological difficulies (Ahey and Sern, 1998). An imporan aspec arising from he research of Milgrom and Robers (1990) 2 is ha an evaluaion of complemenariy should include a producion funcion, somehing which has been overlooked in many of he empirical sudies on complemenariy beween innovaion sraegies. A common pracice in his area of research has been o evaluae complemenariy by focusing on correlaion, a proposal originally pu forward by Arora and Gambardella (1990). By adoping such an approach, raher han producing evidence of complemenariy, hose auhors demonsraed he coexisence of sraegies. The sudies of Cassiman and Veugelers (2006) and Laursen and Saler (2006) provide empirical resuls of he complemenariy beween he inernal generaion and exernal acquisiion of he knowledge. However, hese wo sudies produced conradicory resuls. Cassiman and Veugelers (2006) found ha inernal R&D aciviies and he exernal acquisiion of knowledge have complemenary effecs on firms innovaive aciviies, whereas Laursen and Saler (2006) found ha subsiuion effecs beween he inensiy of R&D and he use of exernal knowledge sources predominae. The absorpive capaciy of innovaive firms Since he 1980s he ineres in analyzing absorpive capaciy in innovaive firms has coninued o grow. Cohen and Levinhal (1989, 1990) inroduced he concep of absorpive capaciy and highlighed he dual naure of R&D in ha i boh generaes new informaion and improves he abiliy of a firm o calibrae and incorporae exernal informaion. Rosenberg (1990) was he firs o argue ha inernal R&D is necessary o undersand how scienific and 2 Milgrom and Robers (1990) were he firs o inroduce and formally develop he concep of complemenariy. According o heir heory of supermodulariy, wo aciviies are complemenary if he change in he producion funcion ha is obained when hey are carried ou separaely is less han when hey are carried ou ogeher. 8

9 echnological informaion flows from exernal sources ino a firm. These sudies emphasize he poenial synergies beween inernal and exernal knowledge. Neverheless, Rosenberg s discussion does no deal wih he mulidimensional naure of knowledge and herefore does no fully explore he sources of hese synergies. For his reason, Arora and Gambardella (1994) proposed disinguishing beween wo ypes of knowledge. The firs ype looks a he capaciy or abiliy of a firm o evaluae exernal informaion, whereas he second ype looks a a firm s abiliy o use exernally generaed informaion. Several years laer, Cassiman and Veugelers (2000) and Arbussà and Coenders (2007), among ohers, carried ou furher research along hese lines. The firs ype of knowledge does no enail complex scienific or echnological knowledge, bu raher knowledge regarding he echnology a user level and knowledge regarding business rends. The second ype allows a firm o no only o discover echnological developmens and business rends, bu also o inegrae complex and absrac exernal knowledge ino is aciviies. Disance of firms from he echnological fronier In recen years he concep of disance from he echnological fronier has been widely employed in he macro, secorial and individual lieraures. In Indusrial Economics, his concep has been very useful for sudying firms innovaion sraegies (Griffih e al., 2004). Research has shown ha a firm s posiion in relaion o he echnological fronier is an imporan facor when deermining he inensiy and naure of innovaion in manufacuring and services firms. Neverheless, here are cerain differences among analyses carried ou a counry and firm level. For example, life cycle and enrance and exi sraegies are less imporan a counry level han a firm level. Likewise, firm mergers or akeovers have no equivalen a counry level. One of he firs conribuions o his lieraure regarding counry level was made by Gerschenkron (1962). He argued ha lagging economies find i easier o shoren he disance ha separaes hem from he echnological fronier because hey can imiae and absorb exernal echnology. This phenomenon was originally known as he advanage of backwardness. Years laer, Abramoviz (1986) qualified Gerschenkron s proposal by saing ha underdeveloped counries need o have a cerain amoun of physical, human and echnological capial o be able o use foreign echnology and o benefi from inernaional echnological exernaliies. And more recenly Griffih e al. (2004) and Cameron e al. (2005) defined he disance from he echnological fronier as he TFP of he counry on he fronier divided by he TFP of he counry under sudy. 9

10 One of he firs conribuions regarding firm level came from Coad (2008), who showed ha he highes quaniles corresponded o leading firms near o he echnological fronier, whereas he lowes quaniles corresponded o backward firms, ha is, hose furhes away from he fronier. 3. Empirical model In his secion we presen a mehod for analysing in deph he role played by a se of facors ha deermine he absorpive capaciy of firms. The wo objecives of his sudy were oulined in he inroducory pages and are: 1) o sudy he mechanisms by which he sources of R&D (making, buying and cooperaing) affec innovaion and produciviy; and 2) o analyze he role of facors linked o absorpive capaciy. To mee hese wo objecives, we use a wo-sage economeric sraegy ha follows, in par, he analyical framework proposed by Crépon, Dugue and Mairesse (1998) (CDM), which has been improved in successive revisions by Mairesse and Mohen, 2004; Kremp e al., 2004; Griffih e al., 2006 and Mohnen e al., The srucural model of CDM links produciviy wih he inpu from innovaion, and his in urn wih he firm s invesmen in R&D, hus opimizing he informaion available in he differen versions of he Communiy Innovaion Survey (CIS). The CDM srucural model has four equaions: he firs deals wih wheher firms decide o carry ou innovaion aciviies; he second deals wih he facors ha deermine invesmen in innovaion aciviies; he hird esimaes he elemens ha deermine he oupus of a firm s innovaion; and he fourh uses a Cobb Douglas producion funcion o sudy he effecs of innovaion aciviies on produciviy. This paper covers wo sages of he CMD srucural model. In he firs sage, we apply a probi model o esimae how he R&D sources, he deerminans of absorpive capaciy and a vecor of he firm s individual characerisics condiion he firm s likelihood of carrying ou produc or process innovaions. The empirical work classifies he firms ino four groups (manufacuring firms versus services firms, high knowledge inensiy versus low knowledge inensiy) and applies fixed effecs per secor (CNAE-2009 divisions) and per financial year ( ). We can express a firm s likelihood of developing an innovaion wih he following expression: inn X [1] * i i i where X i is a vecor of he deerminan facors of innovaion, β is a vecor of he parameers ha are esimaed and μ is a random error erm. In his binary model, he dependen variable he likelihood of innovaing is a laen 10

11 variable (Greene, 2003). Using inn i o denoe he binary ha indicaes ha he firm "i" carries ou innovaion, we ge: * 1 if inn i 0 inn i [2] * 0 if inn i 0 where inn is binary observable ha is equal o 1 if he firm carries ou produc or process innovaions and equal o 0 if i does no. The vecor X i includes a series of variables relaed o he sources ha firms use in heir innovaion processes (inernal R&D, exernal acquisiion of R&D and cooperaion in R&D projecs), as well as he deerminans of absorpive capaciy and a series of facors ha bring ogeher he individual characerisics relaed o he firm s produciviy. We also inroduce a vecor wih secorial dummy variables o capure he heerogeneiy across he secors. In he second sage, we use quanile regressions o esimae he impac of innovaion sources, absorpive capaciy and several of he firm s characerisics on is produciviy. Originally many researchers inerpreed R&D aciviies as a new kind of capial ha is accumulaed afer R&D invesmen, bu coninued research hen led many o observe ha knowledge sock is no only he resul of each firm s invesmen, bu also ha knowledge is a non-rival good ha allows firms o learn from he effors and innovaion carried ou by heir compeiors and he firms wih whom hey share heir surroundings. For his reason, we adop a series of assumpions based on he seminal conribuion by Arrow (1962) regarding he limiaions faced by an innovaive firm when rying o appropriae all he benefis, he accumulaed experience poenial, and he learning and exernaliies. This conribuion opened up a new horizon for inerpreing he barriers and incenives ha preven firms from invesing R&D, and several years laer would resul in endogenous growh models ha suppor he growh rajecories susained by he presence of growing incomes and exernal effecs in he producion funcions (Romer, 1986; Lucas, 1988). In paricular, we are ineresed in adoping a series of facors firs proposed by Griliches in His sudy has become a reference work for all hose ineresed in analysing he naure and effec of R&D aciviies on he produciviy of innovaive firms. A Cobb-Douglas producion funcion, broadened o include R&D aciviies, would be as follows: i i i Yi AK L Rie [3] where he firm s produc i in each period of ime, Y, is obained by means of he convenional facors of work, L, and physical capial, K, from he 11

12 knowledge or echnological capial of he firm, R, and, finally, from he rae of echnical progress relaed o he profile of he secor, he echnological spillovers and he individual characerisics of he firm, A. The parameers,, are he elasiciies of he produc in erms of physical capial, work and knowledge, respecively. Following on from Griliches (1979), he parameer capures he firm s characerisics ha affec he level of produciviy, and μ sands for any limiaions regarding he qualiy of he daa; ha is, i is an indicaor of our lack of cerainy abou heir qualiy. By applying logarihms o he producion funcion [3] we can rewrie he equaion hus: Yi, K ln( ) ln( Ai ) ln( ) ln( Ri, ) [4] L L Where he parameers and correspond o he elasiciies of he produc for each of he facors. When we incorporae a series of variables relaed o knowledge, absorpive capaciy and he firm s characerisics, we obain he following expression: ln( y ) ln( RDin ln( Size K ) 6 ln( L ) ln( RDex ) Group 7 2 ) ( Coop ln( MarkeShare 8 3 ) AC 4 ) Age 9 [5] Where for each individual firm i in each period of he ime, y is he firm s produciviy; RD in is he inensiy of he inernal R&D expendiure; RDex is he inensiy of he exernal R&D expendiure; Coop is he cooperaion wih oher parners; AC is he absorpive capaciy; Size is he size of he firm; K / L is he invesmen in erms of physical capial per employee; Group refers o membership of a group; MarkeShare is he share of he marke; and Age refers o age. Furhermore, 1, 2, 3, 4, 5, 6, 7, 8, 9 are elasiciies referring o he inensiy of inernal R&D expendiure, he inensiy of exernal R&D expendiure, cooperaion, absorpive capaciy, size, he inensiy of he invesmen, membership of a group, he marke share and age, respecively. Despie our effors o give he empirical work a rich vecor of variables o explain firm level produciviy, i should no be forgoen ha he resuls obained presen he endogenous or simulaneous biases ha are common o his ype of esimaion. To guaranee he robusness of our resuls, we have followed he proposals of Coad (2008) and Raymond e al. (2010), among ohers, and have moved he produc deerminans back one period wih regard o he dependen variable, hus adoping he following model: 12

13 ln( y AC 4 ) ln( RDin 1 ln( Size ln( MarkeShare K ) 6 ln( L ) Age 9 ) ln( RDex ) Group 7 ) ( Coop ) [6] In he following secion we presen he mos imporan aspecs of he PITEC ha is made up of he Spanish Naional Insiue of Saisics and he Foundaion for Technical Innovaion and which provided mos of our empirical daa. We also used daa from he Insiuo Valenciano de Invesigaciones Económicas (Valencian Economics Research Insiue) o deermine he sock of physical capial in each secor, daa from he Spanish Naional Insiue of Saisics o deermine sales in each secor, and he Indusrial Prices Index o deflae various variables. 4. Daa 4.1 Sample The PITEC collaes a large amoun of informaion regarding he mos imporan aspecs of a considerable number of manufacuring and services firms. One of he advanages of he PITEC over ransversal daa sources from echnical innovaion surveys is he PITEC s emporal naure, which allows i o obain much more accurae esimaions of he progress of firms and o obain much more robus daa ha beer reflec he heerogeneous naure of he firms. Neverheless, we should highligh cerain limiaions (such as subjeciviy) o some of he quesions. In his regard, how he innovaive characer of a paricular aciviy is evaluaed depends in par on he poin of view of he person answering he survey. Neverheless, he evidence offered by Mairesse and Mohnen (2004) suggess ha subjecive evaluaions of innovaion end o be consisen wih more objecive evaluaions. For he presen sudy, i is also essenial o cleanse he primary daa source and o cleanse he daabase (rea missing values, remove excessively disparae raios regarding he average secorial values, ec.). Afer cleansing he daabase, he sample was reduced from 12,813 firms o 5,575 firms. The mos imporan operaions ha were aken ino accoun hroughou he cleansing process were ha: a) he survey daa should cover he period; b) he chosen secors should be manufacuring and services, wih a disincion made beween secors wih high echnological inensiy and secors wih low echnological inensiy; c) he sample should only include firms ha 13

14 have appeared in he daabase for a leas four years; d) firms ha have no undergone a merger or akeover and e) firms wih en or more employees. 4.2 Variables and descripive saisics We measured he produciviy variable, y, in wo differen ways. Firs, we defined i as he number of sales per employee, in keeping wih mos sudies ha have deal wih his subjec, and secondly, we measured i as he Toal Facor Produciviy (TFP), because we hink ha his form of measuremen is much broader and covers differen aspecs of produciviy. Even so, we should clarify ha our sudy of he TFP only looked a manufacuring firms because we did no have he corresponding informaion for services firms. Table 1.a: Descripive saisics (average values) in (Labour Produciviy) by quaniles High-ech indusries (11,220 Obs.) 10% 25% 50% 75% 90% Labour Produciviy 52, , , , , Facors relaed o knowledge Inernal R&D 4, , , , , Exernal R&D 1, , , Cooperaion (%) Absorpive capaciy I (Exernal facors) MARKET PUBLIC OTHERS Absorpive capaciy II (Inernal facors) INTERNAL Firm characerisics Size Invesmen 5, , , , , Group (%) Share Marke (%) Age Low-ech indusries (14,204 Obs.) 10% 25% 50% 75% 90% Labour Produciviy 44, , , , , Facors relaed o knowledge Inernal R&D 1, , , , , Exernal R&D Cooperaion (%) Absorpive capaciy I (Exernal facors) MARKET PUBLIC OTHERS Absorpive capaciy II (Inernal facors) INTERNAL Firm characerisics Size Invesmen 3, , , , , Group (%) Share Marke (%) Age Source: PITEC Noes: Produciviy in euros, Inernal R&D, Exernal R&D and Invesmen in euros per employee and size in employees. 14

15 We defined he inensiy variable for inernal R&D, RD in, as he inernal R&D expendiure per employee, and hen we defined he inensiy variable for exernal R&D, RDex, as he exernal R&D expendiure per employee. The cooperaion variable, Coop, is a binary variable ha differeniaes beween firms ha do no cooperae (represened by 0) and hose ha do cooperae (represened by 1). Table 1.b: Descripive saisics (average values) in (TFP) by quaniles High-ech indusries (11,220 Obs.) 10% 25% 50% 75% 90% TFP 8, , , , , Facors relaed o knowledge Inernal R&D 4, , , , , Exernal R&D 1, , , Cooperaion (%) Absorpive capaciy I (Exernal facors) MARKET PUBLIC OTHERS Absorpive capaciy II (Inernal facors) INTERNAL Firm characerisics Size Invesmen 10, , , , , Group (%) Share Marke (%) Age Low-ech indusries (14,204 Obs.) 10% 25% 50% 75% 90% TFP 7, , , , , Facors relaed o knowledge Inernal R&D , , , , Exernal R&D Cooperaion (%) Absorpive capaciy I (Exernal facors) MARKET PUBLIC OTHERS Absorpive capaciy II (Inernal facors) INTERNAL Firm characerisics Size Invesmen 14, , , , , Group (%) Share Marke (%) Age 22, Source: PITEC Noes: Produciviy in euros, Inernal R&D, Exernal R&D and Invesmen in euros per employee and size in employees. We considered he wo ypes of absorpive capaciy described previously: he firs allows a firm o scan is immediae environmen for knowledge and he second allows i o inegrae knowledge generaed anywhere ino is own aciviies. Given ha knowledge spillovers canno be measured direcly, we used he variables ha are bes suied o defining hese wo ypes of absorpive capaciy. 15

16 Table 2: Descripive saisics (average values) in (Labour Produciviy) by quaniles Knowledge-inensive services (3,837 Obs.) 10% 25% 50% 75% 90% Labour Produciviy 18, , , , , Facors relaed o knowledge Inernal R&D 12, , , , , Exernal R&D 1, , , , , Cooperaion (%) Absorpive capaciy I (Exernal facors) MARKET PUBLIC OTHERS Absorpive capaciy II (Inernal facors) INTERNAL Firm characerisics Size Invesmen 5, , , , , Group (%) Share Marke (%) Age Oher Services (2,951 Obs.) 10% 25% 50% 75% 90% Labour Produciviy 16, , , , , Facors relaed o knowledge Inernal R&D , , , Exernal R&D Cooperaion (%) Absorpive capaciy I (Exernal facors) MARKET PUBLIC OTHERS Absorpive capaciy II (Inernal facors) INTERNAL Firm characerisics Size Invesmen 1, , , , , Group (%) Share Marke (%) Age Source: PITEC Noes: Produciviy in euros, Inernal R&D, Exernal R&D and Invesmen in euros per employee and size in employees. Following he paper by Arbussà and Coenders (2007), o sudy he firs ype of absorpive capaciy, we used he responses regarding he imporance of exernal sources of informaion o innovaion; hese sources range from marke sources of informaion (MARKET), which include cusomers and compeiors, o public insiuions (PUBLIC), which include universiies, echnological cenres and oher public research insiuions, o oher sources (OTHERS), which include conferences, scienific journals, echnical publicaions, fairs, exhibiions, ec. Given ha firms responses o hese quesions were subjecive, we believe ha hey reflec no only he degree o which knowledge is available in he secor bu also he degree of use and absorpion by he firm. Again following on from Arbussà and Coenders (2007), as a proxy for he second ype of absorpive capaciy, we used he responses o 16

17 he quesions regarding he inernal barriers ha impede innovaion in a firm (INTERNAL), hese being: a) he lack of qualified people, b) he lack of echnological informaion and c) he lack of marke informaion. On he basis of his, we used a summaed scale o consruc hese variables, which have a value of 1 o 4. In oher words, his mehod adds he informaion from various caegorical responses and provides an aggregae value ha is more in line wih he purpose of he sudy, and in doing so reduces he limiaions inheren in he use of individual indicaors ha offer a parial dimension of a complex phenomenon, in our case he absorpive capaciy. We defined he size variable, Size, as he number of employees. The invesmen variable, K / L, measures he gross invesmen of maerial asses per employee. The group variable, Group, is a binary variable ha differeniaes beween independen firms, (represened by 0) and hose ha belong o a group (represened by 1). We defined he marke share variable, MarkeShare, as he firm s sales divided by he value of he sales in is secor. We obained he secorial sales from he Spanish Naional Insiue of Saisics. The age variable, Age, covers he years beween he curren financial year and he financial year in which he firm was creaed. Tables 1.a, 1.b and 2 show he descripive saisics of he variables used hroughou he sudy. Tables 1.a and 1.b show ha for manufacuring firms, informaion from he marke, he public insiuions and oher sources is more imporan for firms ha are close o he echnological fronier. Furhermore, hese firms also have less difficuly in finding qualified saff, ec. Table 2 shows ha for firms providing knowledge inense services, informaion from he marke is more imporan o hose firms ha are closer o he echnological fronier, whereas informaion from oher sources is more imporan o hose firms ha are furher from he echnological fronier. The laer also have more difficuly in finding qualified saff, ec. If we look a firms ha provide oher services, we can see ha informaion from he marke and from oher sources is more imporan for firms ha are nearer o he echnological fronier. On he oher hand, informaion from public insiuions is more imporan for firms ha are furher away from i, and hese firms also have more difficuly in finding qualified saff, ec. 5. Resuls In his secion we discuss he mos imporan resuls from our empirical sudy. In he firs sage, we used a probi esimaion o deermine how R&D sources and absorpive capaciy affec a firm s likelihood of carrying ou innovaions of producs, processes or boh. In he second sage, we use a 17

18 quanile regression o esimae he elasiciy of produciviy regarding a series of sources ha deermine he innovaive poenial of Spanish firms. The resuls obained from he quanile regressions provide he value of each parameer for he hundred disribuion perceniles, and hus provide deailed informaion regarding he changes ha occur in a firm s elasiciy of produciviy as a resul of is posiion in relaion o he echnological fronier. The empirical evidence obained conribues grealy o he exising body of knowledge. Some auhors have esimaed he causal relaion beween R&D and innovaion and have obained ambiguous resuls. For example, Benavene (2006) found ha inense R&D in Chilean manufacuring firms had a negaive impac on innovaion; on he oher hand, Ebersberger e al. (2010) found ha inense R&D in Finnish manufacuring firms had a posiive impac on innovaion and ha he closer firms ge o he echnological fronier, he greaer his impac. In addiion, Segarra (2010) showed ha inense R&D in Spanish manufacuring firms wih high echnological inensiy and in Spanish services firms wih high knowledge inensiy had a posiive effec on produciviy, whereas he impac was negaive for Spanish manufacuring firms wih low echnological inensiy. The resuls in Table 3 indicae ha, for manufacuring firms wih eiher high or low echnological inensiy, if a firm cooperaes wih oher firms and here is an increase in he imporance given o informaion from he marke and oher sources such as conferences, ec., ha consequenly here is a significan increase in he likelihood of he firm carrying ou innovaion. Table 3: Probi esimaion on he deerminans of innovaions High-ech indusries Low-ech indusries Knowledgeinensive services Oher Services Facors relaed o knowledge Inernal R&D (0.0029) (0.0024) (0.0051)** (0.0059)*** Exernal R&D (0.0029) (0.0026)*** (0.0047)* ( (0.0069) Cooperaion (0.0510)*** (0.0457)*** (0.0817)*** (0.1055***) Absorpive capaciy I (Exernal facors) MARKET (0.0241)*** (0.0222)*** (0.0403)*** (0.0583) PUBLIC (0.0309)*** (0.0278)*** (0.0515)** (0.0708)*** OTHERS (0.0315)*** (0.0293)*** (0.0505)*** (0.0701)*** Absorpive capaciy II (Inernal facors) INTERNAL (0.0260) (0.0232) (0.0442) (0.0630)** Secorial dummies Yes Yes Yes Yes Time dummies Yes Yes Yes Yes [pseudo]r Number observaions 9,821 11,200 3,253 1,832 Source: PITEC ***Significan a 1%; **significan a 5%; *significan a 10%; sandard errors in parenheses 18

19 Regarding knowledge-inensive services, he empirical resuls show ha a firm s likelihood of innovaing increases when he firm invess more in exernal R&D, cooperaes wih oher firms and gives imporance o informaion from he marke and oher sources (conferences, meeings, fairs, ec.). Regarding firms ha provide oher services, we find ha heir likelihood of innovaing significanly increases if hey cooperae wih oher firms and give imporance o informaion from oher sources (conferences, meeings, fairs, ec.). The variables regarding firm characerisics (size of he firm, invesmen in physical asses, membership of a group, marke share and age of he firm) are highly significan and as expeced. In Tables 4.a, 4.b and 5 we find empirical evidence of quanile regressions ha link he sources and resuls of innovaion aciviies wih produciviy. As such, we can see he effecs of independen variables on he relaive efficiency of each firm. Tables 4.a, 4.b and 5 presen five esimaions (quaniles 10%; 25%; 50%; 75%; 90%). I is paricularly useful o conras he leas efficien firms, locaed far from he echnological fronier (10%), wih hose firms ha are close o he fronier (90%). The quanile regression parameers were calculaed using boosrapped sandard errors (200 reproducions). The quanile regression coefficiens can be inerpreed as he marginal change y in he quanile h caused by he marginal change in he paricular regressor ΔQ θ (y i x i ) /Δx. Specifically, Tables 4.a and 4.b show ha inernal R&D expendiure has a posiive impac on manufacuring firms wih high echnological inensiy ha are siuaed far from he echnological fronier, whereas i has a negaive impac on hose firms nearer o he fronier. In conras, exernal R&D expendiure always has a posiive impac on firms ha are far from he echnological fronier and hose ha are close o i. These resuls indicae ha inernal R&D and exernal R&D coexis in firms ha are far from he echnological fronier, whereas inernal R&D is subsiued by exernal R&D in firms close o he fronier. These resuls coincide wih hose found in sudies such as Segarra and Teruel (2011). Lokshin e al. (2008) found ha inernal R&D and exernal R&D have a posiive impac on he produciviy of manufacuring firms, bu does no classify hese firms eiher in erms of echnological inensiy or in erms of heir posiion in relaion o he echnological fronier. 19

20 Table 4.a: Quanile regressions on he deerminans of produciviy (Labour produciviy) OLS Quanile Regression 10% 25% 50% 75% 90% High-ech indusries (8,125 Obs.) Facors relaed o knowledge Inernal R&D (0.0011)** (0.0019)*** (0.0013)*** (0.0014)*** (0.0017) (0.0021)** Exernal R&D (0.0009)*** (0.0017)*** (0.0011)*** (0.0009)*** (0.0015)*** (0.0020)*** Cooperaion (0.0152) (0.0248) (0.0185) (0.0164) (0.0221) (0.0284) Absorpive capaciy I (Exernal facors) MARKET (0.0081)*** (0.0124) (0.0098) (0.0091)*** (0.0104)*** (0.0148)*** PUBLIC (0.0100)*** (0.0153) (0.0127) (0.0118)*** (0.0160) (0.0199)** OTHERS (0.0100) (0.0133) (0.0120) (0.0126) (0.0133) (0.0194) Absorpive capaciy II (Inernal facors) INTERNAL (0.0089)*** (0.0146)* (0.0109)*** (0.0105)*** (0.0113)*** (0.0196)** Firm characerisics Size (0.0077)*** (0.0132)*** (0.0096)*** (0.0102)*** (0.0122) (0.0194) Invesmen (0.0011)*** (0.0016)*** (0.0014)*** (0.0014)*** (0.0015)*** (0.0018)*** Group (0.0163)*** (0.0226)*** (0.0202)*** (0.0197)*** (0.0248)*** (0.0324)*** Share Marke (%) (0.0077)*** (0.0069)*** (0.0074)*** (0.0175)*** (0.0273)*** (0.0843)*** Age (0.0094)*** (0.0176)*** (0.0121)*** (0.0126)*** (0.0156) (0.0172)*** Secorial dummies Yes Yes Yes Yes Yes Yes Time dummies Yes Yes Yes Yes Yes Yes [Pseudo]-R Low-ech indusries (9,282 Obs.) Facors relaed o knowledge Inernal R&D (0.0009)*** (0.0015)*** (0.0010)*** (0.0010)*** (0.0012)*** (0.0016) Exernal R&D (0.0010) (0.0014) (0.0011) (0.0011) (0.0014) (0.0018)* Cooperaion (0.0159) (0.0267)** (0.0153)*** (0.0166) (0.0215) (0.0259) Absorpive capaciy I (Exernal facors) MARKET (0.0084) (0.0132) (0.0075)** (0.0102) (0.0108) (0.0130) PUBLIC (0.0106)** (0.0187) (0.0121) (0.0117) (0.0149) (0.0192) OTHERS (0.0104) (0.0154) (0.0115) (0.0119) (0.0137) (0.0184) Absorpive capaciy II (Inernal facors) INTERNAL (0.0091)*** (0.0127)*** (0.0097)*** (0.0096)*** (0.0114)*** (0.0160)*** Firm characerisics Size (0.0078) (0.0121)*** (0.0092)** (0.0099)*** (0.0148)*** (0.0219)*** Invesmen (0.0010)*** (0.0018)*** (0.0010)*** (0.0011)*** (0.0013)*** (0.0020)*** Group (0.0164)*** (0.0252)*** (0.0188)*** (0.0173)*** (0.0241)*** (0.0275) Share Marke (%) (0.0201)*** (0.0245)*** (0.0396)*** (0.0629)*** (0.1190)*** (0.1720) Age (0.0095)*** (0.0130)*** (0.0103)*** (0.0121)*** (0.0120)*** (0.0173)*** Secorial dummies Yes Yes Yes Yes Yes Yes Time dummies Yes Yes Yes Yes Yes Yes [Pseudo]-R Source: PITEC Noes: Inernal R&D, Exernal R&D, Size and Invesmen in logs; Cooperaion and Group are dummies; MARKET, PUBLIC, OTHERS and INTERNAL ake values of 1 o 4 and Marke Share is in percenages. ***Significan a 1%; **significan a 5%; *significan a 10%; sandard errors in parenheses 20