Regional and Sectoral Economic Studies Vol (2012) George, General Secretariat For Research And Technology

Transcription

1 Regonal and Sectoral Economc Studes Vol (2012) INNOVATION AND ECONOMIC PERFORMANCE: the case of Greek SMEs BENEKI, Chrstna, GIANNIAS, Dmtros MOUSTAKAS, George Abstract The relatonshp between nnovaton and economc performance s estmated for Greek SMEs usng ICAP 1999 mcro data and a CDM four equaton model. The emprcal analyss shows an unwllngness of the Greek prvate sector to nvest n R&D and the low productvty of nnovaton. JEL Classfcaton: O30, O31, O32, O34, O52 Keywords: Greece; CDM-Model; Innovaton; Economc performance 1. Introducton The lnk between nnovaton and performance at varous levels of aggregaton has been the focus of attenton n a number of studes n recent decades (Loof and Heshmat, 2006). Early emprcal analyss focused on the relatonshp between research and development (R&D) actvty and productvty growth wthn a Cobb-Douglas producton functon framework (e.g. Grlches, 1979; 1984; 1995; 1998). Pakes and Grlches (1984) were the frst to develop and estmate a full model of both the knowledge producton functon and the fnal output producton functon. As an ndcator of economcally valuable knowledge they used patent counts; where, nvestments n knowledge creaton were modeled usng R&D expendtures. Crépon, Duguet and Maresse (1998) proposed and estmated a structural model for French manufacturng enterprses, whch n Lööf and Heshmat (2002) s labelled CDM. Ths combnes a knowledge-producton functon (relatng R&D actvty to patentng or nnovatve actvtes) and economc performance (measured by labor productvty). The model states that an ncrease n nnovaton nput ncreases nnovaton output and that frm s productvty ncreases along wth nnovaton output. Accordng to Hall and Maresse (2006), the fnal contrbuton of ths artcle on CDM was the development of an explct modelng framework, n order to use approprate estmaton methods n the presence of sample selectvty (due to the frm s choce of whether or not to undertake R&D), potental endogenety of some of the rght-hand sde varables, and the partally qualtatve nature of some of the dependent varables (bnary or categorcal). Many researchers adopted, n one way or another, the CDM-model framework to emprcally explore the mpact of nnovaton on productvty. Hall and Maresse (2006) presented a lst of fve selected papers 1 whch apply a model of the R&D, nnovaton, and productvty nterrelatons at the frm level, more or less smlar to that of CDM, to Benek, Chrstna, Te Ionan Islands, Gannas, Dmtros, Hellenc Open Unversty, Moustakas, George, General Secretarat For Research And Technology 1 Benavente, 2006; Loof and Heshmat, 2006; Jefferson et al., 2002; Klomp and Leeuwen, 2001; Mohnen et al., 2006

2 Regonal and Sectoral Economc Studes Vol (2012) countres as dfferent as Chle, Sweden, Chna and the Netherlands, and to a comparson of seven European countres. The CDM model has become popular n many countres among scholars usng data from the Communty Innovaton Survey (CIS). Startng n the 1990s, CIS has provded a bass for research brngng a three-stage, frm orented dataset by dstngushng between the nput, throughput and output stage of the nnovaton process (Klomp, 2001). In specfyng the emprcal model, the usual problem s that nnovaton output s not observable. Standard proxes are patents or the share of nnovatve sales. The use of data obtaned from the CIS has allowed the use of a more drect measure of the nnovatve output, whch s a bnary varable ndcatng whether an nnovaton has been ntroduced or not (Duguet, 2006; Pars et al., 2006; Maresse and Mohnen, 2005; Grffth et al., 2006; Robn and Maresse, 2008; Musoles and Huban, 2010). There s an extensve lterature that studes how publc and prvate R&D nvestments nfluence the productvty of countres (Department of trade and ndustry-dti, 2006). The majorty of the emprcal analyses rely on an extended producton-functon approach, whch ncludes R&D (or alternatve measures of nnovaton effort) as another nput to producton. Frm level studes, ncludng Lchtenberg and Segel (1991) on the US, Hall and Maresse (1995) and Maresse and Mohnen (2005) on France, Klette and Johansen (1996) on Norway, Lott and Santarell (2001), Pars et al. (2006) on Italy, Lööf and Heshmat (2002) on Sweden, Bönte (2003) on Germany, van Leeuwen and Klomp (2006) on The Netherlands, Janz et al. (2004) on Germany and Sweden, and Zacharads (2004) on 10 OECD countres, provded ndcatons of the correlaton between R&D and productvty. Moreover, the econometrc results of Grffth et al. (2006) suggested that process nnovaton s only assocated wth hgher productvty n France; n the other countres there s no such connecton. Product nnovaton was assocated wth hgher productvty n France, Span, and the UK, but not n Germany. The study of Robn and Maresse (2008) (re-)nvestgated the lnks between R&D, nnovaton and productvty usng French data from CIS3 and CIS4. Fndngs resulted that product nnovaton s assocated wth a hgher labour productvty. The effect of process nnovaton appears to be weak. Bloch (2008) found that the nterac ton of customer drven nnovaton and nnovaton output has a postve sgnfcant mpact on productvty. The collaboraton varable s also not sgnfcant n the productvty equatons. Hall, Lott and Maresse (2009) nvestgated the relatonshp between nnovaton and productvty specfcally wth Italan small and medum-szed frms by tryng to mprove on the CDM orgnal specfcaton. Results showed that frm sze and R&D ntensty, along wth nvestment n equpment, have a postve mpact on frm s productvty, especally process nnovaton. Musoles and Huban (2010) found that nnovaton s frequent n Knowledge Intensve Busness Servces and has a strong and postve effect on productvty, usng France mcro-data. More recently, Huergo and Moreno (2010) adapted also a varaton of the CDM-model for a sample of Spansh manufacturng frms between 1990 and Fndngs reflected the exstence of true state dependence both n the decson of R&D nvestment and n the producton of nnovatons. Last but not least, the am of the research of Vakhtova and 32

3 Benek,C., Gannas,D. Moustakas,G. Innovaton And Economc Performance of Greek SMEs Pavlenko (2010) was to estmate a lnk between nvestment n nnovatons and productvty n Ukranan manufacturng frms. The fndngs suggested that only the effectve government nnovaton polcy may actual postvely contrbute to the productvty. In ths study, the CDM-model framework addresses the relatonshp between nnovaton and economc performance n Greek frms. To our knowledge, ths s the frst attempt for estmaton, usng CDM-model, wth frm-level data from Greece. The paper s organzed as follows. Frstly, the CDM-model s presented. Secton 3 provdes data set descrpton and outlnes the econometrc model behnd the study and the estmaton process. Secton 4 presents the estmaton results. The fnal secton concludes the dscusson. 2. The Theoretcal Model The CDM model s a four-equaton knowledge producton functon model whch ncorporates three relatonshps: the productvty equaton relatng nnovaton output to productvty, the knowledge producton functon (or nnovaton equaton) relatng nnovaton nput to nnovaton output, and the research nvestment equaton lnkng nnovaton nput to ts determnants. An addtonal equaton concerns nvestment decsons (Crépon et al., 1998; Lööf and Hesmat, 2002). The structure of the CDM- model The nvestment model: In order to descrbe frms research behavor, we use a model wth two equatons. In the frst equaton, frms decde whether to engage n research actvtes or not (the selectvty equaton) and n the second, frms decde how much to nvest n research actvtes condtonal that they have already decded to make an nvestment (the research nvestment relaton). That s: g x u wth g 1 f g 0, else g 0 (1) k / g wth x 1 u k f k 0, else k 0 (2) k where, g s a dependent varable (decson varable) for the frm, that s the decson of frm to engage or not n R&D actvtes, k s the captal (captal used n R&D) of frm used n R&D and x s a vector of explanatory varables. Consder that g and k are the correspondng observable measures. The β 0 and β 1 are vectors of unknown parameters to be estmated reflectng the mpact of certan factors on the probablty on engagement n nvestment and the actual level of R&D nvestment. The u and 0 u are random error terms wth mean zero, constant varances and not correlated wth the explanatory varables. 1 33

4 Regonal and Sectoral Economc Studes Vol (2012) The Innovaton Equaton: The Innovaton equaton, descrbes the procedure takng place n order to transform nnovaton nput, such as R&D captal and actvtes, to economc effectve nnovaton output. Ths s a procedure to create Knowledge and Innovaton called, too, The Knowledge Producton Functon, Grlches (1994). Ths equaton s specfed below. t k x u, (3) k where t s the nnovaton output, k vector of explanatory varables (nnovaton nput) and s the elastcty of nnovaton output, 2 x s the u s the random error term wth mean zero, constant varance and not correlated wth the explanatory varables. The Productvty Equaton: The last equaton of CDM-model s the frm s economc effcency functon and descrbes the relaton between nnovaton output and the frm s economc performance. Ths functon s an enrched Cobb-Douglas producton functon descrbed as: 3 3 q t x u, (4) where t 3 q s value added from nnovatve products of frm, of scale wth respect to nnovaton output, 3 2 t s the effcency elastcty x s the usual vector of explanatory varables 3 n a common Cobb-Douglass producton functon and u s the error term. The basc econometrc problems that the CDM solves are selectvty and smultanety bases. When only the nnovaton sample s used, the frms are not randomly drawn from the larger populaton, and selecton bas may arse. Therefore, t s napproprate to lmt the focus only on nnovatve frms, and the method suggested by CDM adds a selecton equaton (Lööf and Hesmat, 2002). 3. Methodology Data and descrpton of varables Our sample conssts of Greek enterprses that responded n 1999 to the survey Measurement of R&D actvtes n enterprses 2. The nformaton was collected through the use of a structure standardzed questonnare provded by EUROSTAT. Responsble for the collecton of data n the last survey was ICAP SA (ICAP). The samplng procedure was Stratfed Random Samplng. The stated objectves of the survey were: a) the collecton of relable data assessng prvate R&D expendture b) the measurement of the real partcpatng personnel n R&D actvtes. In order to dstngush frms by ther nnovatve nature, we selected a sample of 510 nnovatve frms. 3 Innovatve frms, accordng to our defnton, consst of frms wth 2 Ths was funded by the General Secretarat of Research and Technology (GSRT) n Greece. The survey had a natonal coverage and had been conducted rregularly from 1986 to 1999 (sx tmes). The last survey covered the year 1999 and s used n ths artcle. 3 The ICAP dataset ncludes data of 3841 frms of them had non-negatve workng cycle and from them the 510 were nnovatve frms. 34

5 Benek,C., Gannas,D. Moustakas,G. Innovaton And Economc Performance of Greek SMEs postve expendtures for R&D. The key economc varables of our study nclude total personnel, gross captal, workng cycle, nvestments, total personnel s revenue, net proft (before tax), number of patents requested, prvate sector s expendtures for R&D, publc sector s expendtures for R&D, number of researchers, and total R&D personnel. Total R&D personnel s expressed as the sum of the number of researchers, the number of techncans and the number of other supportng personnel, and publc sector s expendtures for R&D as the sum of government s and European Communty s expendtures for R&D, respectvely. The abbrevatons of the varables are shown n Table 1. Table 1. Descrpton of Varables Abbrevaton Measure Abbrevaton Measure TP Total Personnel PR Net Proft (Before tax) K Gross Captal BRD Prvate Sector s Expendtures for R&D WC Workng Cycle GRD Publc Sector s Expendtures for R&D IN Investment SA Total Personnel s Revenue APN Number of Patents TRP Total R&D Personnel Requested RP Number of researchers Descrptve statstcs For jont sample of nnovatve and non-nnovatve frms and separately for only nnovatve frms, smple summary statstcs of the data are presented n Tables 2 and 3. Table 2. Summary statstcs of data: Total frms (N=1.469) Varables Mean Std. Dev. Mnmum Maxmum TP K E+07 SA WC E+07 IN PR BRD GRD RP TRP APN

6 Regonal and Sectoral Economc Studes Vol (2012) Table 3. Summary statstcs of data: Innovatve frms (N=510) Varables Mean Std. Dev. Mnmum Maxmum TP K E+07 SA WC E+07 IN PR BRD E GRD RP TRP APN The It s shown that for fve varables (TP, K, WC, PR, GRD) the mean values of Greek nnovatve frms are about twce hgher than those of total frms n Greece. Furthermore, for two varables (SA, TRP) the mean values of Greek nnovatve frms are approxmately three tmes greater than those of total frms, and the average nvestment of nnovatve frms s 2.5 tmes greater than that of total frms. Fnally, the average number of patents of nnovatve frms s almost the same wth (slghtly smaller than) wth that of total frms. Model Specfcaton The econometrc analyss of ths study follows the basc dea of Crépon et al. (1998) artcle. Our model conssts of four equatons, two of whch are estmated by two step approach, namely tobt estmaton n the frst step 4 followed by OLS estmaton n the second step, and the last two equatons of whch are estmated n a smultaneous equaton system framework. In estmatng the system of equatons we have adopted the Seemngly Unrelated Regresson (SUR) technque 5. Next we present the specfcatons of the four equatons of the CDM. Selectvty Equaton 0 The explanatory varables x of the selectvty equaton consst of number of total personnel and publc sector s expendtures n R&D. Research Investment Relaton 1 The determnants of nnovaton nput labeled as x vector consst of the number of researchers, publc sector s expendtures n R&D, and net profts. The research nvestment equaton has prvate sector s expendtures n R&D as dependent varable. 4 The method of Tobt elmnates the problem of selectvty bas we would have f we used a sample wth only companes that nnovate. Instead we take the estmated values from the selectvty equaton to correct the error n the research nvestment relaton and, subsequently, that n nnovaton equaton and productvty equaton. 5 The last two equatons were estmated n a 2SLS smultaneous equaton system, as well. However, SUR gave better results, and therefore, these are reported n ths artcle. The 2SLS results are avalable upon request. 36

7 Benek,C., Gannas,D. Moustakas,G. Innovaton And Economc Performance of Greek SMEs Innovaton Equaton 2 The determnants of nnovaton output, the x vector, consst of the number of researchers, publc sector s expendtures n R&D, and prvate sector s expendtures n R&D. Productvty Equaton Fnally, as ndependent varables (besdes the nnovaton output) of the productvty equaton we use the well known Cobb-Douglas varables,.e. Captal and Labor. For the factor Captal we use alternatvely ether Captal or Investments, wth the second gvng more credble results than the frst. For the factor Labor we use alternatvely ether total personnel or Total Personnel s Revenue wth the second gvng margnally better results. The dependent varables nclude prvate sector s expendtures n R&D, k, n equaton (2); t s defned n two ways n equaton (3), frst as the number of patents and secondly as the total R&D personnel 6 ; q n equaton (4) s defned as the workng cycle, due to the fact that t s more easly measurable and more earnestly declared than the net proft. The specfcatons of all dfferent models are gven n Table 4. Table 4. Specfcaton of eght CDM-models; where the dependent varable n the nnovaton equaton s ether the number of patents (APN) or the total R&D personnel (TRP) Models Selectvty Equaton Dep. varable Independent TP TP TP TP TP TP TP TP varables GRD GRD GRD GRD GRD GRD GRD GRD Research Investment Relaton Dep. varable BRD BRD BRD BRD BRD BRD BRD BRD RP RP RP RP RP RP RP RP Independent varables GRD GRD GRD GRD GRD GRD GRD GRD PR PR PR PR PR PR PR PR Innovaton Equaton Dep. varable APN APN APN APN TRP TRP TRP TRP BRD BRD BRD BRD BRD BRD BRD BRD Independent GRD GRD GRD GRD GRD GRD GRD GRD varables RP RP RP RP RP RP RP RP Productvty Equaton Dep. varable WC WC WC WC WC WC WC WC APN APN APN APN TRP TRP TRP TRP Independent varables K K IN IN K K IN IN TP SA TP SA TP SA TP SA 6 As an alternatve to patents, CDM-models use the nformaton provded by the CIS, n partcular, the share of sales of nnovatve products, to measure nnovaton output, whch s probably more approprate (Hall and Maresse, 2006). Unfortunately, ths s not provded by the ICAP survey. 37

8 Regonal and Sectoral Economc Studes Vol (2012) 4. Estmaton results As t was mentoned above, the CDM-model s estmated twce assumng frst, the varable output of nnovaton to be the number of patents (models 1-4), and secondly the R&D personnel (models 5-8). The estmaton process was performed by LIMDEP. The estmaton results of productvty equaton for the eght models are presented n Table 5. Table 5. Estmaton results of productvty equaton functon Model 1 Model 2 Model 3 Model 4 Coeffcent Coeffcent Coeffcent Coeffcent Constant (0.0006) (0.0000) (0.0006) (0.0000) Number of patents requested (0.0027) (0.9753) (0.8677) (0.0063) Gross Captal (0.2437) (0.0000) Investment (0.0079) (0.9862) Total Personnel (0.0000) (0.0000) Total Personnel s Revenue (0.0000) (0.9795) Durbn-Watson R Model 5 Model 6 Model 7 Model 8 Coeffcent Coeffcent Coeffcent Coeffcent Constant (0.0056) (0.0000) (0.0049) (0.0000) Total R&D Personnel (0.9412) (0.0000) (0.1245) (0.4057) Gross Captal (0.4031) (0.0000) Investment (0.0030) (0.0000) Total Personnel (0.0000) (0.0000) Total Personnel s Revenue (0.0002) (0.7118) Durbn-Watson R

9 Benek,C., Gannas,D. Moustakas,G. Innovaton And Economc Performance of Greek SMEs The estmaton of the CDM models usng the Tobt method results to a satsfactory predcton 7, 68,96%. The relevant analyss of the predctons of the bnary model are gven n Table 6. Table 6. Analyss of Bnary Choce Model Predctons Predcton Success Predcton Falure Senstvty = actual 1s correctly predcted Specfcty = actual 0s correctly predcted Postve predctve value = predcted 1s that were actual 1s Negatve predctve value = predcted 0s that were actual 0s Correct predcton = actual 1s and 0s correctly predcted % % % % % False pos. for true neg. = actual 0s predcted as 1s False neg. for true pos. = actual 1s predcted as 0s False pos. for predcted pos. = predcted 1s actual 0s False neg. for predcted neg. = predcted 0s actual 1s False predctons = actual 1s and 0s ncorrectly predcted 1.043% % % % % The two best models accordng to the R 2 and the t-statstc sgnfcant level are shown n the table below: WC Con APN K IN TP SA D-W R 2 Model Model 7 (0.0006) (0.0027) (0.2437) (0.0000) Con TRP K IN TP SA D-W R (0.0049) (0.1245) (0.0030) (0.0000) 7 That s, the estmated values wth the Tobt method, whch we used to "fx" the problem of selectvty bas n the system of nnovaton and productvty equatons, are 68,958% correct. 39

10 Regonal and Sectoral Economc Studes Vol (2012) In Table 7 below, we can see the estmated values for the nnovaton equaton for the best two models. Unlke model 1, we can see that Model 7 has a suffcency R 2 and s sgnfcant, based on the t-statstcs coeffcents In concluson, Model 7 s the overall the best one. Table 7. Estmates of the Innovaton Equaton TRP Con BRD GRD RP D-W R 2 Model (0.3588) (0.3944) (0.3816) (0.4396) Model (0.0000) (0.0000) (0.0000) (0.0000) 5. Conclusons Ths study was amed at nvestgatng the relatonshp between nnovatveness and frm economc performance n Greece. The data set used contans detaled nformaton on varous key characterstcs of Greek frms. Ths ncludes total personnel, gross captal, workng cycle, nvestments, total personnel s revenue, net proft (before tax), number of patents requested, prvate sector s expendtures for R&D, publc sector s expendtures for R&D, number of researchers, and total R&D personnel. The estmaton strategy employed n the study s based on a CDM-model framework wth regards to two dfferent varable outputs of nnovaton; the number of patents and the R&D personnel. The above analyss shows a negatve correlaton between nvestment and turnover and a postve between nvestment and nnovaton (whch s the result of research and technology). At the same tme we know that, whle nvestment grows over tme, prvate spendng on R & D does not follow the same path; ths exsts rather as a prerequste for attractng publc expendture on research and development. These let us confrm a paradox, namely, that Greek SMEs nvest (especally n new technology), but productvty s reduced. In addton to the above, t s worthmentonng that we fnd a postve correlaton between nnovaton, publc sector s expendtures for R&D and number of researchers. On the contrary, nnovaton s negatvely assocated wth prvate sector s expendtures for R&D. The above analyss dentfes a severe problem of the Greek Innovaton System, namely, the unwllngness of the prvate sector to nvest n R&D and the low productvty of nnovaton. Our emprcal fndngs underlne the mportance of leveragng prvate nvestment n nnovaton through publc nvestment. There s a need for a new plan and a new target for the Greek nnovaton system, a new nstrument and a new way to buld new R&D programs, n order to ncrease the productvty of nnovaton. That s what we can call an Effcent Innovaton, thus an nnovaton resultng n greater economc performance or greater socal productvty. 40

11 Benek,C., Gannas,D. Moustakas,G. Innovaton And Economc Performance of Greek SMEs References Benavente, J. M. (2006) The role of research and nnovaton n promotng productvty n Chle, Economcs of Innovaton & New Technology, 15(4-5): Bloch, C. (2008) The market valuaton of knowledge assets, Economcs of Innovaton and New Technology, 17(3): Bönte, W. (2003) R&D and productvty: Internal vs external R&D. Evdence from WestGerman manufacturng ndustres, Economcs of Innovaton and New Technology 12(4): Crépon, B., Duguet, E., Maresse, J., (1998) Research, Innovaton, and Productvty: An Econometrc Analyss at the Frm Level, Economcs of Innovaton and New Technology, 7(3): DTI (2006) UK productvty and compettveness ndcators 2006, Department of trade and ndustry, London, UK. Duguet, E. (2006) Innovaton heght, spllovers and TFP growth at the frm level: evdence from French manufacturng, Economcs of Innovaton and New Technology, 15: Grffth, R., Huergo, E., Maresse, J. and Peters, B. (2006) Innovaton and Productvty across Four European Countres, Oxford Revew of Economc Polcy, 22(4): Grlches Z. (1979) Issues n Assessng the contrbuton of R&D to productvty growth Bell Journal of Economcs, 10: Grlches Z. ed. (1984) R&D, Patents and Productvty, Chcago:Unversty of Chcago Press. Grlches Z. (1994) Patent Statstcs as Economc Indcators : A Survey, Amercan Economc Revew, 84:1-23 Grlches Z. (1995) R&D and Productvty: Econometrc Results and Measurement Issues, n P. Stoneman (ed.), Handbook of the Economcs of Innovaton and Techncal Change, Blackwell Handbooks n Economcs. Grlches Z. (1998) R&D and Productvty: The Econometrc Evdence, Chcago:Unversty of Chcago Press. Hall, B.H. and Maresse, J. (1995) Explorng the Relatonshp between R&D and Productvty n French Manufacturng Frms, Journal of Econometrcs 65(1): Hall, B.H. and Maresse, J. (2006) "Emprcal Studes of Innovaton n the Knowledge Drven Economy", Introducton to a specal ssue on: Emprcal studes of nnovaton n the knowledge drven economy, Economcs of Innovaton and New Technology, 15(4/5), Hall, B., Lott, F. and Maresse, J. (2009) Innovaton and productvty n SMEs: emprcal evdence for Italy, Small Busness Economcs, Sprnger, 33(1): Huergo, E. and Moreno, L. (2010) Does hstory matter for the relatonshp between R&D, Innovaton and Productvty?, EERI Research Paper Seres, EERI_RP_2010_41, Economcs and Econometrcs Research Insttute (EERI), Brussels. Janz, N., Lööf, H., Peters, B. (2004) Innovaton and productvty n German and Swedsh manufacturng frms: Is there a common story? Problems & Perspectves n Management, 2:

12 Regonal and Sectoral Economc Studes Vol (2012) Klette, T. J., and Johansen, F. (1996) Accumulaton of R&D captal and dynamc frm performance: a not-so-fxed model. Dscusson Paper no 184, Statstcs Norway. Klomp, L. (2001) Measurng Output from R&D actvtes n Innovaton Surveys, Proceedngs of sesson 53of the nternatonal statstcal nsttute (ISI) conference, Invted Paper Meetng (IPM) 63, Seoul (Korea), augustus 2001, book 2, pp Klomp, L., and van Leeuwen, G. (2001), Lnkng nnovaton and frm performance: a new approach, Internatonal Journal of the Economcs of Busness, 8(3): van Leeuwen, G., and Klomp, L. (2006) On the contrbuton of nnovaton to multfactor productvty growth, Economcs of nnovaton and New Technology, 15: Lchtenberg, F.R., Segel D., (1991) The Impact of R&D Investment on Productvty. New Evdence Usng Lnked R&D-LRD Data, Economc Inqury, 29(2): Lott, F., and Santarell, E. (2001) Lnkng Knowledge to Productvty: A Germany-Italy Comparson Usng the CIS-Database, Emprca, 28: Lööf, H., and Heshmat, A. (2002) Knowledge captal and performance heterogenety: A frm-level nnovaton study, Internatonal Journal of Producton Economcs, 76 (1), Lööf, H. and Hesmat, A. (2006) On the relatonshp between nnovaton and performance: A senstvty analyss Economcs of Innovaton and New Technology, Taylor and Francs Journals, 15(4-5): Maresse, J., Mohnen, P. (2005) The mportance of R&D for nnovaton: a reassessment usng French survey data, Journal of Technology Transfer 30(1 2): Mohnen, P., Maresse, J. and Dagenas, M. (2006) Innovatvty: A comparson across seven European countres Economcs of Innovaton and New Technology, 15(4-5): Musoles, A. and Huban, J.P. (2010) Innovaton and productvty n knowledge ntensve busness servces, Journal of Productvty Analyss 34: Pakes, A. and Grlches, Z. (1984) Patents and R&D at the frm level: A frst look n Grlches (1984). Pars, M.L., Schantarell F., and Sembenell, A. (2006) Productvty, Innovaton and R&D: Mcro Evdence for Italy, European Economc Revew, 50: Robn, S., Maresse, J. (2008) Innovaton and productvty: a frm-level analyss for French manufacturng and servces usng CIS3 and CIS4 data ( and ), Paper presented at the 25 th Celebraton Conference 2008 on Entrepreneurshp and Innovaton Organzatons, Insttutons, Systems and Regons, Copenhagen, CBS, Denmark, June 17-20, Vakhtova, G. and Pavlenko, T. (2010) "Innovaton and Productvty: a Frm Level Study of Ukranan Manufacturng Sector," Dscusson Papers 27, Kyv School of Economcs. Zacharads, M. (2004) R&D-nduced Growth n the OECD? Revew of Development Economcs, 8(3): Journal publshed by the EAAEDS: 42