The impact of government support to industrial R&D on the Israeli economy. Final report. English translation from Hebrew

Transcription

1 The mpac of governmen suppor o ndusral R&D on he Israel economy Fnal repor Englsh ranslaon from Hebrew Prof. Saul Lach, Dr. Shlom Parza, Danel Wasserel June 2008 E.G.P Appled Economcs Ld. Research and Consulng n Economcs, Markeng and Socal Scences. 48 Menachem Begn Rd., Tel-Avv Phone: , Fax:

2 Inroducon Inroducon Ths repor summarzes a research conduced by he undersgned for he Mnsry of Fnance and he Offce of he Chef Scens ("OCS") n he Mnsry of Indusry, Trade and Labor. The obecve of he research was o esmae he mpac of governmen suppor o ndusral R&D on he Israel economy. By combnng he resuls of several economerc models esmaed usng daa from he manufacurng ndusry and he R&D surveys from 1996 o 2003, we are able o presen esmaes, of a knd never prevously presened n Israel, regardng he expeced oal ncremen o ndusral GDP resulng drecly from he governmen's suppor o R&D. We would lke o express our graude o he charman of he professonal seerng commee, Dr. Yehoshua (Shuk) Gleman whose vas experence and useful remarks were of grea benef o he qualy of he research. We hank he members of he seerng commee, Mr. Yar Ama (MATIMOP), Mrs. Smcha Bar-Elezer (Cenral Bureau of Sascs), Mr. Yaackov Barkay (Mnsry of Fnance), Mr. Ruby Gunel (Manufacurers Assocaon of Israel), Mr. Mchel Hver (MATIMOP), Mr. Gad Levn (coordnaor of he commee - MATIMOP), Mrs. Lda Lazanes (Offce of he Chef Scens a he Mnsry of Indusry, Trade and Labor), Mrs. Haya Mller (MATIMOP), Dr. Mchel Srawchnsky (Bank of Israel), and Mr. Haga Ido (Mnsry of Fnance) who conrbued her nsghs and experse hroughou he enre work process. A specal graude s reserved for Mrs. Smcha Ben-Elezer and he saff of he Cenral Bureau of Sascs for processng he daa, negrang he fles of ndusry surveys and R&D surveys, and provdng he calculaons and analyses ha form he bass of hs research. Researchers: Dr. Shlom Parza, Appled Economcs Ld. Danel Wasserel, Cenral Bureau of Sascs Academc Advsor: Prof. Saul Lach, The Deparmen of Economcs a he Hebrew Unversy of Jerusalem 2

3 Table of Conens & Lss of Fgures and Tables Table of Conens Inroducon Summary and conclusons Obecves and research mehodology General Esmang he reurns o he economy from R&D acvy Technologcal and marke proxmy beween frms Buldng poenal spllover pools Economerc mehodology Esmang he mpac of governmen suppor on he level of execued R&D (addonaly/crowdng ou) Descrpon of he daa Daabases Descrpve sascs Frm heerogeney Indusry concenraon of producon and R&D Fnancng sources for R&D acvy Inal ndcaors of R&D oupu n he manufacurng ndusry Research resuls The mpac of governmen suppor on he levels of execued R&D Esmang he reurns o he economy from R&D acves Prvae reurns on R&D nvesmens Esmaon of R&D spllover effecs

4 Table of Conens & Lss of Fgures and Tables 4.3. An negraed model for esmang he reurns o he economy from governmen R&D suppor Reurns on R&D nvesmens vs-à-vs nvesmens on physcal capal Appendx 1: Ls of varables n he daa bases Appendx 2: Prce adusmens Appendx 3: Buldng capal seres Appendx 4: Issue of double counng R&D acvy Appendx 5: Mergng he daa from he manufacurng ndusry surveys and he R&D n manufacurng surveys Appendx 6: Grlches model Appendx 7: Blundell and Bond model Appendx 8: Esmaon oupu - Grlches producon funcon Appendx 9: Esmaon oupu - Blundell and Bond 1998 producon funcon Appendx 10: Esmaon oupu - addonaly n he manufacurng ndusry Appendx 11: Esmaon oupu - addonaly n he compuer servces, sofware and R&D branches Appendx 12: Ls of ndusral branches by echnology nensy Appendx 13: Surveys of manufacurng and surveys of R&D n manufacurng and n he compuer servces, sofware and R&D branches - dscusson

5 Table of Conens & Lss of Fgures and Tables Ls of Fgures Fgure 1: Dsrbuon of R&D expendures as a percenage of revenues n hghechnology esablshmens, Fgure 2: Selec dsrbuonal characerscs of R&D employees n manufacurng esablshmens Fgure 3: GDP dsrbuon by echnologcal nensy - manufacurng ndusry surveys, Fgure 4: R&D expendures dsrbuon by echnologcal nensy - manufacurng ndusry surveys, Fgure 5: CR4 ndex of ndusry concenraon n R&D expendures whn echnologcal nensy groups - surveys of R&D n he manufacurng ndusry, 1996 Vs Fgure 6: CR4 ndex of ndusry concenraon n R&D expendures - surveys of R&D n he compuer servces, sofware and R&D branches, 1997 Vs Fgure 7: Number of observaons by echnologcal nensy and exernal R&D fnancng - R&D n manufacurng surveys, Fgure 8: Number of observaons by man branch and R&D fnancng from exernal sources - R&D surveys n he compuer servces, sofware and R&D branches, Fgure 9: Frms' GDP per workng hour by echnologcal nensy - percenage dfference beween frms ha performed R&D and frms ha ddn', surveys of he manufacurng ndusry and R&D n he manufacurng ndusry, on average Fgure 10: R&D oupu Toal Facor Producvy (TFP), on average Fgure 11: Correlaon coeffcens beween R&D expendures and TFP, on average Fgure 12: Toal R&D ncremen n he economy as a resul of governmen grans (addonaly) manufacurng ndusry, 2003 NIS n mllons

6 Table of Conens & Lss of Fgures and Tables Fgure 13: Toal R&D ncremen n he economy as a resul of governmen grans (addonaly) compuer servces, sofware and R&D branches, 2003 NIS n mllons Ls of Tables Table 1: Sze of samples manufacurng and R&D n manufacurng surveys Table 2: Sze of samples R&D survey n he compuer servces, sofware and R&D branches Table 3: Observaons by echnologcal nensy - manufacurng Table 4: Dsrbuon of he number of years an esablshmen akes par n he samples R&D n manufacurng surveys and manufacurng surveys Table 5: Dsrbuon of he number of years an esablshmen akes par n he sample R&D survey n he compuer servces, sofware and R&D branches Table 6: Man characerscs of he R&D expendures dsrbuon - manufacurng ndusry, on average Table 7: Man characerscs of he R&D expendures dsrbuon - compuer servces, sofware and R&D branches, Table 8: Indusry concenraon of GDP manufacurng ndusry surveys, Table 9: Indusry concenraon of revenues R&D n he manufacurng ndusry surveys, Table 10: Indusry concenraon of revenues - R&D surveys n he compuer servces, sofware and R&D branches, Table 11: The dynamcs of R&D suppor by he Offce of he Chef Scens o he manufacurng ndusry, Table 12: The dynamcs of R&D suppor by he Offce of he Chef Scens o he compuer servces, sofware and R&D branches,

7 Table of Conens & Lss of Fgures and Tables Table 13: Manufacurng esablshmens and compuer servces, sofware and R&D frms ha were graned governmen fundng, by source Table 14: Dsrbuon of R&D exernal fundng (governmen and oher sources) manufacurng ndusry branches Table 15: Dsrbuon of R&D exernal fundng (governmen and oher source) compuer servces, sofware and R&D branches Table 16: Addonaly esmaes manufacurng ndusry branches, 2003 NIS n mllons Table 17: Addonaly esmaes - compuer servces, sofware and R&D branches, 2003 NIS n mllons Table 18: Number of Esablshmens by revenue groups, echnology nensy and R&D execuon Table 19: Margnal GDP ncremen as a resul of an addonal NIS 100 housand of R&D nvesmens - Grlches model Table 20: Margnal GDP ncremen as a resul of an addonal NIS 100 housand of R&D nvesmens - Blundell and Bond model Table 21: Esmaes of R&D spllovers from medum-large frms (NIS mllon) - Grlches model, gven a NIS 100 housand ncremen n he branch R&D level Table 22: Esmaes of R&D spllovers from medum-large frms (NIS mllon) - Blundell and Bond model, gven a NIS 100 housand ncremen n branch R&D level Table 23: Esmaes of R&D spllovers from very large frms (NIS 300+ mllon) - Grlches model, gven a NIS 100 housand ncremen n he branch R&D level Table 24: Esmaes of R&D spllovers from very large frms (NIS 300+ mllon) - Blundell and Bond model, gven a NIS 100 housand ncremen n branch R&D level Table 25: Toal reurns o he economy due o R&D governmen suppor Grlches model, medum-large frms (NIS mllon), assumng a NIS 5 mllon governmen gran

8 Table of Conens & Lss of Fgures and Tables Table 26: Toal reurns o he economy due o R&D governmen suppor Blundell and Bond model, medum-large frms (NIS mllon), assumng a NIS 5 mllon governmen gran Table 27: Toal reurns o he economy due o R&D governmen suppor Grlches model, very large frms (NIS 300+ mllon), assumng a NIS 5 mllon governmen gran Table 28: Toal reurns o he economy due o R&D governmen suppor Blundell and Bond model, very large frms (NIS 300+ mllon), assumng a NIS 5 mllon governmen gran Table 29: Raos of he reurn on R&D nvesmens o he reurn on physcal capal nvesmens

9 1. Summary and Conclusons 1. Summary and conclusons For many years, he Israel governmen has allocaed vas resources o he suppor and promoon of cvl research and developmen (R&D) acves by he prvae secor. Ths suppor has oaled approxmaely 3 bllon NIS over he years The obecve of hs research was o quanfy he mpac of governmen suppor o ndusral R&D on he Israel economy, and n parcular, o provde polcy makers wh a quanave esmae of he ndusral GDP ncremen expeced from each addonal Shekel of governmen suppor dreced o ndusral R&D n Israel. Publc suppor of R&D carred ou by he busness secor s usfed n he presence of marke falures ha whou governmen nervenon resuls n an equlbrum where prvae nvesmen n R&D s sub-opmal from a socal vewpon (Arrow 1962) 2. The man falure ensues from he fac ha nvesng n R&D yelds new knowledge whose assmlaon or adopon by frms ha dd no bear he developmen coss canno be prevened fully. Ths phenomenon s known as R&D spllovers. The prvae reurn o R&D, whch forms he bass for decson-makng n frms, does no nclude he spllover componen and s herefore smaller han he oal reurn o he economy. Furher marke falures occur n he capal markes when enrepreneurs and frms fal o ransfer busness rsks n full o rsk-neural players such as banks and nsurance frms due o asymmerc nformaon beween he dfferen players. The obecve of governmen suppor o R&D s o ensure ha knowledge s "produced" a an opmal level from a socal vewpon. Due o he exsence of knowledge spllovers and varous consrans over nvesmen decsons n frms, he governmen ams o ncrease R&D expendures of frms o a level hgher han he level obaned due o 1 Source: Offce of he Chef Scens n he Mnsry of Indusry, Trade and Labor. The daa represen ne suppor afer deducon of royaly paymens receved over he years by he governmen. 2 Arrow, K. J., 1962, "Economc Welfare and he Allocaon of Resources for Invenon", n The Rae and Drecon of Invenve Acvy, Prnceon Unversy Press: Prnceon, pp

10 1. Summary and Conclusons prvae consderaons of prof and loss only. Ths may be acheved by removng consrans o nvesmen or by provdng drec suppor for he execuon of R&D acves, hereby enablng frms o oban a reurn ha s equal o he socal reurn and choose her nvesmens accordngly. In lgh of he aforemenoned, hs repor was requred o presen quanave answers o he followng quesons: 1. Wha s he new R&D level ha drecly resuls from governmen nervenon and ha oherwse would no have been underaken? 2. Wha s he GDP ncremen o he economy relaed o hs new R&D? The frs queson, known n he economc leraure as he ssue of crowdng ou/addonaly, concerns he proporon beween frms' prvae resources and publc fundng. One may consder a suaon n whch a frm s graned governmen fnancng for a research proec ha would have been carred ou by ha frm even whou he same governmen suppor. Ths suaon s referred o n he leraure as crowdng ou - a suaon n whch governmen suppor "crowds ou" prvae funds ha would have been used o fnance he proec had he governmen suppor no been graned. Naurally, n he even of crowdng ou, publc funds do no creae new R&D, bu smply replace prvae fnancng. On he oher hand, s possble ha governmen funds consue an ncenve for addonal nvesng greaer han ha whch would have been carred ou under prvae marke erms. Ths suaon, known as addonaly, refers o a suaon n whch publc funds smulae frms o nves n R&D up o a level ha s hgher han he one ha would be acheved solely as a resul of frms' prvae consderaons. The second queson s conneced o he ncremen n GDP ha sems from a successful R&D process. Ths ncremen s comprsed of he drec effec o he GDP of he frm ha performed he R&D and he ndrec (spllover) effec over oher frms n he ndusry. Therefore, we mus esmae: (1) he mpac of R&D nvesmen on he execung frm's 10

11 1. Summary and Conclusons performance, and (2) he R&D spllovers mpac -.e. he mpac of one frm's R&D on he performance of oher frms n he ndusry. The resuls of hs repor are based on a unque daabase compled by combnng he surveys of he manufacurng ndusry and he surveys of R&D n he manufacurng ndusry for he years The daabase comprses dealed nformaon abou some 2,800 ndusry esablshmens (plans) ncludng, ner ala, daa on added value, sales, nvesmens, employees and R&D expendures. Concurrenly, an analyss was conduced usng daa from he R&D surveys of he compuer servces, sofware and R&D branches, whch nclude approxmaely 470 esablshmens durng he years A hs pon, we would lke o noe ha he daabase suffers from a sgnfcan scarcy of observaons of frms ha engaged n R&D n he radonal manufacurng branches. Ths fac makes dffcul o draw sascally robus conclusons regardng hese branches. The man resuls from he research are as follows: A. The R&D ncremen derved from governmen suppor Governmen suppor nduces he "creaon" of new R&D, whch, were no for he suppor, would no have been underaken, of up o 2 o 3 mes he amoun of he gran gven (dependng on he level of gran). Ths resul refers o he margnal governmen suppor and s sascally sgnfcan and robus hroughou dfferen economc branches boh n he manufacurng ndusry and n he compuer servces, sofware and R&D branches, and wh regard o frms whose echnologcal level and sze dffer from one anoher. I follows from hs ha he suppor mechansms mplemened durng he years covered by our analyss dd no crowd ou prvae funds (namely, dd no fnance nvesmens whch would have been aken up anyhow) bu raher resul n sascally sgnfcan R&D nvesmen addonaly regardless of he branch or sze of he suppored frms. 11

12 Addonal R&D - Mllons of NIS 1. Summary and Conclusons The esmaes show ha n he manufacurng ndusry branches, a governmen gran amounng o NIS 1 mllon creaes an ncremen n prvae R&D expendures amounng o NIS 1.28 mllon meanng a oal ncremen of NIS 2.28 mllon of R&D expendures n he economy. In he compuer servces, sofware and R&D branches, he ncremen n prvae R&D expendures s NIS 1.81 mllon, and he oal R&D ncremen o he economy s NIS 2.81 mllon. These are lower bound esmaes for addonaly de faco, snce hey were esmaed based on gross grans daa whou deducng royaly paymens o he governmen. The followng fgure presens he cenral resuls derved from analyzng he addonaly ha sems from governmen R&D suppor. The columns (Y axs) represen he oal R&D ncremen o he economy as a resul of governmen grans (X axs). The dagonal lne s a 45 lne. Smlar fndngs were aaned for R&D governmen suppor n he compuer servces, sofware and R&D branches. Toal R&D ncremen n he economy as a resul of governmen grans (addonaly) manufacurng ndusry 2003 NIS, mllon Gran Sze - Mllons of NIS 12

13 1. Summary and Conclusons B. Toal reurns o he economy The reurns o he economy derved from R&D governmen suppor are very hgh even n branches where mos R&D nvesmens are made (hgh-echnology/medum-hgh echnology) and n whch daa observaons abound. The reurns o he economy derved from governmen suppor were esmaed usng wo separae producon funcon models (Grlches,Blundell & Bond), well known n he economc leraure. The resuls obaned n regard o boh he prvae reurns o R&D and he spllovers esmaes are conssen wh he resuls from research on he subec elsewhere. The resuls of he analyss show ha mos of he R&D spllovers are derved from medum-large frms (urnover of NIS mllon) and very large frms (urnover greaer han NIS 300 mllon). The resuls aaned reflec a mnmal mulpler of 5 o 6 mes beween he governmen nvesmen and he oal fuure ndusry GDP ncremen (for a gran of NIS 5 mllon) o frms wh a sales urnover of NIS mllon; and a mnmal mulpler of 1.5 o 2 beween he governmen nvesmen and he oal fuure ndusry GDP ncremen (for a gran of NIS 5 mllon) o frms wh a sales urnover of NIS 300+ mllon. Hgher reurns are aaned n he medum-low, medum-hgh and low-echnology branches. The man fndng, noneheless, s ha even whn he hgh-echnology branches where he vas maory of R&D expendures n manufacurng are concenraed, and whch have receved hgh levels of governmen suppor hroughou he years a mulpler of 4.7 o governmen funds s aaned. Tha s o say, ha even n branches wh hgh levels of R&D and governmen suppor and a large quany of daa observaons avalable - posve and hgh reurns of governmen suppor o prvae R&D are aaned. 13

14 1. Summary and Conclusons C. Addonal fndngs 1. Ths research presens for he frs me daa regardng he dynamcs of he suppor of he Offce of he Chef Scens o frms (a he esablshmen or plan level) n he manufacurng ndusry, compuer servces, sofware and R&D branches. The facs ndcae ha he suppor of he Offce of he Chef Scens s graned for wo consecuve years o approxmaely 70% of he frms and for hree consecuve years o approxmaely 50% of he frms. Tha s o say, a frm whose reques was approved for he frs me n a ceran year has an average expeced probably of 70% o receve connued suppor n he followng year and a 50% probably o receve suppor for hree consecuve years. From he vewpon of he frms, he daa reflec a consderable degree of uncerany regardng he probably of change n he level of suppor graned. Among he frms whch are graned suppor for wo consecuve years, 50% wll receve a suppor ha s abou 50% hgher han he suppor receved n he prevous year, and he oher half wll receve suppor ha s abou 40% lower han ha receved n he prevous year. Noneheless, an analyss smlar o hs one, of he daa fle from he Offce of Chef Scens a he sngle proec level, raher han a he frm (esablshmen) level, as was conduced n hs research, s called for. 2. A large degree of heerogeney was found beween he frms n regard o economc varables relevan o polcy-makng. For nsance, average R&D expendures as a share of frms' GDP n manufacurng amouned o 8% n he years The sandard devaon, ha measures he spread around he average, equaled approxmaely 50%! The dsrbuon medan was only 2% - abou one fourh of he average. Furhermore, he dsrbuons whn groups of echnologcal nensy and sze do no provde a more homogenous pcure. 3. A sysemac examnaon of he daa reveals a hgh and perssen concenraon of frms' GDP and R&D expendures beween and whn he relevan branches. In parcular, examnng he share of large frms n he oal branch GDP/revenues reveals ha abou wo hrds of manufacurng producon orgnaes n 14

15 1. Summary and Conclusons approxmaely 10% of he frms only. A pcure us as exreme emerges from he R&D surveys n he compuer servces, sofware and R&D branches. 4. R&D expendures are also concenraed manly n a lmed number of large frms. In he hgh-echnology branches, where mos of he manufacurng ndusry R&D expendures are concenraed (86%), abou one hrd (!) of R&D expendures are made by he four larges frms. In oher words, four hgh-echnology frms alone are responsble for approxmaely 30% of he overall R&D expendures n he enre Israel manufacurng ndusry. 5. The margnal reurn o R&D nvesmens s smlar o ha receved on capal nvesmens n he hgh-echnology branches alone. In all oher branches, he reurn on R&D nvesmens s 6 o 200 mes hgher han he reurn o nvesmen n physcal capal. Due o scarcy of observaons n some of he branches, we esmae ha he reasonable reurn raos are up o 8 mes n favor of ndusral R&D nvesmens. The facs show ha from he vewpon of resource allocaon n he economy R&D nvesmens are, n mos cases, o be preferred over capal nvesmens. Based on he analyss and fndngs above, we presen he followng man conclusons: 1. The hgh addonaly found demonsraes ha he governmen suppor mechansm s effecve and succeeds n selecng proecs, whch would no have been aken up were no for governmen suppor. In lgh of he aforemenoned, s reasonable o assume ha a smlar performance s o be expeced wh regard o curren suppor mechansms n branches for whch exensve pas daa s unavalable (boechnology, nanoechnology). 2. The hgh reurn o he economy semmng from R&D governmen suppor usfes he shfng of governmen budge allocaons o hs acvy. Ths concluson ceranly apples o he relavely hgh levels of suppor provded n he years , and even more so o he suppor levels currenly provded. These allocaon 15

16 1. Summary and Conclusons shfs should be raded off vs-à-vs governmen suppor schemes ha bear lower reurns. 3. The offcal se of publcaons whch gudes polcy makers should explcly show and refer o he hgh level of frm heerogeney beween and whn dfferen seemngly homogenous branches - n regard o relevan varables such as GDP, R&D expendures, workers, ec. We beleve ha average changes n he varables are no enough o enable he formulaon of effcen polcy ools. 4. Enablng he Offce of he Chef Scens o comm n advance o mul-year fnancng should be consdered. We beleve ha f frms would be aware n advance of he fnancng me horzon, hey could be spared he uncerany assocaed wh he approval of requess on a yearly bass. I seems ha n he case of hgh-rsk processes such as R&D, where par of he raonale for governmen suppor s he reducon n busness relaed rsk, s mporan no o creae a new uncerany assocaed wh requred regulaory processes for he approval of suppor. Ths ssue can be examned opmally usng he Offce of he Chef Scens daa a he proec level. 5. R&D spllovers s bu one of he nsances n whch governmen nervenon s usfed n he conex of R&D. Thus, for example, he survval and growh of small frms depends, ner ala, on access o capal markes, he dsance from produc markes, and effcen rsk managemen, whch, n mos cases, necessae governmen nervenon. 6. Lasly, we found benefcal o nclude, as an appendx o hs documen, a se of recommendaons and dscusson pons wh regard o daa colleced n he framework of he manufacurng ndusry surveys and he R&D n he busness secor surveys. By dong so, we hope o provde polcy makers wh clear, focused answers o some of he quesons, brough o our aenon whls preparng hs research, whch were lef unanswered due o nsuffcen daa. 16

17 1. Summary and Conclusons The srucure of he repor s as follows: he second chaper presens he man research queson of hs work and he chosen mehodology o address. The hrd chaper descrbes he daabases used whn he framework of hs research and her sources. Subsequenly, we presen comprehensve, descrpve sascs regardng frms n he manufacurng ndusry, compuer servces, sofware and R&D branches n Israel. The fourh chaper presens he resuls of he research. I presens economerc esmaes enablng a mehodcal examnaon of he amouns of new R&D creaed n he economy as a resul of R&D governmen suppor. These resuls are negraed no a combned model ha enables he esmaon of he oal reurns o he economy derved from governmen suppor o ndusral R&D n he years

18 2. Obecves and Research Mehodology 2. Obecves and research mehodology In hs chaper we presen he queson of hs research and he mehodology used o answer opmally: As specfed n he ender, he obecve of he research s o esmae he mpac of governmen suppor o ndusral R&D on he Israel economy. The concepual framework for he analyss of hs mpac s presened below: 2.1. General The erm "Research and Developmen" s defned and wdely acceped as a mehodcal process for he creaon of scenfc or novel echnologcal 3. The economc oupu of he R&D process les n he applcaon of hs knowledge o mprove producs and/or exsng producon processes and/or o develop such processes from scrach. The ranson from a phase of generang knowledge va research o an applcaon whn he framework of a frm conrbues o producvy growh and provdes he frm wh compeve advanages n he markeplace. I s commonly hough, as shall be presened laer, ha R&D processes are characerzed by a socal reurn ha s hgher han he prvae reurn - a fac ha accordng o economc heory usfes he subsdzng of R&D acvy. Relyng on hs prncple, he Israel governmen has been supporng and encouragng frms o perform processes of knowledge accumulaon as menoned above (suppor of R&D). The obecve of he research presened here s o esmae quanavely he reurns expeced n erms of GDP for each Shekel of ndusral R&D governmen suppor - a he level of he frm. The queson of he research may be broken down no wo 3 Research and Developmen a mehodcal, orgnal acvy amed a creang scenfc knowledge or new echnologcal knowledge or alernavely o develop a novel applcaon of exsng scenfc or echnologcal knowledge. See he nroducon chaper o he R&D n manufacurng survey, he Cenral Bureau of Sascs. 18

19 2. Obecves and Research Mehodology secondary quesons. The frs queson refers o he exen of he mpac ha he level of execued R&D has over ndusral performance n Israel, and he GDP of he manufacurng ndusry n parcular. The second queson refers o he mpac ha governmen R&D suppor has over he level of ndusral R&D execued by frms. The oulne of he research can herefore be formulzed by he followng smple dfferenal equaon: dy ds Y RD RD s (1) s s Where Y s he GDP a he frm level, RD s he R&D sock of ha same frm and s s he R&D governmen gran provded. Thus, he lef-hand sde of he equaon shows he change n GDP as a resul of an sδ ncremen of governmen R&D suppor. Ths change s comprsed of he expeced change n GDP as a resul of he ncrease n R&D (frs rgh-hand sde expresson n he equaon) mulpled by he expeced change n R&D as a resul of he change n he level of governmen suppor (second rgh-hand sde expresson n he equaon). The research mehodology presened below oulnes he emprcal pah requred o quanfy each of he expressons menoned. The oal effec of a successful R&D process s comprsed boh of a drec effec on he frm effecuang and an ndrec effec on oher frms n he ndusry (spllovers). In order o quanfy he frs rgh hand expresson above, one mus quanfy: (1) he mpac of he R&D nvesmen on he performance of he execung frm and (2) he mpac of he R&D spllovers. Tha s o say, he R&D mpac of ha same frm on he performance of oher frms n he ndusry. In secon 2.2 below, we presen wo wdely acceped models used o esmae producon funcons ha nclude R&D componens and R&D spllovers. 19

20 2. Obecves and Research Mehodology In order o quanfy he second rgh-hand expresson above, we mus examne wha would he R&D expendures of each frm have been, f governmen suppor had no been graned. Clearly, no daabase enables o drecly answer hs queson. There s no "parallel unverse" from whch we can learn wha would have been he R&D nvesmen consderaons of a frm had no receved governmen suppor. In secon 2.3 below, we shall presen a combnaon of wo esmang mehods acceped n he conemporary economc leraure ha allow for he execuon of he requred comparson. Lasly, s possble o on he wo derved esmaes above no one combned model enablng a forecas of he GDP change n he ndusry branches ha sems from governmen suppor o R&D n hese branches over he years. 2.2 Esmang he reurns o he economy from R&D acvy R&D acves performed by frms generae novel echnologcal knowledge; generally, a frm does no succeed n reanng exclusvely hs knowledge. The erm "echnologcal spllovers" refers o he par of generaed knowledge whch spreads among frms, branches, secors and saes wh no remuneraon or compensaon o he eny whch creaed ha knowledge. Therefore, he reurns o he economy from frms' R&D acves comprse boh he prvae reurns o he frm execung he R&D plus he reurns o oher frms as a resul of echnologcal spllovers. The man and wdely acceped argumen for governmen nervenon amed o promoe R&D s he exsence of a marke falure ha leads o sub-opmal prvae nvesmens n R&D from a socal vewpon (Arrow 1962). 4 Ths marke falure occurs bascally because 4 Arrow, K. J., 1962, "Economc Welfare and he Allocaon of Resources for Invenon ", n The Rae and Drecon of Invenve Acvy, Prnceon Unversy Press: Prnceon, pp

21 2. Obecves and Research Mehodology he spllovers of new knowledge and s adopon by frms ha do no bear he developmen coss canno be fully prevened. Ths phenomenon s known as a posve exernaly. In such a suaon, he oal socal benef from knowledge s greaer han he drec benef expeced o be ganed by he developng frm by vrue of s nvesmen n R&D. In oher words, he eny ha makes he nvesmen decson (he frm) wll be nclned o nves less han he amoun ha would have been decded on by all he benefcares of he fru of ha same nvesmen (boh he nvesng/developng frm and oher frms). In addon, R&D nvesmens are characerzed by a hgh level of uncerany and herefore, n hs conex, one can also expec cred markes o provde less-hanneeded cred due o nheren neffcences n rsk allocaon and asymmerc nformaon beween enrepreneurs/developers and nvesors (see Romer 1190, Grlches 1998) 5. I s commonly held ha he phenomenon of cred shorage has a hgher mpac on small frms whose access o capal markes s lmed han on large, well-esablshed frms. Grlches 1979 defned wo knds of echnologcal spllovers: Pure knowledge spllovers and ren spllovers. A knowledge spllover s defned as he spread of knowledge semmng from R&D acves among varous frms n an unconrolled, unguded manner ha leads o or suppors echnologcal mprovemens. Such spllovers can ake place n dverse ways such as: adopon of new echnologes leadng o her learnng and assmlaon by he adopers n a process called reverse engneerng, cooperaon agreemens, ranson of employees beween frms, ec. Ren spllovers refer o suaons where he prces of echnologcally mproved producs drop relavely rapdly as a resul of he mprovemens. In suaons as such, echnologcal spllovers are manfesed n he ransfer of value from he frm whch mproved he produc o a frm whch purchases he produc as an npu for producon. 5 Grlches, Z., 1998, R&D and Producvy, Unversy of Chcago Press: Chcago. Romer, P., 1990, Endogenous Technologcal Change", Journal of Polcal Economy,

22 2. Obecves and Research Mehodology The spllovers phenomenon and s mporance o he echnologcal and economc progress of economes have been well known o and acceped by economss for several decades. Formulang governmen polcy for he suppor of R&D consdered o be he man source of spllover - requres, herefore, knowledge abou he scale and nensy of hs phenomenon de faco. Noneheless, here are serous consrans wh regard o he denfcaon and quanfcaon of he connecons and he scope of hs phenomenon. We presen bellow a mehodologcal framework for esmang he exen of knowledge spllovers whn and beween he manufacurng ndusry branches n Israel. Our framework s based on curren leadng emprcal mehodology from he economc leraure Technologcal and marke proxmy beween frms The acves of frms ake place smulaneously n wo man spheres: The echnologcal sphere where he producon process s performed and he sphere of he arge markes for he producs ("he markeng sphere"). Accordngly, dfferen frms can be more or less close o one anoher n each of he abovemenoned felds. Thus, for example, n he 80's, Kodak and Polarod compeed over her arge markes (roll flms) ye he echnologes a he bass of her producs were que dfferen. The companes were close n erms of her arge markes and dsan n erms of her producon processes. An addonal example s sofware frms, whch use he same knd of producon echnology - use of he same programmng language o sell dfferen applcaons o dfferen markes (a prvae marke or an nsuonal marke, ec.). 6 See: (1) Jaffe, 1986, Technologcal Opporuny Spllovers of R&D: Evdence from Frms' Paens, profs, and Marke Value", Amercan Economc Revew, 76(5), pp (2) Bloom, Shankerman and Reenan, Idenfyng Technologcal Spllovers and Produc Marke Rvalry: Theory and Evdence from a Panel of U.S. Frms", mmeo July,

23 2. Obecves and Research Mehodology The performance level of he frms wll be affeced by successful R&D acves of oher frms accordng o her proxmy n he wo dscussed spheres, and n oppose manners: A successful execuon of R&D acves by frm A wll have a posve mpac on frms ha are close o n erms of producon echnologes, hrough knowledge spllovers, whch wll enable hem o mprove her producon processes oo. Bu he same successful execuon of R&D acves by frm A wll have a negave mpac on frms ha compee wh n he end-produc markes (bu are echnologcally dsan) snce hey wll suffer from relave nferory (or wll lose a echnologcal advanage hey possessed). The posve mpac can be hough of as a complemenary effec of R&D and he negave mpac can be referred o as he rvalry effec of R&D. The posve mpac of R&D beween frms grows as her producon echnologes are closer and he dsance beween her arge markes s greaer. A grea proxmy boh n producon echnology and n arge markes can generae a oal R&D mpac ha s posve, nsgnfcan, or even negave. Ths conrass wh he common noon regardng he posve connecon beween R&D and he performance of frms n he same ndusry - a fac ha calls for reconsderaon he desrable scope of governmen nerference and s mpac on he economy's branches. Naurally, s lkely ha a successful R&D acvy underaken by a frm wll probably have no mpac on frms wh dsan producon echnologes and arge markes. Bloom, Shankerman and Van Reenen (2005) examned a sample of Amercan frms beween he years They were he frs o presen a model combnng he rvalry effec and he spllover effecs of R&D. The research showed ha boh R&D effecs are quanavely sgnfcan, and ha her overall mpac on he economy was posve. Tha s o say, he posve mpac of R&D on producon echnology was found o be greaer han he negave R&D mpac va he arge markes of he producs. I s mporan o noe, early on, ha we do no have any daa enablng us o separae he wo ypes of effecs wh regard o he Israel manufacurng ndusry. Thus, we only esmae he oal effec of R&D spllovers and her mpac hrough he arge markes. 23

24 2. Obecves and Research Mehodology Buldng poenal spllover pools The mpac of R&D s esmaed by buldng "poenal spllover pools" for each frm n he sample a any gven me. These pools reflec he oal sources from whch knowledge may spll over o any frm. For any gven frm, spllover pools are bul on he bass of he R&D expendures/r&d sock of oher frms weghed by he level of proxmy beween he specfc frm and each of he oher frms. Weghng by he level of echnologcal proxmy mplemens he noon ha knowledge complemenares exs beween frms ha are close o each oher on he sphere of producon echnology. Opposely, weghng by he level of proxmy n arge markes mplemens he noon of knowledge rvalres whch exss beween rval frms. A echnologcally-weghed spllover pool beween close frms shall be defned for each frm a a gven me as he sum of he R&D expendures/r&d sock of all he oher frms weghng her echnologcal proxmy o frm : )2) TSP TechSplloverPool TechClosseness n, R & D Where R & D are he R&D expendures/r&d sock of any frm a a gven me and n s he oal number of frms. The spllover mpac derved from hs pool for each frm s expeced o be posve Economerc mehodology We esmae he level of R&D spllovers usng wo models wdely acceped n he economc research leraure for esmang producon funcons. Ther basc assumpon s ha one can wre he producon of a frm as an explc funcon of s level of capal 24

25 2. Obecves and Research Mehodology nvesmens, maerals purchases and workers. A he same me, he oal producvy componen of he frm can be expressed as a funcon of s own R&D nvesmens (prvae reurns on R&D) and he poenal spllovers sock (echnologcal complemenares vs. rvalres n arge markes). In he frs model presened heren and developed by Prof. Zv Grlches, an R&D capal sock s bul for each company smlarly o he physcal capal sock. 7 The underlyng assumpon s ha reurns on R&D nvesmens las over several years (lke equpmen and physcal machnery). Therefore, n regard o R&D acvy, he R&D capal sock s he nfluencng facor over a frm's fnancal resuls. Ths model s among he mos wdespread, acceped models for esmang producon funcons wh R&D nvesmens. Please refer o appendx 3 for a dealed descrpon of he manner n whch he varables of capal and R&D sock are bul n hs conex. Noneheless, many researchers have crczed he Grlches model due o he specfc srucure mposes on he evoluon of frms' producvy over me (represened by he developmen of he R&D capal sock). In parcular, we mus assume a deprecaon rae for R&D nvesmens n order o buld he R&D capal sock, and we mus assume ha he capal sock a any gven me s a lnear funcon of pas R&D expenses/nvesmens. In hs respec, oher models were developed for esmang producon funcons whch do no necessae he assumpons above. The second model esmaed whn he framework of hs research was developed by Blundell and Bond (1998). In hs model, producvy and s evoluon over me are esmaed from usng avalable daa (see appendx 6) 8. 7 Grlches, Z., 1979, Issues n Assessng he Conrbuon of Research and Developmen o Producvy Growh", Bell Journal of Economcs, 10, pp Blundell, R. and Bond, S., 2000, "GMM Esmaon wh Perssen Panel Daa: An Applcaon o Producon Funcons", Economerc Revews, 19(3),

26 2. Obecves and Research Mehodology The esmaed equaons for each frm a any gven me may be wren n he followng general form: Y ( R & D, TSP )3) Own R&D Ohers' R&D, K, L, Conrols, TechConrols) Where Y s producon or added value (GDP), R&D are R&D expendures (Blundell and Bond model) or he R&D capal sock (Grlches model). The varable TSP s he relevan spllover pool for each frm a a gven me. K s he physcal capal sock of a frm and L sands for work npus. The conrol varables are used o elmnae fxed me effecs from he explaned varable. Typcal conrol varables are: annual dummy varables, frms' ndusry codes, ec. Conrol varables of specal mporance n he conex of esmang echnologcal spllovers are he echnology groups o whch frms belong TechConrols n he equaon above. These varables are se o capure he mpac of "echnologcal opporunes" whch can be used by frms ha belong o smlar echnology groups. These mpacs are no spllover mpacs ye may be nerpreed as such whenever hey are no explcly conrolled for. 2.3 Esmang he mpac of governmen suppor on he level of execued R&D (addonaly/crowdng ou) The obecve of governmen nervenon n he area of R&D s o brng abou he execuon of new R&D whch would no have been execued oherwse. Ths, wh he am ha knowledge s "produced" a an opmal level from a socal vewpon. The governmen neres s o ncrease R&D expendures n frms o a level ha s hgher han ha would be effecuaed based solely on prvae consderaons of prof and loss. 26

27 2. Obecves and Research Mehodology Ths s acheved by provdng frms wh ncenves so ha he prvae level of reurn hey see, ncludng he governmen subsdy, s equal o he socal reurn and he frms choose her nvesmens accordngly. The man queson asked n hs conex s ha of he crowdng ou/addonaly assocaed wh he rao beween publc funds and a frm's prvae resources. Le us consder a suaon n whch a frm s graned governmen fnancng for a research proec whch would have aken up even whou governmen suppor. Such a suaon s referred o n he leraure as crowdng ou a suaon n whch governmen suppor "crowds ou" prvae funds ha would have been used o fnance he proec had no been for he governmen suppor. Naurally, n a crowdng ou suaon, publc funds do no creae new R&D bu smply replace prvae fnancng. From he economy's pon of vew hs suaon s undesrable. On he oher hand, s possble ha governmen funds serve as an ncenve for addonal nvesng beyond he level ha would have prevaled under prvae marke erms only. Ths suaon s named addonaly and refers o a suaon n whch publc funds smulae frms o nves n R&D up o a level ha s hgher han ha whch hey would have nvesed n, based on her prvae consderaons alone. The esmaon of R&D expendures spen by a subsdzed frm on R&D had no been for governmen subsdes s he foundaon sone on whch he esmaon of he mpac of governmen suppor o R&D s based. Naurally, we do no drecly observe hese expendures, and here les he man dffculy of analyss. Furhermore, drecly comparng he R&D expendures of frms whch were graned fnancng o hose of frms whch were no graned fnancng would be concepually flawed n hs conex. Ths s due o he selecon process whch characerzes governmen suppor. The sample of frms ha were graned governmen fnancng for her R&D acves s no a random sample of he populaon of frms. Any comparave process mus bear hs n mnd. Oherwse, he resuls aaned wll be based. Nowadays, forunaely, he economc leraure offers mehods ha enable he esmaon of he mpac of governmen suppor 27

28 2. Obecves and Research Mehodology schemes hrough a srucured comparson of subsdzed frms o "couner facual" unsubsdzed frms whls conrollng for her observed characerscs. In order o esmae he mpac of governmen suppor o ndusral R&D, he nonparamerc "propensy scores" mehod was used. 9 The propensy scores mehod s a sascal ool hrough whch we can defne he smlary beween frms - subsdzed and unsubsdzed n erms of her nclnaon or propensy o receve an R&D gran. By dong so, we can creae a sor of conrol group agans whch we can compare he mpac of a subsdy on he subsdzed frms (effec of reamen on he reaed). The obecve of he analyss s o creae a suaon n whch we are able o denfy frms whose characerscs are such ha hey have a smlar or equal probably of recevng governmen suppor ex-ane. Among hem, we assume ha he granng of a subsdy s random and s no coordnaed wh he frms' characerscs as specfed above. Therefore, we are able o compare frms ha were graned governmen R&D suppor n he general populaon wh hose of he conrol group ha were no graned R&D governmen suppor. Esmang he propensy scores s frsly acheved by esmang he probably o receve a gran among all he frms. Ths s done usng prob ype regressons, whch enable he esmaon of he probably of recevng governmen fnancng accordng o he known characerscs of a sample of frms - ncludng hose whch evenually dd no receve fnancng (amoun of employees, mpor volume, sales, ec.): )4( Pr obably of Subsdy F( x ˆ ) 9 See: Czarnzk, D., and Fer A., 2002, Do Innovaon Subsdes Crowd ou Prvae Invesmen? Evdence from he German Servce Secor, ZEW Dscusson Paper No , Manhem. 28

29 2. Obecves and Research Mehodology Where s a vecor of relevan predefned parameers, whose value s unknown and X s a vecor of observable varables among he frms of he sample. Followng he esmaon of he regresson, one can calculae for each frm he specfc probably of recevng a subsdy. These values are called propensy scores. Where ˆ s he esmae for he vecor of parameers n he populaon and x are he varable values of a specfc frm. Followng he calculaon of he propensy scores for each frm, one can creae pars of frms where: (a) one of hem receved fnancng and he oher dd no; (b) boh have an almos dencal propensy (score) of receve fnancng ex-ane; (c) boh are smlar n her sze and he echnologcal feld characerscs. Thus, n fac we creae a conrol group whch comprses a "parallel unverse" n whch we observe unsubsdzed frms ha are smlar as much as possble o he frms ha were graned governmen suppor - on a frm on frm bass. The nex phase nvolves a comparson beween he R&D expendures of subsdzed frms o hose of unsubsdzed frms n he conrol group. We conduc he comparson by esmang a lnear regresson wh he followng general srucure: (5) Y, Y ( ), X, X ( ), s,, observables besdes subsdes observables besdes subsdes subsdy effec me specfc effec random effec Where: = frm ha receved R&D suppor (subsdy) = a frm ha belongs o he conrol group, s a mach of frm bu ha dd no receve R&D suppor. Y, = R&D expendures of a frm ha receved a subsdy (ne, followng deducon of governmen suppor). Y(), = R&D expendures of frm ha belongs o he conrol group, s a mach of frm bu ha dd no receve a subsdy. 29

30 2. Obecves and Research Mehodology X, = a vecor of varables mpacng frms' R&D expendures. = a vecor of he coeffcens for all observed varables mpacng he explaned varable besdes he subsdy varable. s, = he level of subsdy receved. = he mpac of he governmen subsdy on R&D expendures he man coeffcen for he purpose of analyss. = mpacs derved from unobserved varables muual o all frms (boom or bus n he economy, rse of general npu prces such as peroleum, ec.), = random shocks. Under he funconal assumpons n he specfcaon above, one can demonsrae ha he coeffcen s equal o he mpac of a subsdy on he subsdzed frms, whch we could have esmaed, f we had daa regardng wha her R&D expendures would have been whou he subsdy.e. couner facual daa, see Lach (2002) for a more dealed dscusson of he esmaon mehod 10. Ths coeffcen, our focus of neres, measures o wha exen are he R&D expendures of subsdzed frms bgger or smaller han hose n oher non-subsdzed frms. The addonaly/crowdng ou esmaes and he esmaes of he reurns o R&D n producon erms can be combned no a sngle model ha enables he esmaon of he oal reurns o governmen R&D suppor. If we assume a ceran level of governmen R&D suppor, he esmaes obaned regardng he addonaly/crowdng ou of R&D suppor enable us o esmae he volume of addonal nvesmen n R&D - beyond he nvesmen ha would have been execued whou he subsdy - ha s expeced o be underaken n he economy as a resul of hs R&D subsdy. Then, hs new R&D 10 Lach, S., 2002, "Do R&D Subsdes Smulae or Dsplace Prvae R&D? Evdence from Israel", The Journal of Indusral Economcs, December, Vol. L, No. 4, pp

31 2. Obecves and Research Mehodology nvesmen can be ranslaed no producon (added value or GDP) fgures usng he esmaes of he reurns o R&D ha may be obaned as descrbed above. 31

32 3. Descrpon of he Daa 3. Descrpon of he daa Ths chaper descrbes he daa used n he framework of he research, s sources and he nal processng underaken n order o presen descrpve sascs wh regard o manufacurng frms and frms n he compuer servces, sofware and R&D branches n Israel. The mporance of he daa presened here les n he fac ha hey show, for he frs me n Israel, addonal characerscs of he dsrbuon of economc varables n hese branches (varance, medan and percenles), beyond he average fgures publshed n he offcal publcaons of he Cenral Bureau of Sascs. Thus, we presen a broader pcure of he economc phenomenon n focus - R&D, enablng hus polcy makers o learn abou he hgh level of frm heerogeney characersc of he researched branches. Upon decdng o conduc hs research and s mehodology, was agreed upon wh he Cenral Bureau of Sascs ha a daabase wll be specally prepared for he purpose of hs analyss. The daabase was prepared and processed by employees of he Cenral Bureau of Sascs based on he surveys of he manufacurng ndusry and he surveys of R&D n manufacurng and n he compuer servces, sofware and R&D branches. The ables, regressons, and resuls appearng n hs work are he bass for he analyss and have been prepared by he Cenral Bureau of Sascs guded by Dr. Shlom Parza and dreced by Prof. Saul Lach Daabases A concse revew of he srucure of he surveys and he varables s presened below. A full specfcaon of he ls of varables n each survey s presened n appendx 1. The manufacurng ndusry survey comprses dealed frm-level daa. The acual un from whch daa are colleced n he framework of he surveys descrbed here s he 32

33 3. Descrpon of he Daa esablshmen, whch may alernavely be par of a bgger frm or be a frm by self. The erms frm and esablshmen are hus used alernavely hroughou hs repor referrng eher o frms or esablshmens. The daa nclude repors of sales volume, R&D expendures (n a paral manner), npu purchases, capal nvesmens, producon (oupu and added value or GDP) employees, labor hours, expors and addonal mporan economc varables. The survey s conduced once a year among a represenave sample of manufacurng frms 11. One can herefore "append" he daa for each frm over he years and creae a frm-level panel daa se 12. The daa from he manufacurng ndusry surveys are from years 1995 o The survey of R&D n he manufacurng ndusry s conduced accordng o a smlar mehodology o ha of he manufacurng ndusry surveys, bu examnes only manufacurng frms ha underook research and developmen acves durng he surveyed year. The daa base conans dealed economc varables regardng he frms' of research and developmen number of employees engaged n R&D acves and her earnngs, and educaon; frms' R&D expendures, R&D fnancng from sources ousde he frms (ncludng from he Offce of Chef Scens, oher governmenal enes, nernaonal funds, and so forh), acquson of paens or her developmen as well as oher economc varables relaed o he frm's acves (oal revenues, expors, ec). The manufacurng ndusry R&D surveys conan frm-level daa from a represenave sample of manufacurng frms ha were engaged n R&D beween years 1996 o An overlappng exss beween he esablshmens parcpang n he surveys of manufacurng and he surveys of R&D n manufacurng (n parcular regardng he large ones). Therefore, he daa from boh surveys may be merged n order o creae a unfed fle of analyss as was done here (see appendx 5). The R&D survey of he compuer servces, sofware and R&D branches s conduced dencally o he manufacurng ndusry R&D survey regardng frms from 2-dg code 11 Frms ha belong o he 2-dg ndusry branches 13 o 39 accordng o he Israel Cenral Bureau of Sascs' "Unfed classfcaon of economy branches, 1993". 12 Panel daa s he erm gven o a se of daa whch enables researchers o monor he developmen of economc acvy of a sngle frm n he course of me. 33

34 3. Descrpon of he Daa branches 72 - compuer servces and sofware, and 73 research and developmen frms 13. The followng ables presen daa regardng he quany of observaons n each survey. Table 1: Sze of samples manufacurng and R&D n manufacurng surveys Year Manufacurng surveys Number of esablshmens R&D n manufacurng surveys Boh surveys merged , , , , , , , , , Toal 17,102 1, The manufacurng ndusry surveys nclude an average of approxmaely 1,800 observaons (esablshmens) a year spannng 22 sub-branches (a an aggregaon level of 2 dgs). Therefore, he amoun of observaons n he manufacurng ndusry surveys enables he analyss of he daa whn 2-dg ndusry branches. The R&D surveys of he manufacurng ndusry branches nclude only abou 200 frms a year (from he same 22, 2-dg ndusry branches) and he merged se of daa (esablshmens parcpang n he same year n boh he manufacurng ndusry survey and he manufacurng R&D survey) ncludes an average amoun of observaons of approxmaely 120 frms per year. A smlar average amoun of yearly observaons exs n he R&D survey of he compuer servces, sofware and R&D branches - able bellow. Furher on, we shall 13 The compuer servces and sofware branch (72) ncludes acves such as daa processng, manenance of nformaon banks, sofware or programmng servces, consulaon regardng compuerzaon, ec.. The research and developmen branch (73) ncludes frms engaged n basc research n naural scences, lfe scences, ec. (as dsngushed from research and developmen acves performed n oher economc branches). Sar-up frms are ncluded n hs branch. Daa regardng frms ha ook par n he OCS Technologcal Incubaor program a he me of conducng he survey were excluded from he sample. 34