Technological Opportunities, Absorptive Capacities, and Innovation

Transcription

1 Technologcal Opportuntes, Absorptve Capactes, and Innovaton Wolfgang Becker* and Jürgen Peters** Abstract The am of ths paper s to analyze the effects of technologcal opportuntes on the nnovaton actvtes of frms, dependng on ther absorptve capactes. The mportance and mpacts of the ablty of frms to use external knowledge sources were nqured especally for external knowledge stemmng from scentfc research. Usng a smple theoretc model, dfferent nnovaton effects were emprcally outlned for the German manufacturng ndustry for the frst tme. On the nnovaton nput sde, the effects of scencerelated technologcal opportuntes n combnaton wth absorptve capactes varables are stronger on the ntenstes as n the estmatons wthout such proxes. Further, the nnovaton output of frms s postvely nfluenced by the ablty to adapt external knowledge effcently. Frms n the German manufacturng ndustry wth nhouse absorptve capactes and a hgh mportance of scentfc knowledge are characterzed by hgher sales shares of new and mproved products and hgher probabltes of patent regstratons than other frms. Key words: JEL classfcaton: Technologcal Opportuntes, Absorptve Capactes, Innovaton, Scentfc Knowledge O31, I20, L20, L60 * Unversty of Augsburg, Department of Economcs, Unverstätsstr. 16, D Augsburg Phone: ; Emal: wolfgang.becker@wso.un-augsburg.de (correspendence) ** Deutsche Bahn AG, Corporate Development, ZKE 3, Holzmarkstr. 17, D Berln Phone: ; Emal: juergen.j.peters@bku.db.de

2 1 1. Introducton R&D actvtes determne the economc success and compettveness of frms. As a consequence of the dynamcs of technologcal change, nnovatve frms contnuously have to expand and optmze ther nhouse R&D by applyng technologcal opportuntes (Dos 1988; Grlches 1995; Maresse/Saasenou 1991). Ths s closely related to the ncreasng mportance of mult- and nterdscplnarty of R&D and the strengthened nterrelaton of basc research wth ndustral research (Becker 1996; Mansfeld 1995; Nelson/Wolff 1997). In general, the nnovaton capabltes of frms depend on the nteracton of nhouse R&D and the extent to whch technologcal opportuntes from outsde can be used for own purposes (Cohen 1995; Cohen/Levn 1989; Klne/Rosenberg 1986). By utlzng technologcal opportuntes frms can expand ther nnovaton capabltes wth postve effects on the research effcency, the probablty of beng successful n R&D and on the qualty of new technologes (Cohen/Levnthal 1989; Harab 1995; Klevorck et al. 1995). The mprovement of nnovaton actvtes resultng from the use of external knowledge leads to larger competences of the research personnel and to more effcent producton processes. The extent to whch frms can mplement exogenously generated knowledge, however, depends on ther absorptve capactes. Emprcal studes underlne the partcular role of absorptve capactes n the nnovaton process (Cantner/Pyka 1998; Peters/Becker 1998a; Trpsas 1997, Veugelers 1997). Level as well as qualty of absorptve capactes are dverse, varyng from one frm to another and, to a certan degree, from ndustry to ndustry. Capactes to absorb external knowledge are especally sgnfcant for hgh-tech frms n R&D ntensve ndustres, such as the computer ndustry, arcraft ndustry, astronautcs, and the pharmaceutcal ndustry. Frms n nnovatve ndustres have to nvest n complementary nhouse R&D n order to understand and use the results of externally performed R&D and to obtan full access to the research fndngs of other frms and nsttutons (Brockhoff 1995; Mowery/Rosenberg 1989). The degree to whch frms (can) use technologcal opportuntes s closely correlated wth ther capabltes pror to external knowledge used (Arvants/Hollensten 1994; Levn/Ress 1988). Further, emprcal evdence could be found for dfferences n the knd or qualty of absorptve capactes necessary to adapt technologcal opportuntes (Arora/Gambardella 1994; Malerba/Torrs 1992; Peters/Becker 1998b). Partcular nformaton generated n unverstes and other scentfc nsttutons requre hgher absorptve capactes snce the development of new and mproved products ncreasngly depends on the (generc) fndngs and results of academc research (Davd 1994; Narn/Hamlton/Olvastro 1997; Rosenberg/Nelson 1994). However, there hardly s any emprcal evdence supportng that absorptve capactes enhance frms technologcal potentals and, along wth t, ther nnovatve actvtes. Therefore, we want to nvestgate the queston, to whch extent absorptve capactes affect the nput and output level of nnovatve frms. Veugelers (1997) fnds emprcal support for the mportance of absorptve capactes to the adaptaton of external knowledge n the Flemsh manufacturng ndustry. She dentfes sgnfcant postve effects of R&D co-operatons (as a proxy for hgh technologcal opportuntes) on the level of R&D nvestments - f

3 2 frms have establshed absorptve capactes as a full-tme staffed R&D department and negatve effects otherwse. In addton, Peters/Becker (1998a) have shown for the German automoble supply ndustry that R&D arrangements wth unverstes have no sgnfcant effects on the nnovaton outputs f absorptve capactes (exstence of R&D labs) are neglected. Therefore, we want to analyze the functon of the absorptve capactes n the nnovaton process n more detal connectng theoretcal and emprcal nvestgatons. Aganst the background of the state of the art, the mportance and mpacts of the ablty of a frm to absorb external knowledge on the development of new and mproved products (technologes) are nqured. Usng nput-, output- and market-related varables, dfferences n the nnovaton actvtes dependng on the level of frms' absorptve capactes are emprcally outlned for the German manufacturng ndustry for the frst tme. The paper s organzed as follows: In the second secton, the nterrelatons of technologcal opportuntes, absorptve capactes and nnovaton actvtes are theoretcally dscussed. In the thrd secton, the data set and the defntons of the varables for the regressons are presented and the specfcaton of the emprcal models s gven. Thereafter, the results of the econometrc nvestgatons regardng nnovaton effects of technologcal opportuntes, especally from scentfc research, and the absorptve capactes of frms n the German manufacturng ndustry wll be presented n the fourth secton. The man fndngs are summarzed n the ffth secton. 2. Theoretcal Analyss The development of new and mproved products s a well-drected search and learnng process, whch nvolves techncal as well as economc uncertantes and, n case of success, sets n wth a lag (Cohen 1995; Dos 1988; Flag/Stadler 1998). Innovatons are the results of a combnaton act of frm-specfc determnants (R&D actvtes, frm sze, etc.) and external nfluences (technologcal opportuntes, R&D spllovers, etc.). Moreover, both factor groups are to be nterpreted wthn the context of ndustry-specfc condtons (sectoral technology levels, market dynamcs, etc.). R&D actvtes play a fundamental role referrng to frm nternal 'technologcal capabltes'. They am at a systematc broadenng of the exstng stock of knowledge and ts effcent applcaton n the nnovaton process. R&D actvtes are also meant to ncrease the probablty of the arrval rate of product and process nnovatons, respectvely. Furthermore R&D s performed on grounds of the assocated postve mpact of R&D-nduced mprovements and extensons n techncal and organzatonal know-how on economc magntudes such as an ncrease n productvty, turnover and profts (Cohen/Levn 1989; Grlches 1995; Maresse/Saasenou 1991). Addressng the external sphere of nnovaton-determnng factors, t s the total amount of the currently exstng and explotable resources - technologcal opportuntes - to be consdered (Arvants/Hollensten 1994; Dos 1998; Klevorck et al. 1995). Such opportuntes are dverse, varyng n the knd and usefulness of technologcal knowledge not only from ndustry to ndustry but also from one

4 3 frm to the other. "Due to varatons n the degree of avalablty of these technologcal opportuntes, nnovatons are 'cheaper' to realze... Ths factor stands - n combnaton wth others - behnd the emprcally observable nter-ndustral dfferences n the rates of techncal progress, of total factor productvty and of economc growth" (Harab 1995, p. 67) The Role of Technologcal Opportuntes n the Innovaton Process Technologcal opportuntes are related to the contrbuton of external sources to the nnovaton actvtes of frms. In our nvestgatons, we concentrate on the stock of knowledge (nformaton) frms are faced wth. These stock can be spltted nto ndustral and non-ndustral knowledge sources. Technologcal opportuntes brought up by the nnovaton actvtes of supplers, customers and compettors defne the ndustral stock of knowledge. Technologcal opportuntes arsng from nnovaton actvtes outsde the prvate sectors, e.g. the academc sphere, consttute the non-ndustral stock of knowledge. Technologcal opportuntes generated by scentfc research actvtes are of major nterest for frms n R&D-ntensve technology branches such as mcroelectroncs, chemcal ndustry, ecologcal technologes, botechnology, etc. Knowledge generated n scentfc nsttutons s an mportant nnovaton resource for frms wth hgh-leveled R&D actvtes due to the close nterrelaton of basc research and ndustral research. Scherer (1992, p. 1424) ponts out that "... the mysterous concept of technologcal opportuntes was orgnally constructed to reflect the rchness of the scentfc knowledge base tapped by frms". In the early 60 s, Nelson (1959) and Arrow (1962) emphaszed the mportance of 'new scentfc knowledge' as a drvng force behnd nnovaton, technologcal and economc progress. 1 Ever snce, ts magntude n developng new and mproved products has contnuously grown (Grupp 1994; Mansfeld 1995; Rosenberg/Nelson 1994). The ncreasng dynamc of the technologcal progress as well as the ncreasng complexty of nnovaton process account for ths. "What unversty research most often does today s to stmulate and enhance the power of R&D done n ndustry..." (Rosenberg/Nelson 1994, p. 340). The bottom lne s, "... as scentfc knowledge grows, the cost of successfully undertakng any gven, scence-based nventon declnes..." (Rosenberg 1974, p. 107). Ths leads - ceters parbus - to a rse n productvty of the frms own nnovaton actvtes. "The consequence s that the research process s more effcent. There s less tral-and-error; fewer approaches need to be evaluated and pursued to acheve a gven technologcal end. From ths perspectve, the contrbuton of scence s that t provdes a powerful heurstc gudng the search process assocated wth technologcal change" (Cohen 1995, p ). For further analyss we defne the pool of technologcal opportuntes Ω as Ω = Ω( TO_ CUCO, TO_ SUPP, TO_ SINT) (1) 1 To the role of knowledge n the nnovaton process n general, see: Beckmann et al. 1998; Navarett et al. 1998; Tamborn 1997.

5 4 where TO_CUCO represents the technologcal opportuntes stemmng from customers/compettors and TO_SUPP captures the knowledge pool generated by supplers. TO_SINT reflects the contrbutons of scentfc (unversty) knowledge to the technologcal capactes of frm. We postulate that the margnal effects of ncreasng the external research resources EX_R, wth EX_R = TO_CUCO, TO_SUPP, TO_SINT on the level of technologcal opportuntes s strctly postve ( Ω / EX _ R > 0 ) wth constant, 2 2 dmnshng or ncreasng returns ( Ω /( EX _ R ) < > 0 ), dependng on the ntal level of frms technologcal capabltes. Strength and sources of technologcal opportuntes are mportant factors explanng frm-specfc and cross-ndustry varatons n R&D ntensty and R&D productvty (Klevorck et al. 1995; Nelson/Wolff 1997; Sterlacchn 1994). The adaptaton of technologcal opportuntes changes the characterstcs and nfluences the performance of nputs requred for nnovatons. For the recpents, specfc utlzaton of exogenously generated knowledge leads to an mproved qualty of the factor nputs. Technologcal opportuntes nduce factor emboded technologcal progress whch does not affect all producton technologes to the same extent but unfold ther effects selectvely and cumulatvely wthn a technologcal problem-solvng process embedded n the scentfc, methodologcal, analytcal frame of a technologcal paradgm. "A technologcal paradgm defnes contextually the needs that are meant to be fulflled, the scentfc prncples utlzed for the task, the materal technology to be used. In other words, a technologcal paradgm can be defned as a 'pattern' for soluton of selected techno-economc problems based on hghly selected prncples derved from natural scences.... Puttng t another way, technologcal paradgms defne the technologcal opportuntes for further nnovatons and some basc procedures on how to explot them. Thus they also channel the efforts n certan drectons rather than others: a technologcal trajectory... s the actvty of technologcal progress along the economc and technologcal trade-offs defned by a paradgm..." (Dos et al. 1988, p ). Emprcal studes (Becker/Peters 2000; Gerosk 1990; Harab 1995) emphasze that technologcal opportuntes - besdes frm sze, approprablty condtons, market structures, etc. - do have a crucal nfluence on the type, range and results of frms nnovatve actvtes, e.g. the level of R&D expendture and the share of new or mproved goods n sales. Provded that the necessary, organzatonal requrements are avalable (such as qualfed R&D personnel), the adaptaton of external sources of knowledge enlarges and mproves the frms' technologcal capabltes wth postve effects on the probablty of beng successful n R&D and on the qualty of technologes. The mprovement of producton performance resultng from the use of technologcal opportuntes leads to more effcent producton processes, larger technologcal know-how and competences of the research personnel. Along ths lne, the hgher the level of technologcal opportuntes, the larger the ncentve of frms to nvest n nnovaton/r&d wll be.

6 The Role of Absorptve Capactes n the Innovaton Process Absorptve capactes refer to frms' ablty to adapt and apply external knowledge. Correspondngly, Cohen/Levnthal (1989, p. 569) defne these abltes as the capablty "... to dentfy, assmlate, and explot knowledge from the envronment". Absorptve capactes represent the analytcal lnk between the external stock of technologcal opportuntes and the nhouse capabltes n developng new and mproved products (Cantner/Pyka 1998; Cohen/Levnthal 1990; Malerba/Torrs 1992). The competence "... to evaluate and utlze outsde knowledge s largely a functon of the level of pror related knowledge. At the most elemental level, ths pror knowledge ncludes basc sklls or even a shared language but may also nclude knowledge of the most recent scentfc or technologcal developments n a gven feld" (Cohen/Levnthal 1990, p. 128). To solve technologcal tasks and problems t s drawn back on experences and knowledge from the past n terms of a 'hstory dependency of technologcal progress'. In order to mplement externally generated knowledge, especally from scentfc research, nnovatve (and R&D-ntensve) frms have to nvest a certan share of ther fnancal endowment n the mantenance and mprovement of ther absorptve capactes unless they want to lose ther compettveness. The pont s to antcpate potental and relevant technologcal tendences n order to make use of them for one s own, frm-specfc objectves. Ths s about the nvestment n one s own R&D actvtes n order to understand the results of externally performed R&D on the one hand (Cohen/Levnthal 1989; Veugelers 1997) and the nvestment n nnovaton management to acheve a well-drected and effcent explotaton of technologcal opportuntes on the other (Brockhoff 1995; Gaugltz-Lüter 1998; Rothwell/Dodgson 1991). In general, scence-based technologcal opportuntes bascally ask for a hgher level of absorptve capactes than those generated by other knowledge sources such as customers (Nelson/Wolff 1997) Frms' Inhouse Capabltes, Technologcal Opportuntes, and Innovaton Actvtes To analyze the functon of absorptve capactes n the nnovaton process as an ntermedate factor between technologcal opportuntes and the nhouse capabltes of frms n more detal, we have to dfferentate the varous effects of external knowledge sources on the nnovaton nput and output sde Innovaton Input Basc assumpton s that frm has to nvest n dosyncratc and generc R&D (nnovaton actvtes). 2 Whereas dosyncratc R&D (R d ) prmarly create specfc knowledge for frms own R&D process, 2 R&D s a part of the engagement of frms to develop new or mproved products (technologes). Innovaton actvtes nclude also expendtures for product desgn, tral producton, purchase of patents and lcense, and tranng of employees etc. In the followng, the theoretcal dscusson s concentrated on R&D as the man part of frms' nnovaton actvtes.

7 6 generc R&D ( R ge ) produce nformaton havng more the character of a publc good (Nelson 1992). In the extreme case, new generc nformaton can spll over to other actors wthout purchasng the rght to do so (R&D spllovers). 3 A further basc assumpton relates to the fact that frms dosyncratc R&D can be d dev dvded n one part, whch s necessary only to develop new technologes ( R / ) and n another part whch s only necessary to adapt, transform and use technologcal opportuntes for nhouse R&D d abs ( R / ). By ths, the whole R&D (nnovaton) nput R can be wrtten as R d ge = R R, R, Ω ), (2) ( d d/ dev d/ abs 2 d ge wth R = r( R, R ). In ths context, we suppose that R / R R > 0 and 2 R d / R Ω > 0. Further, we postulate that the generc part of nhouse R&D ( R ge ) can be substtuted by externally generated knowledge defned by Ω. Hereby, the rate of substtuton s can vary wthn the range of 0 s 1. As argued above, the extent to whch frm can mplement technologcal opportuntes s determned by ther absorptve capactes. Therefore, we take nto consderaton the parameter λ (wth 0 λ 1) reflectng frms' ablty to make use externally generated knowledge. 4 Dependng on the knd of Ω, absorptve capactes have to be mantaned, focused or enlarged (Aora/Gam-bardella 1994; Klevorck et al. 1995; Peters/Becker 1998b). The level of λ s a functon of the extent of assumng postve margnal effects d abs R /, wth /abs λ = λ( d ), (3) λ / R d abs R / lne, technologcal opportuntes frm can factual use accumulated knowledge and λ : > 0 wth constant, ncreasng or decreasng rates. In ths use Ω s a functon of the stock of externally Ω use d/ abs = λ ( R ) Ω. (4) Usng the relatons (3) and (4), the R&D (nnovaton) nput R (2) can be re-wrtten as d/ dev ge d / abs [, R, ( R Ω ] R = R R λ ). (5) The process of nternal transformaton of technologcal opportuntes can be descrbed as follows (see also the graph): (1) Frms (decde to) substtute nhouse generc R&D ( R ) by usng externally generated knowledge Ω. 5 d/ abs (2) Frms nvest a part of dosyncratc R&D to establsh absorptve capactes R ge R&D spllovers are externaltes of R&D actvtes beyond ther prmary defnton, where not alone the nnovator has the beneft, but also other actors can apply them for ther nnovatve actvtes (Dos 1988; Elasson 1996; Grlches 1992). For smplcty, we assume that frms have not to nvest n actvtes to make use of ther nhouse generc R&D. In the graph we assume full substtuton of R ge by λ Ω.

8 7 (3) Dependng on the level of absorptve capactes λ, technologcal opportuntes (external generc use R&D) can be transformed and used for own purposes Ω. (4) The R&D (nnovaton) nput R defnes frms' nternal and external resources usable for the development of new technologes. Graph: Internal transformaton process of technologcal opportuntes Knd of R&D (Innovaton) Input Idosyncratc R&D d dev R / (4) Transformaton Process Dependng on Absorptve Capactes Usable R&D (Innovaton) Input R d/ dev ge d/ abs [ R, R, λ( R ) Ω ] d abs R / (2) ge d abs Generc R&D R λ =λ( R / ) (1) (3) (4) Technologcal opportuntes Ω use d abs Ω =λ ( R / )Ω After havng hghlghted the mportance of λ, n the nternal transformaton process of external knowledge, we have to dscuss the mpacts of λ on frms' R&D (nnovaton) nput as a whole. Ω Frst, the nfluence of λ Ω on R depends on productvty effects of usng external resources. Productvty effects relate to the argument that the ncentve for frms to nvest n (dosyncratc) R&D s postve correlated wth the extent of usable technologcal opportuntes. Ths comprehenson "... corresponds to the functon that maps the flow of R&D nto ncreases n the stock of knowledge" (Klevorck et al. 1995, p. 188). At ths, the stock of knowledge expands wth dmnshng returns of nhouse R&D at the margn because technologcal opportuntes exhaust wth further progress n a gven technologcal area the more frms (markets) have to nvest n R&D (Coombs 1988). In ths lne, hgher levels of technologcal opportuntes (e.g. large flow of scentfc knowledge) enhance the productvty of nhouse R&D on product qualty or cost reducton wth stmulatng mpacts on R&D nvestments. Hereby, t s mportant to remember that the optmal level of R&D (nvestments) s determned by the nterrelaton d/ dev of R and the degree technologcal opportuntes can really used for own purposes. Second, the mpact of λ Ω on R depends on substtuton effects of usng externally generated knowledge. If - as mentoned n (2) - nhouse generc R&D ( R ge ) and λ Ω are substtutes, frms are faced wth the decson of ether generatng ths knd of knowledge by oneself or adaptng the necessary nformaton from external sources. The decson to substtute generc R&D by λ Ω s determned by the ge costs of nhouse R&D c ( R ) and by the costs of adaptaton external knowledge c( λ Ω ). If ge c λ Ω ) c ( ), there wll be no motvaton for frm to mplement technologcal nformaton generated ( R outsde. The adaptaton of usable technologcal opportuntes wll be an effcent strategy, f the costs of

9 8 ge searchng and mplementng are lower than the costs of generc R&D nhouse done: c( λ Ω ) < c ( R ). In ths stuaton, frms can use external knowledge to reduce ther costs of generc R&D. 6 But the substtuton effects on frms' total R&D costs necessary to develop new or mproved products c(r ), wth R descrbed n (5), are ambguous. To llustrate ths, let us assume thatλ Ω and R ge have the d/ dev ge same productvty regardng to R as well as c λ Ω ) < c ( ). In order of these assumptons the d/ dev man object of nterest s c( R ) ( d/ dev. Let us defne R R as the proft maxmzaton level of dosyncratc R&D wthout usng technologcal opportuntes. In ths case, we have to dffer between two stuatons: a.) As long as frms substtute the generc part of R&D only up to the cost level of generc R&D done ge formerly nhouse c(λ Ω ) R, the costs of R&D nvestments c( d dev R / ) can not be hgher than formerly wth nhouse actvtes n generc R&D. In the case of spllovers, the margnal costs of usng technologcal opportuntes are zero, because λ Ω = 0. Ths strategy of substtuton wll reduce R remarkably. 7 b.) Because of postve productvty effects of λ as descrbed above, frms whch decde to utlze Ω more generc knowledge than they had formerly generated nhouse ( λ Ω dev more n dosyncratc R&D ( R / > d d dev R / > R ge ) wll also nvest ). In ths stuaton t s not possble to make a clear statement about the total R&D costs because > ge d dev d dev c ( λ Ω) c( R ) and c( < )>c( ). For abs frms wth a hgh effcency n R / - that mplcates a low level of c λ Ω ) - t s more lkely that d dev R / d dev R / d d abs R / ( c( )<c( ). But a hgh effcency of also rses frms ncentves to absorb more external nformaton than formerly generated nhouse. By ths, the whole level of R can ncrease. In general, the substtuton effects depend on the elastcty of dosyncratc R&D λ R / d dev R / R / wth regard to Ω. For example, f the elastcty s small (hgh), the total R&D costs can be lower (larger) wth the utlzaton of external knowledge than formerly wth generc R&D actvtes done nhouse. Thus, the level of R&D expendtures wll be lower n the case of hgh levels of technologcal opportuntes than n the case of low levels. The results of the theoretcal dcusson of the productvty and substtuton effects of λ Ω on R can be summarzed - n the sense of hypotheses for the emprcal analyss - as follows: For an ncreasng effcency n the utlzaton of generc R&D t s more lkely that frms wll substtute ther nhouse producton of generc knowledge by technologcal opportuntes stemmng, e.g. from scentfc research. If technologcal opportuntes are used as substtutes (complements) for nhouse R&D, t s more lkely that the adaptaton of Ω wll dscourage (encourage) the R&D (nnovaton) actvtes of frms. But to reach a hgh level of effcency frms have to nvest more n ther absorptve capactes. By ths, frms wth hgh (low) λ have more (less) potentals to use external sources for own purposes and therefore can 6 7 As Sterlacchn (1994) notes, basc research on ther own can be more expensve and less effectve for frms than fundng academc research to realze an nnovaton (gven frms' level of absorptve capactes). For more theoretcal dscussons, see: Harhoff 1996.

10 9 mplement external generc R&D more (less) effcently. 8 If frms have hgh (low) absorptve capactes, the effects of λ Ω on the productvty of nhouse dosyncratc R&D wll also be hgh (low). Fnally, frms wth hgh (low) level of usable technologcal opportuntes Ω use d / abs = λ ( R ) Ω wll nvest more (less) n R&D (nnovaton actvtes) R Innovaton Output To dscuss the effects of λ Ω on the nnovaton (R&D) output y, ndcated by the qualty of new products or by the extent of cost reductons, we assume that y s gven as a functon of R&D nvestments, the usable pool of technologcal opportuntes, and µ, the 'propensty to nnovate' (Scherer 1965) reflectng other factors above and beyond research nputs (ncludng smple luck): d ge y = g R, R, λ, Ω, µ ) (6) ( As argued above, we assume that dosyncratc and generc R&D are necessary to realze nnovatons. Further, the avalablty of technologcal opportuntes Ω can substtute frms own generc R&D nvestment ( R ge ) dependng on the level of absorptve capactes. In general, frms' nvestment n λ does not drectly stmulate the nnovaton output such as ncreasng frms probablty to be nnovatve. Rather, n up-szng the level as well as the effcency of usng technologcal opportuntes, λ has an ndrect postve mpact on y. By ths, the relatonshp between y, frms dosyncratc and generc R&D, and λ can be expressed by the followng condtons: Ω y d / R > 0; y / R > 0 ; / λ Ω > 0, (7) ge y 2 y / ( R d ) 2 > 2 ge 2 > < 0, y / ( R ) < 0, 2 y λ 2 / ( Ω ) > < 0, 2 y / R R > 0, y / R λ Ω > 0, 2 d ge y / R R > 0. ge d 2 d Hgher nvestments n dosyncratc or generc R&D enlarge the nnovaton output - e.g. mproves the product qualty of own products or reduce the cost of producton - wth dmnshng, constant or ncreasng rates of return, dependng on the ntal level of nhouse R&D. The same condtons apply for the mpacts of usable technologcal opportuntes on y. Thus, gven the level of nhouse R&D, an expanson of λ Ω ncreases y. ge The adaptaton of technologcal opportuntes wll be an proft enhancng strategy, f c( λ Ω ) <c( R ). The nnovaton output depends on frms nhouse R&D actvtes prmarly, f c λω ) c( R ). Frms wll ( ge 8 Gambardella (1992) found emprcal support for ths assumpton n the US drug ndustry.

11 10 have no motvaton to nvest n own generc R&D, f adequate nformaton s avalable n the form of R&D spllovers. Under the postulated condtons, the followng hypotheses can be formulated: The nnovaton output y wll ncrease wth the extent of usable technologcal opportuntes. The hgher (lower) the level of absorptve capactes, the larger (smaller) the ncentves for frms to be engaged n R&D (nnovaton actvtes). By ths, an effcent adaptaton of externally generated knowledge enlarges frms' nhouse capabltes wth stmulatng effects on y. 3. Data Set, Varables and Specfcaton of the Emprcal Models 3.1. Data Set and Varables for the Regressons The data for the emprcal nvestgatons orgnate from the frst wave of the Mannhem Innovaton Panel (MIP) conducted n Germany n About 2,900 frms partcpated n the survey and flled n a questonnare about ther nnovaton actvtes for the perod of They answered a broad range of questons related to nput-, output- and market-related aspects of nnovaton actvtes. The orgnal survey covers nnovatve as well as non-nnovatve frms. Innovatve frm are defned as companes whch have ntroduced new or mproved products n the years or have ntended to do so n the perod of We restrct our analyss to nnovatve frms. 10 After havng excluded the nonnnovatve frms from the orgnal data set, 1988 nnovatve frms were ncluded n the emprcal analyss. Further, we concentrate the emprcal analyss on the manufacturng ndustry. In Germany, more than 90 per cent of the entre R&D nvested by prvate companes s performed by frms from the manufacturng sector (Federal Mnstry of Educaton, Scence, Research and Technology 1998). To estmate emprcally the mportance and effects of the absorptve capactes as an ntermedate factor between the external stock of technologcal opportuntes and the nhouse capabltes for nnovatve frms n the German manufacturng ndustry, a set of nput- and output-related varables of frms' nnovaton actvtes was used (see also Table 1). The nput-related nnovaton varables reflect the engagement of frms to generate product and process nnovatons. Ths engagement ncludes nnovaton expendtures for R&D, product desgn, tral producton, market analyss, purchase of patents and lcense. We dstngush between three proxes of frms' nnovaton nput: (1) the nnovaton ntensty (INNO_INT), measured by the log of frms' nnovaton expendtures to sales rato, (2) the R&D ntensty (R&D_INT), measured by the log of frms' R&D 9 10 We thank the ZEW for the permsson to use the censored verson of the survey data (Verson 98-1). For a detaled descrpton of the data and the data collecton, see: Felder et al. 1995; Harhoff/Lcht We have tested the model specfcatons for all frms n the ZEW data set. In the regresson, no basc dfferences related to the nfluences of the exogenous varables on the nnovaton nput and output could be found. Further, we have splt the data set n a sub-sample wth west German frms only. No fundamental dstnctons between the regressons results for the west frms and all frms were observable.

12 11 expendtures to sales rato, and (3) the R&D employment ntensty (R&D_E_INT), measured by the rato of frms' R&D employment to ther total employment. The log of these ntenstes are computed because of the problems wth non-normal dstrbutons of these varables. Unfortunately, gven a lack of data, no dstncton between dosyncratc and generc R&D nvestments could be made. Therefore, we are not able to estmate the dfferent effects of technologcal opportuntes regardng to the engagement of frms n both knds of nhouse R&D. The output-related nnovaton actvtes are measured by the sales shares of new and mproved products (INNO_NOV, INNO_IMP), based on ordered nformaton (ten-pont scale). The regstraton of patents (INNO_PA) s used to dentfy the relevance of nventon for frms. We argue that patents are a specfc output factor of R&D actvtes wthout drect market references. 11 Unfortunately, there were no nformaton avalable on the number of patents. Table 2 shows the exogenous varables used n the regressons to explan the nnovaton actvtes of frms n the German manufacturng ndustry: mportance of external knowledge sources (technologcal opportuntes), ablty to absorb externally generated nformaton (absorptve capactes), extent to whch nhouse technologcal knowledge can protect from other frms (approprablty condtons), and other market-related control varables lke frm sze, nternatonal dversty, or sales expectatons. To measure the mportance of dfferent types of technologcal opportuntes, especally from the academc sphere, the scores of factor analyss on external knowledge sources are employed (see Appendx 1). As n the theoretcal part of the paper mentoned, we dstngush technologcal opportuntes stemmng from compettors/customers (TO_CUCO), supplers (TO_SUPP), and scentfc organzatons (TO_SINT). Followng Levn/Ress (1988) we assume that the degree, frms rate scentfc nsttutons (unverstes, techncal nsttutons) as relevant sources of nformaton for nnovaton actvtes, s closely correlated wth the level of ther technologcal opportuntes (see also Arvants/Hollensten 1994; Felder et al. 1996; Harab 1995). In general, the hgher the mportance of unverstes and other scentfc nsttutons as external knowledge sources, the hgher frms' technologcal opportunty whch affect ther nhouse capabltes to develop new or mproved products postvely. Further, n the estmatons a varable reflectng separately the role of unverstes as knowledge sources (TO_UNIV) s used. Moreover, the emprcal evdence of R&D cooperatons as the most systematc form of knowledge transfer between unverstes wll be checked. COOP_UNI defnes frms havng developed new or mproved products jontly wth unverstes formally or nformally n the year The emprcal measurement of the absorptve capactes of frms s dffcult. Due to the lack of data, only two proxes for the absorptve capactes of frms could be used n the estmatons. 12 Frstly, we used the For the general dscusson of the status of patents n the nnovaton process, see: Archbug 1992; Köng/Lcht 1995; OECD We have also tested the sales shares of expendtures for tranng and professonal educaton of employees as an other proxy for absorptve capactes. As Rothwell/Dodgson (1991) show, the level of qualfcaton s strongly connected wth the ablty to establsh external lnkages to other organzatons. Especally n the case of utlzng

13 12 dummy varable R&D_REG ndcatng frms beng regularly actve n nhouse R&D. The dea behnd these modelng s that frms whch contnuously are engaged n R&D have establshed nhouse capabltes for ther dosyncratc R&D as well as for the adaptaton of external knowledge effcently. As Malerba/Torrs (1992) underlne, frms whch do not permanently nvest n R&D have a smaller access especally to technologcal opportuntes stemmng from scentfc research. Secondly, a dummy varable for the exstence (but not on the level) of nhouse absorptve capactes s used. R&D_LAB defnes frms havng one or more own staffed R&D labs n Here, the same arguments of contnuty can be made as for R&D_REG (Veugelers 1997). The degree to whch external (technologcal) nformaton s nternalzed depends on the extent frms can protect ther knowledge from other companes (Cohen/Levnthal 1989; Köng/Lcht 1995; Levn et al. 1987). Approprablty condtons defne the ablty of nnovators to retan the returns of R&D. The hgher (lower) the approprablty condtons of frms, the less (more) R&D spllovers wll occur. 13 The technologcal opportuntes a sngle frm s faced wth are not ndependent of these condtons. "For example, one frms feasble advances n technology may be blocked by the property rghts of another" (Klevorck et al. 1995, p. 186). Therefore, several varables related to the frms' approprablty condtons are used n the regressons (AP_). We employ the scores of factor analyss (see Appendx 2) on a specfc mechansm of protectng technologcal knowledge from other frms regardng to (product and process) nnovatons. Market-related varables, well-known from other emprcal studes (for an overvew, see Cohen 1995), are ntroduced n the estmatons to reflect further factors determnng the nnovaton process such as frm sze (EMPL_MI, EMPL_G), 14 sales expectatons n the medum term (SALE_EXP), degree of dversfcaton (DIVERS) and ntensty of nternatonal sales (INTERNAT). These varables allow us to control the nfluence of sze and demand factors on nnovaton actvtes, whch may explan dfferences n nnovaton actvtes on the frm level. The role of frm sze s a pror unknown, because these varables "... can be used as a proxy for varous economc effects" (Arvants/Hollensten 1996, p. 18). It can be expected that sales expectatons (Klenknecht/Verspagen 1990; Schmookler 1966), hgh export shares of sales (Felder et al. 1996; Lunn/Martn 1986) and a hgh degree of dversfcaton (Kamen/Schwartz 1982; Nelson 1959) wll nfluence postvely the nnovaton actvtes of frms. scentfc knowledge, a hgh level of qualfcaton of the employees seems to be necessary. In our data set, only a small fracton of frms have gven nformaton about ther expendtures to qualfy ther employees. Due to the reducton of the sample sze n a to large extent, we have to exclude ths proxy from the emprcal analyss In the case of ntra-ndustry spllovers, the ncentves of frms to nvest n R&D wll dmnsh (Harhoff 1996; Levn/Ress 1988; Spence 1984). In the case of spllovers stemmng from sources outsde the market (e.g. from unverstes), postve effects of R&D-spllovers on the effcency and output of nnovaton actvtes can be expected (Bernsten 1989; Nadr 1993; Peters 1998). Frm sze s a categoral varable wth three extensons. We defne the category 'small frms' (up to 50 employees) as basc group. EMPL_MI (50 up to 249 employess) and EMPL_G (250 and more employees) are used as dummes n the estmatons.

14 13 Restrctons of the data set make t mpossble to nclude varables on the degree of competton n the frms' market (10-frm concentraton rato or Herfndahl ndex) n the model specfcatons. But other studes (Arvants/Hollensten 1996; Crépon/Duget/Kabla 1996) have shown that the degree of competton has an mpact of no sgnfcance or of a comparable small order of magntude on the nnovaton or R&D actvtes of frms, f the estmatons are controlled by varables of technologcal opportuntes. In the ZEW data set, the manufacturng ndustry encloses eleven sectors. Accordng to the classfcaton of the OECD (1992), these sectors are dvded n three technology groups (LOW, MEDIUM, HIGH) dependng on the sales shares of R&D. The varable MEDIUM s used as basc group n the regressons. The results, reported here are the same usng dummes of dfferent ndustres Specfcaton of the Emprcal Models The basc model specfcaton for explanng the nnovaton actvtes x of frms n the German manufacturng ndustry s as follows: x = α + α TO_ CUCO + α TO_ SUPP + α TO_ SINT + α AP + α MR + ε (8) where TO_CUCO, TO_SUPP and TO_SINT represent proxes of technologcal opportuntes stemmng from customers/compettors, supplers and scentfc nsttutons, AP stands for the approprablty condtons of frms, and MR represents market-related determnants such as frms sze, sales expectatons, etc. ε s an unobserved, addtve error term. To assess the nnovaton effects of frms absorptve capactes, we modfy the basc equaton (8) to x = α + α TO_ CUCO + α TO_ SUPP + α TO_ SINT + α AC + α AP + α MR + ε (9) where AC measures the exstence of absorptve capactes (λ ) alone. At least, we try to estmate the nteracton of absorptve capactes and technologcal opportuntes (λ Ω ) n form of α AC = α ( λ, TO_ SINT ). (10) 5 5 wth concentraton on TO_SINT as a proxy for scence-based Ω. The estmaton of our model specfcatons rases several statstcal problems. On the one sde, the nnovaton and R&D ntenstes were censored n the upper tal of the dstrbutons by 0.35 resp 0.15 to avod the dentfcaton of frms. On the other sde, some nnovatve frms dd not performed any R&D as well as had no nnovaton expendtures. Acceptng a msspecfcaton of the model, the problem can be solved by usng a Tobt model wth censorng n both tals of the dstrbutons. Possble msspecfcaton may be attrbuted to the fact that ndependent parameters smultaneously determne the probablty as well as the expendtures of nnovaton actvtes (Cohen/Levn/Mowery 1987). Therefore, we use the two-step verson of the Heckman model (Heckman 1979). Ths model specfcaton allows us to dentfy the

15 14 parameters affectng a frm's decson to partcpate n R&D (nnovaton actvtes) and the degree of ts ntensty. In the case of dscrete varables, we employ the probt methods for dchotomous response varables and ordered probt models for the varous multnomal varables. 4. Emprcal Analyss In the followng, the results of the emprcal analyss testng the hypotheses formulated n the theortcal part of the paper are presented. We nvestgate the effects of technologcal opportuntes on the nnovaton actvtes of frms n the Germany manufacturng ndustry, dependng on ther absorptve capactes. The mportance and mpacts of the ablty of frms to use external knowledge sources were nqured especally for external knowledge stemmng from scentfc research. Our regresson strategy s as follows: In a frst step, we measure the mpacts of technologcal opportuntes among other factors on the nnovaton nput and output actvtes of German frms wthout varables reflectng frms absorptve capactes. By ths, we nvestgate whether the complementarty effects of usng external knowledge domnate the substtuton effects. In a second step, the nfluence of nhouse (absorptve) capactes on the nvestments of frms and on the results of ther nnovaton/r&d actvtes are analyzed. Herewth, we want to show f the nfluence of technologcal opportuntes reman for frms wth hgh nhouse capactes. By ths, emprcal evdence of absorptve capactes as an ntermedate factor between external knowledge sources and the engagement (performance) of frms to develop new or mproved products can be specfed wll be gven. Due to the close connecton between scentfc knowledge and the development of new or mproved products, the estmatons are focussed on the effects of technologcal opportuntes stemmng from scentfc research. We analyze these effects under three aspects: Frstly, we test the mpacts of the mportance of scentfc nsttutons (unverstes, techncal nsttutes, ndustry-fnanced research nsttutes, etc.) on the level of the factor scores on external sources of nformaton (TO_SINT). Secondly, we check the contrbuton of unverstes separately as knowledge sources (TO_UNIV). Thrdly, we ncure the dummy varable COOP_UNI to dentfy frms havng developed new products or processes jontly wth unverstes Innovaton Effects of Technologcal Opportuntes wthout Absorptve Capactes Innovaton Input Effects The mpacts of the exogenous varables are estmated wth regard to the queston, f and how they can explan the probablty that frms are engaged n the development of new or mproved products and the ntensty of ther nnovaton/r&d actvtes. The regresson results for INNO_INT, R&D_INT, and R&D_E_INT are summarzed n Table 3.

16 15 Usng the two-step verson of the Heckman model, no sgnfcant effects of the scence-related varables TO_SINT, TO_UNIV and COOP_UNI on the probablty beng engaged n the nnovaton process could be found. But TO_SINT and COOP_UNI have very hgh sgnfcant effects on the ntenstes of nnovaton actvtes. Further, the regressons ndcate hghly sgnfcant effects (at the level) of the scence-related varables on the partcpaton n R&D. Hgh assessments to scentfc/unversty knowledge and an establshment of R&D cooperatons wth unverstes ncrease the probablty of beng actve n R&D. Wth excepton of COOP_UNI, the estmatons underlne the mportance of stmulatng effects of TO_SINT and TO_UNIV on the ntensty of R&D. The strongest effects could be found regardng the estmatons of the R&D employment ntensty. Technologcal opportuntes stemmng from scentfc research ncreases the probablty as well as the ntensty of R&D nvestments n human captal (at the level). The regressons underlne that scentfc knowledge sources are used as complements n the German manufacturng ndustry. The adaptaton of these knds of technologcal opportuntes encourage and stmulate the nput actvtes of nnovatve frms because of an ncreasng productvty of nhouse research or of hgher R&D nvestments n ther absorptve capactes. Inhouse capabltes can be expanded wth postve effects on the engagement and ntensty of developng new or mproved products. In ths context, Nelson/Wolff (1997) gves emprcal support on the level of certan lnes of US busness that the outcome of scence can be regarded as pure opportunty enhancng. On the other hand, t has to be mentoned that "... by far the largest share of the work nvolved n creatng and brngng to practce new ndustral technology s carred out n ndustry, not n unverstes" (Rosenberg/Nelson 1994, p. 340). However, as Harab (1995) and Klevorck et al. (1995) remark, the mpacts of scentfc research on R&D and nnovaton actvtes can dffer across ndustres. In some technology felds, the results of academc research are used as substtutes for ndustral research. Peters/Becker (1998a), for example, fnd substtutve effects of scentfc knowledge on nhouse actvtes n the German automoble supply ndustry. Some knds of nnovaton actvtes, such as testng and prototype buldng are outsourced by German automoble supplers to unversty and scentfc laboratores, whch yelds remarkable savngs n nnovaton costs (see also Peters/Becker 1998b). In ths case, the cost savngs are larger than the stmulatng (complementng) mpacts of scentfc research on nhouse R&D. In terms of technologcal opportuntes, the regressons ndcate no sgnfcant effects of TO_SUPP as the stock of external knowledge generated by supplers on the nnovaton/r&d actvtes. The negatve sgn of the coeffcents for the nnovaton ntenstes can partly be explaned by the fact that nformaton from supplers tend to be substtutes for nhouse actvtes, as also found n studes for the US (Cohen/Levnthal 1989; Levn/Ress 1988; Nelson/Wolff 1997). Aganst ths, technologcal opportuntes stemmng from customers/compettors unfold stmulatng and sgnfcant effects on the ntenstes of nnovaton/r&d. The coeffcents of TO_CUCO related to the probablty and the ntensty of R&D nvestments n human captal are postve but, n general, wthout sgnfcance.

17 16 We dentfy stmulatng effects of the approprablty condtons on the nvestments of German frms n the development of new or mproved products. Frm-specfc strateges to protect knowledge from other companes (AP_FIRM) ncrease the probablty of an engagement n nnovaton/r&d and the level of expendtures sgnfcantly. Mechansms of protectng nnovatons by law (AP_LAW) have strongly sgnfcant effects on the partcpaton n R&D, but not on the ntensty of R&D nvestments. Decson and ntensty of frms to nvest n R&D employment are postvely nfluenced on the sgnfcant level by AP_FIRM as well as by AP_LAW. Hghly sgnfcant explanatory power could also be found of frm sze classfcatons (EMPL_MI, EMPL_G) on the probablty for engagng n nnovaton/r&d and for nvestng n R&D employment. Large and mddle-szed frms n the German manufacturng ndustry have a hgher probablty of nvestng n nnovaton nput actvtes. The effects of the ncurred frm sze varables on the ntensty of nnovaton/r&d expendtures are negatve. These fndngs are n lne wth studes n other countres (Cohen/Klepper 1996; Evangelsta et al. 1997; Klenknecht 1996). In general, large frms have a hgher probablty to engage n the nnovaton process than small frms but - f they have nvested - they spend less money n ther nnovaton actvtes. Other marked-related control varables have the expected sgns. Hgh shares of exports (SALE_EXP) stmulate the nvestment n nnovaton/r&d (sgnfcant at the level) whch strengthens the hypothess of Schmookler (1966). The ntensty n nternatonal sales (INTERNAT) and the degree of dversfcaton (DIVERS) have postve and hghly sgnfcant effects only on the decson of German frms to be actve n R&D. Foregn trade encourages sgnfcantly (at the level) the probablty as well as the level of nvestments n R&D employment. Fnally, the regressons ndcate the expected (sgnfcant) effects of the technology level of ndustry groups (LOW, HIGH). The hgher the technology level of ndustres, the more ntensve the nvestments n nnovaton/r&d and R&D employment are Innovaton Output Effects To estmate the mpacts of technologcal opportuntes, especally of scentfc research on the nnovaton output n the German manufacturng ndustry, the set of explanatory varables as on the nput level s used wth slght modfcatons. The effects are analyzed for the sales shares of new products (INNO_NOV), the sales shares of mproved products (INNO_IMP), and the regstraton of patents (INNO_PA). As shown n Table 4, technologcal opportuntes stemmng from scentfc nsttutons (TO_SINT) have postve effects on the three nnovaton output varables, but wthout statstcal sgnfcance. Only TO_UNIV reflectng nformaton separately from unverstes has postve effects at the sgnfcance level on the probablty of patent regstraton n the German manufacturng ndustry. Ths reveals a pattern confrmed by the fndngs of Grupp (1996). In general, these outcomes verfy the presumpton that technologcal opportuntes stemmng from scentfc research mprove the qualty of products more ndrectly by ncreasng frms' R&D effcency rather than by generatng techncal advance drectly.

18 17 Smlar to the nnovaton nput estmatons, the regressons pont out postve, stmulatng (sgnfcantly) effects of COOP_UNI. R&D cooperatons wth unverstes as the most systematc form of knowledge transfer ncrease the probablty of hgher sales shares of mproved products (at the level) and of the regstraton of patents (at the level). The mpacts on the probablty of hgher sales shares of new products are very weak and nsgnfcant. Ths result s contrary to the fndngs of other studes (Faulkner/Senker 1994; Mansfeld 1995). They found strong postve and sgnfcant effects of scentfc knowledge sources on the sales shares of new products. More research have to been done to analyze the data set n more detal, especally to fnd out sectoral peculartes, to reveal the reasons for the specfc constellatons n Germany. Comparng the regresson results wth secton 4.1.1, the explanaton power of TO_SINT, TO_UNIV and COOP_UNI are much weaker than n the estmatons for the nnovaton nput varables. Especally the applcaton of the Ordered Probt models for INNO_NOV yeld nsuffcent results, whch can be seen by the level of the McKelvey/Zavona R 2. Lookng at the other knds of technologcal opportuntes, the regressons uncover the followng remarkable connectons: TO_SUPP has postve and hghly sgnfcant mpacts on INNO_NOV. The hgher frms rank the mportance of external knowledge from supplers, the hgher the sales shares of new products are. The estmatons ndcate stmulatng and strong effects (sgnfcant at ) of technologcal opportuntes stemmng from customers and compettors (TO_CUCO) only on INNO_IMP. Further, the effects of the approprablty condtons on the nnovaton output of frms n the German manufacturng ndustry are postve and mostly sgnfcant. The econometrc nvestgatons pont out the evdence of the market-related factors. By ths, the effects of the frm sze varables (EMPL_MI, EMPL_G) are ambguous. On the one hand, larger frms have lower sales shares of new products (negatve coeffcent wth lack of sgnfcance). On the other hand, postve effects on the sales shares of mproved products could be dentfed for medum-szed frms (at the sgnfcance level). At last, large frms have a hgher probablty of patentng ther new fndngs. These fndngs strengthen the presumpton that t s more lkely that larger frms nvest n process nnovatons than for smaller frms, whch perform product R&D rather than process R&D (Peters 1998). Further, hgh levels of exports (SALE_EXP) have sgnfcantly stmulatng mpacts on both sales shares. The ntensty n nternatonal trade (INTERNAT) unfolds emprcal evdence only on the probablty of patent regstraton (at the level), the degree of dversfcaton (DIVERS) only on the sales shares of new products (wth lack of sgnfcance). Fnally, the regressons reflect stmulatng and sgnfcant effects of the ndustry group HIGH on INNO_NOV. Frms n ndustres wth hgh technology levels undertake R&D to develop novel products rather than frms n low technologcal.