Technological Opportunities, Academic Research, and Innovation Activities in the German Automobile Supply Industry

Transcription

1 Technologcal Opportuntes, Academc Research, and Innovaton Actvtes n the German Automoble Supply Industry Jürgen Peters and Wolfgang Becker* February 1998 Abstract In ths paper the mportance and the effects of technologcal opportuntes, stemmng from academc research, on the nnovaton actvtes of frms n the German automoble supply ndustry are nvestgated. We observe that the contrbuton of academc research to frms nnovaton actvtes s less mportant than the relevance of ndustral sources but yet, the most lkely partners for formal R&D co-operatons are unverstes. Usng measures of supplers nnovaton nput and output, we can outlne dfferences n the effects of academc research on supplers nnovaton behavor. Although the proxmty to academc research stmulates supplers actvtes n own R&D, the unversty knowledge substtutes supplers nvestment n R&D and other nnovaton actvtes. Unversty co-operatons seem to have a postve mpact on the mprovement of exstng rather than on the development of new products or processes. But we can show that the nfluence of academc research on supplers nhouse nnovaton actvtes depend on ther absorptve capactes. Key words: JEL classfcaton: Innovaton Actvtes, Technologcal Opportuntes, Academc Research, Absorptve Capactes, Automoble Supply Industry O31, I20, H40 Dr. Jürgen Peters, PD Dr. Wolfgang Becker: Department of Economcs, Unversty of Augsburg, Unverstätsstr. 16, D Augsburg, Germany, Phone: , Fax: , Emal: juergen.peters@wso.un-augsburg.de. Fnancal support by the research fund of the Unversty of Augsburg (F-423) s gratefully acknowledged. We thank Jürgen Detz for helpful support.

2 1 I. Introducton In Germany, the development of automobles belongs to those economc actvtes - next to the electrcal and electronc ndustry as well as the chemcal ndustry - wth the hghest R&D levels (Federal Mnstry of Educaton, Scence, Research, and Technology 1996). In 1995, more than one ffth of the total R&D expendtures n the prvate sector was used for the research and development of new automobles or parts (lnes of economcs 244), such as motors, tres, electronc components, plastcs, etc. A share of 5.9 per cent of R&D expendtures on sales n frms engaged n R&D s much hgher n the automoble ndustry than n the German manufacturng ndustry as a whole wth 3.4 per cent. The same s vald for the share of R&D employment to total employment (6.0 per cent to 3.8 per cent). 1 In recent years, the mportance of supplers for the nnovaton process of automakers has steadly ncreased (VDA 1996, Wldemann 1993). German automakers (e.g. Aud, BMW, Damler Benz, VW) have changed ther busness strateges for developng new automobles. Most of the R&D, nnovaton and producton actvtes formerly done n-house are now outsourced to suppler frms. Nowadays, automakers purchase more parts from outsde than they produce themselves (Messner et al. 1994). As a consequence of outsourcng, suppler frms have more responsblty for the development and producton of automoble parts (e.g. n the context of desgnng, R&D, prototypng and testng). On the other hand, the ncreasng pressure on the producton costs n the automoble ndustry reduces the margn of suppler frms to nvest n basc and appled research actvtes n order to develop new and technologcally novel parts. Ths requres an effcent utlzaton of the nhouse R&D capactes as well as an mprovement of complementary technologcal knowledge from outsde (Davd 1994). Thus, frms n the automoble supply ndustry have to ncrease the use of external sources of new technologcal knowledge to expand ther technologcal capactes. One possblty of replenshng scarce R&D resources s to adapt knowledge generated n academc nsttutons as a part of the total stock of technologcal opportuntes frms can utlze from outsde for ther own purposes. As stressed by Rosenberg and Nelson (1994), major nnovaton mpulses, e.g. n modern mcro-electroncs, are based on academc research, generc technologcal knowledge and ts drect adaptaton and utlzaton n appled ndustral research. Also, for the automoble supply ndustry, the mportance of scentfc knowledge has grown contnuously over tme because the development and producton of automoble parts depends strongly on the fndngs and results from academc research. Automakers ncreasngly rely on external developers closely connected to academc nsttutons. The reasons for ths can been seen n the nterdependence of technologcal progress and scentfc research, and the steadly ntensfed nterrelaton of fundamental and ndustral research (Freeman 1982, Narn and Olvastro 1992). 1 Accordng to the classfcaton of the level of average R&D ntenstes done n the Frascat Manual (OECD 1994), the automoble ndustry can therefore add to the hgher valued technologcal sectors.

3 2 The dependence on scentfc verfcaton of technologcal developments has led to a rsng demand for hghly qualfed personnel (scentsts, engneers, techncans, etc.) n the automoble supply ndustry 2 and for research results from the academc sphere (Peters 1997). Ths s caused by the ncreasng mportance of multdscplnarty of R&D actvtes (Stephan 1996), the growng complexty of nnovaton processes (Becker 1996), and the ncreasngly shorter ntervals for developng new automoble parts (Clark and Fujmoto 1993). For the US manufacturng ndustry, the relevance of academc research n the nnovaton process, measured by varables regardng nnovaton nput and output, has been nvestgated thoroughly (Acs et al. 1992, Jaffe 1989, Mansfeld 1991). For German ndustry, however, the mportance and effects of acadamc research on ndustral nnovaton actvtes on both levels are theoretcally and emprcally less nvestgated. 3 An excepton s the econometrc study of Becker and Peters (1997). They have analysed the relevance and mpacts of academc research on the nnovaton nvestments and nnovaton output of frms n the German manufacturng sector, usng data from the Mannhem Innovaton Panel (MIP) for the perod of They found hghly sgnfcant and stmulatng effects of German unverstes as external sources of technologcal nformaton on frms nnovaton nput, even when controllng for ndustry effects. But nfluences of scentfc-based technologcal opportuntes on the nnovaton output of frms n the German manufacturng ndustry were ambguous. No emprcal evdence was found for sgnfcant stmulatng mpact of scentfc and unversty knowledge on the development of new products, but all the more on the development of mproved products. In ther concluson, Becker and Peters (1997) assume that the mpact of academc research dffers n the technology felds used and n the peculartes of ndustres as ponted out by Cohen (1995). The am of ths paper s to analyse the mportance and mpacts of technologcal knowledge generated n academc research on the nnovaton actvtes n the German automoble supply ndustry as a specfc sector wth a hgh level of R&D actvtes and to compare these results wth the fndngs of Becker and Peters (1997) for the German manufacturng ndustry on the whole. In the automoble supply ndustry we observe dfferent technologes n dfferent lnes of busness for automoble parts, components or systems but all lnes of busness show almost the same demand structures. Thus, for a gven ndustry we can nvestgate whether the varaton of nnovaton and R&D actvtes can be explaned by dfferences n the technologcal opportuntes caused by the adaptaton of knowledge generated by academc 2 In the tme perod the share of hghly qualfed personnel wth academc educaton has been ncreased n the automoble ndustry (West Germany) from 6.2 per cent to 7.9 per cent (Federal Mnstry of Educaton, Scence, Research, and Technology 1996). 3 The exstng studes n Germany focus ther nvestgatons on specal aspects of the scence-technologynnovaton nterface, e.g. the mportance of academc research n scence-based felds of technology (Grupp 1994, 1996), the role of unverstes n the technology transfer process especally for small and medum-szed frms (Bese et al. 1995, Harhoff and Lcht 1996) or the relevance of the regonal scence and research nfrastructure on the formaton and locaton of frms (Harhoff 1995, Nerlnger 1996).

4 3 research. In ths context, we also examne the nfluence of dfferences n supplers absorptve capactes because suffcent expertse s requred n order to understand and mplement external nformaton for ther own purposes (Veugelers 1997). In the followng secton, basc theoretcal consderatons are made about the nterrelaton of technologcal opportuntes, academc research, and nhouse nnovaton (R&D) actvtes. In the thrd secton, the role of academc research for the German automoble (supply) ndustry s dscussed, by usng descrptve statstcs. Thereafter, the data sets at hand are descrbed, and the defntons of the nnovaton measures are gven. In the ffth secton, we present the results of several regresson equatons related to the effects of academc research as an external knowledge source on the nnovaton nput and output of frms n the German automoble supply ndustry. In the sxth secton, we checked some of our estmatons for dfferences n supplers capactes to absorb technologcal nformaton by contractual co-operatons wth unverstes. The last secton closes wth concludng remarks. II. Theoretcal Consderatons about Technologcal Opportuntes, Academc Research, and Innovaton Actvtes It s generally accepted that the level of the nnovaton actvtes of frms and/or the techncal progress n ndustres depends on ther opportuntes to acqure and nternalze usable technologcal (techncal) knowledge (Gerosk 1990, Harab 1995, Klevorck et al. 1995). As recent econometrc studes show, varances n R&D expendtures and nnovaton actvtes can be explaned better by dfferences n the level of technologcal opportuntes each frm or ndustry s faced wth rather than by other factors, such as sze, market structure, approprablty, etc. (for a survey, see Cohen 1995). The theoretcal concept of technologcal opportuntes dffers from ths pont of vew. Wthn neo-classcal theory technologcal opportuntes can be descrbed as "... the set of producton possbltes for translatng research resources nto new technques of producton that employ conventonal nput" (Cohen and Levn 1989, p. 1083). In the framework of the evolutonary theory of techncal progress, technologcal opportuntes are seen as potentals of new technologes wth (sometmes unknown) relatonshps to other technology felds or as techncal advance along a technologcal trajectory (Coombs 1988, Dos 1988). But t s not the am of ths paper to make a contrbuton to obtan a...consensus on how to make the concept of technologcal opportunty precse and emprcally operatonal (Cohen and Levn 1989, p. 1083). Rather we assume - n lne wth Cohen 1995, Dos 1988 and Levn et al that technologcal opportuntes relate to the contrbuton of external (knowledge) sources to frms nnovaton actvtes, and that there are dverse knds of sources of technologcal opportuntes wth varyng usefulness not only from ndustry to ndustry but also from one frm to the other (Harab 1995). One way for frms to expand ther technologcal opportuntes s to apply external knowledge orgnatng from supplers, customers and/or compettors. Another possblty s the adaptaton of knowledge from sources besde the

5 4 ndustral sector, here n partcular from the academc sphere. Scherer (1992, p. 1424) ponts out that "... the mysterous concept of technologcal opportunty was orgnally constructed to reflect the rchness of the scentfc knowledge base tapped by frms". Informaton from scentfc nsttutons are mportant sources for frms to expand ther technologcal capactes because of the close nterrelaton of fundamental academc research and ndustral research. We assume that the degree to whch frms use technologcal nformaton stemmng from scentfc nsttutons (unverstes, techncal nsttutons) s closely correlated wth ther technologcal capactes to develop new or mproved products (Arvants and Hollensten 1994, Harab 1995). In ths lne, the adaptaton of external knowledge from unverstes ncrease frms (nhouse) nnovaton capactes wth postve effects on the nnovaton process. But the mportance and usefulness of academc research for frms nnovaton actvtes dffer n the felds of unversty research. Klevorck et al. (1995, p. 201) dstngush ndustres that are more strongly nfluenced by advances n basc research (e.g. drug ndustry) from those that are much less nfluenced (e.g. arcraft), and ndustres wth rch technologcal opportuntes (e.g. electronc components) from those whch are not (e.g. metal products). However, before we emprcally analyse the effects of the knd and level of technologcal opportuntes, nfluenced by the adaptaton of external knowledge from academc research, on the nnovaton nput and output of frms n the German automoble supply ndustry emprcally, we wll present a smple theoretcal framework descrbed n more detal by Becker and Peters (1997). Followng ths framework, we assume that the technologcal opportuntes Ω result from the sum of the total approprablty stock of externally accumulated knowledge frm s faced wth: Ω C I S = Ω( R, R, R ), (1) where R C I and R represent the technologcal opportuntes brought forth by compettors and S other ndustres (e.g. supplers, customers). R reflects the contrbuton of scentfc knowledge to the technologcal capactes of frm. Because the technologcal opportuntes (1) are represented by external nformaton, the margnal effects of ncreasng the external C I S resources R, wth R = R, R, R, on the level of technologcal opportuntes s strctly 2 2 > postve ( Ω / R > 0 ) wth constant, dmnshng, or ncreasng returns ( Ω /( R) < 0 ), dependng on the ntal level of frms technologcal capactes. We assume that the nnovaton output w s nfluenced by the level of technologcal opportuntes Ω and frms own R&D nvestments R : 4 w = w( R, Ω ), (2) 4 Ths vew dffers from the defnton of Grlches (1979, p. 98) who characterzes technologcal opportuntes as "... one or more parameters n a producton functon relatng research resources to ncrements n the stock of knowledge, wth the stock of knowledge enterng n turn as an argument, along wth conventonal nputs, n the producton for output." In our defnton, technologcal opportuntes are nterpreted as an argument contaned drectly n the producton functon of nnovaton w.

6 5 wth followng condtons: w 2 / R > 0, / Ω >0, w / R Ω > 0 (2 ) w w /( R ) > 2 2 < 0, /( Ω ) > < w Hgher nvestments n nhouse R&D enlarge frms nnovaton output (e.g. mprove the qualty of own products or reduce the costs of producton) wth dmnshng, constant, or decreasng rates of return, dependng on the level of nvestments. The same condtons reman for the mpact of the level of technologcal opportuntes on frms nnovaton output w. Thus, gven the level of own R&D, an expanson of the technologcal opportuntes ncreases frms nnovaton capactes (potentals) and nfluences the nnovaton output. Whereas the mpacts of usng external knowledge (from unverstes and other scentfc nsttutons) on frms nnovaton output seem to be clear, the effects on the nnovaton nput (e.g. R&D expendtures) are ambguous. The decsve queston s, how and to whch extent an expanson of Ω by utlzng external generated knowledge nfluences the nnovaton actvtes of frm on the nput sde. We concentrate our consderatons on technologcal opportunes resultng from knowledge from the academc sphere: Ω = Ω( ). Accordng to former consderatons (Becker and Peters 1997), we dstngush three effects of Ω on frms own (nnovaton) R&D nput: (1) the productvty effect, (2) the absorptve capacty effect, and (3) the substtuton effect. (1) The nput effect of the level of technologcal opportuntes on the margnal return relates to the argument that the ncentve of frm to nvest n R&D s postvely correlated wth the level of technologcal opportuntes. Ths concept "... corresponds to the functon that maps the flow of R&D nto ncreases n the stock of knowledge" but wth dmnshng returns of nhouse R&D at the margn (Klevorck et al. 1995, p. 188). In ths context, hgher levels of technologcal opportuntes enhance the margnal effect of nhouse R&D on product qualty or cost reducton (as shown n (2 )) as well as on frms proft whch stmulate ther R&D nvestments. The productvty of nhouse R&D whch maxmzes profts s an ncreasng functon of the level of technologcal opportuntes. It s much easer for frms to realze a gven (product or process) nnovaton wth nhouse R&D, f the pool of technologcal opportuntes s replenshed (Rosenberg 1974). (2) The effect of the level of technologcal opportuntes on the absorptve capacty relates to the technologcal opportuntes usable for frms nhouse R&D ~ Ω R S = λ Ω whch depends on frms ablty to adapt external generated knowledge for ther own purposes. The parameter λ descrbes the extent to whch frm can absorb external generated knowledge and the easness to whch t has access to such resources, wth 0 λ 1 (Cohen and Levnthal 1989). At λ = 1 the degree of utlzaton s complete. λ = 0 shows there are no possbltes of utlzng external knowledge.

7 6 As Mowery and Rosenberg (1989) stress, frms have to nvest n complementary nhouse R&D to understand and mplement the results of externally performed R&D and to obtan full access to the research fndngs of other frms and nsttutons ( ~ Ω Ω or λ 1). 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 external sourcng on the level of R&D spendng, f frms have absorptve capacty establshed as a full-tme staffed R&D department. In ths context, unversty knowledge should requre hgher absorptve capactes than nformaton from other sources because t s related more to basc than to appled scence. As Cohen and Levnthal (1989, p. 589) remark "... basc scence s less targeted to the needs and concerns of the frm" whch mplcates that frms must nvest more n ther absorptve capactes to assmlate and explot scentfc knowledge. 3) The substtuton effect refers to the fact that the adaptaton and mplementaton of external resources can reduce frms nnovaton expendtures, f external knowledge (Ω ) can be used as a substtute for own (generc) R&D. Ths s the case when frms have to nvest n dosyncratc and generc R&D to realze an nnovaton output. To clearfy ths, we modfy the relatonshp (2) to d ge w = w( R, R, Ω ), (3) wth the same condtons shown before. Whereas dosyncratc R&D actvtes ( R d ) prmarly create frm-specfc knowledge, generc R&D actvtes ( R ge ) produce nformaton whch have more the character of a publc good (Nelson 1992). New generc nformaton (knowledge) can spll over to other actors wthout purchasng the rght to do so. 5 We argue n lne wth Harhoff (1996) that nvestments n dosyncratc and generc R&D are strategc complements: decreasng (ncreasng) the level of generc R&D lowers 2 d ge (stmulates) the nvestment n dosyncratc R&D ( w / R R > 0, 2 d w / R Ω > 0 ). The adaptaton of external (academc) knowledge wll be a proft enhancng strategy, f the costs of searchng and usng externally generated knowledge are lower than the generaton of generc knowledge nhouse. As Harhoff (1996) shows, f frms substtute ther generc part of own R&D up to the level of generc R&D done formerly nhouse, they wll strctly reduce ther whole R&D cost. Gven the effcency of generc R&D nformaton, the cost of generc R&D wll decrease, whereas the amount of dosyncratc R&D nvestment can not be hgher than formerly wth nhouse engagement n generc R&D. Only f frms decde to utlze more generc knowledge from external sources than 5 R&D spllovers are externaltes of R&D actvtes beyond ther prmary defnton, where not the nnovator alone has the beneft, but whch can be appled also by other actors for ther own nnovatve actvtes (Dos 1988, Elasson 1996, Grlches 1992). They bascally change the characterstcs of product factor nputs requred for nnovatons. R&D spllovers between compettors can reduce frms ncentve to nvest n R&D because of the nherent approprablty problem (Spence 1984).

8 7 they have formerly generated nhouse, the level of dosyncratc R&D wll rse. But t s mpossble to make a clear statement about the level of frms R&D nvestment on the whole. If the elastcty of dosyncratc R&D wth regard to generc R&D s small (hgh) the entre R&D costs can be lower (larger) wth the utlzaton of scentfc knowledge than formerly wth generc R&D done nhouse. Thus, the extent of nnovaton actvtes and the level of nnovaton expendtures can be lower n the case of a hgh level of technologcal opportuntes than n the case of a low level. The whole mpact of technologcal opportuntes, e.g. stemmng from academc research, on the nnovaton nput depends on the strength of the dfferent effects of usng external sources of technologcal knowledge. But t also depends on the nteracton of the three effects mentoned. For example, f frms have hgh (low) absorptve capactes to utlze external knowledge, the nfluence of technologcal opportuntes on the productvty of ther nhouse R&D may also be hgh (low). It can be assumed that frms wth hgh (low) absorptve capactes have more (less) potentals to use external generated knowledge for own purposes and therefore can mplement external knowledge more (less) effcently nhouse. 6 For ncreasng effcency n the utlzaton of generc R&D, t s also more lkely for frms wth hgh absorptve capactes that the substtuton effect of technologcal opportuntes s of low mportance regardng to the level of R&D spendng. Because of the ncreased effcency, frms wth hgh absorptve capactes use more external generc R&D than formerly done nhouse whch enhances the margnal effect of dosyncratc R&D. At least, they wll nvest more n ther nnovaton actvtes than other frms. Summarzng, whereas the adaptaton and mplementaton of technologcal opportuntes stemmng from academc research may dscourage (substtute) or encourage (complement) frms' nnovaton/r&d nvestment, t s more lkely that hgh (low) levels of technologcal opportuntes ncrease (decrease) the level of frms' nnovaton output. Wthn ths theoretcal framework the mportance and effects of academc research on the nnovaton nput and output of frms n the German automoble supply ndustry wll be emprcally nvestgated n the followng sectons. III. Academc Research and Innovaton Actvtes n the German Automoble Supply Industry Due to the lack of offcal data for descrbng the nnovaton actvtes of automoble supplers and ther lnks to academc sources of external knowledge, we used data derved from two sample surveys conducted n the German automoble ndustry n the summer of 1995 and n the sprng of 1996 (Peters 1997). These samples contan automoble supplers actng on more than 18 lnes of economcs (SYPRO four dgt) and 30 lnes of busness. 6 Gambardella (1992) found emprcal support for ths assumpton n the US drug ndustry. Frms wth hgher levels of nhouse R&D mplement nformaton of scentfc research more effectvely than other frms.

9 8 The frst questonnare contaned data on the nnovaton actvtes of automoble supplers, ther technologcal opportuntes, approprablty condtons, and on the knd of lnkages to ther customers (automakers or other automoble supplers). Out of an ntal sample of 1,306 automoble supplers, 460 frms returned the questonnare. After excludng frms n the felds of engneerng, consultng and tunng, 401 nnovatve as well as non-nnovatve supplers of automoble parts, components or subsystems remaned. An nnovatve suppler was defned as a frm whch has ntroduced new or mproved products or processes n 1993 or In both of these years 78.0 per cent of the automoble supplers had successfully ntroduced new products n ther markets, whereas 74.5 per cent had successfully mplemented new processes n ther frms. In our analyss we ncluded only the 348 nnovatve supplers. The probablty of realzng new or mproved products or processes depends heavly on the sze of the frm. Smaller frms wth less than 250 employees realzed much less nnovatons than larger frms (Peters 1998). On average, 38 per cent were small frms (less than 250 employees), 23 per cent were medum-szed frms ( employees), and 39 per cent large frms (500 and more employees). Next, n 1996, all automoble supplers whch had contrbuted to the ntal survey were also asked to ndcate the mportance of academc research for ther nnovaton actvtes and the knd of co-operatons wth unverstes. 138 nnovatve automoble supplers returned the questonnare, whereas four supplers n the feld of engneerng and consultng were excluded. The Innovaton Actvty of Automoble Supplers Before we turn to the emprcal results related to the mportance of the academc research to enhance frms technologcal opportuntes, we gve a bref descrpton of the nnovaton actvtes of German automoble supplers n A more complete descrpton of the sample can be found n Peters (1997). The data clearly llustrate the hgh level of nnovaton actvtes n the German automoble ndustry. On average, the R&D-employment ntensty - whch relates to the percentage of employees performng R&D tasks wthn or outsde R&D departments (no full tme equvalents) - was 5.3 per cent wth a maxmum value of 30.0 per cent (Fgure 1). About 75 per cent of the nnovatve automoble supplers n the data set had a formal R&D lab n One ffth of them nvested n basc research, whereas three quarters of the supplers were actve n appled R&D regulary. However, 90 per cent remarked that nhouse R&D s neccessary for ntroducng new products or processes successfully. Solely for foundres, R&D s much more unmportant for realzng nnovatons than on average. As Klenknecht and Rejnen (1991) state, for a successful ntroducton of new products or processes frms not only need to perform R&D but to engage n related actvtes as well. The nnovaton ntensty captures further characterstcs of frms nnovaton nput, such as desgn and concepton, constructon, engneerng, and fabrcaton of prototypes, tral producton, plot

10 9 plans, etc. 7 The nnovaton expendtures per unt of sales (wthout costs of patentng, lcensng, and staff tranng) were 5.5 per cent n average. As seen n Fgure 1, R&D and nnovaton ntenstes vary among the lnes of economcs. Frms producng chemcal materals receve the hghest ntenstes, whereas foundres, drop forges and steel shapng frms receve the lowest ntenstes. In addton, nnovatve supplers spend about 354,000 German Mark per unt of employee or 10,600 German Mark per unt of R&D employee n ther nnovaton actvtes. INSERT FIGURE 1 HERE The survey also provdes nformaton on the nnovaton output, e.g. on the mportance of nnovaton. For 42.8 per cent of the nnovatve supplers the ntroducton of new automoble products was more mportant than the mplementaton of new processes. But they try to mprove exstng products or processes rather than to develop basc nnovatons. Only for 11.5 per cent of the supplers the realzaton of fundamentally new nnovatons was more mportant than the techncal mprovement of exstng automoble products or processes. Sources and Acquston of External Knowledge After ths bref descrpton of the nnovaton behavor of frms n the German automoble supply ndustry, we want to look at the relevance of academc (scentfc) research as an external source of knowledge to ncrease frms technologcal opportuntes. Dfferent measures were used to show the nfluence of technologcal opportuntes and academc research on the nnovaton actvtes. Followng Levn and Ress (1988), we assume that the degree to whch frms are nterested n scentfc nsttutons (unverstes, techncal nsttutons) as relevant sources of technologcal nformaton s closely correlated wth the level of ther technologcal capactes (see also Arvants and Hollensten 1994, Felder et al. 1996, Harab 1995). Therefore, the nnovatve frms were asked to rate on a seven-pont scale the mportance of several external sources of technologcal nformaton for ther nnovaton actvtes n the years As shown n Table 1, nter-ndustral sources (customers and supplers) were rated as the most mportant nformaton sources for the nnovaton actvtes, followed by compettors as an ntra-ndustral nformaton sources. Academc nsttutons (unverstes or professonal assocatons) were ranked as less mportant sources for external technologcal knowledge. INSERT TABLE 1 HERE These results are consstent wth the fndngs of Becker and Peters (1997) for the German manufacturng ndustry on the whole, usng data from the frst wave of the Mannhem Innovaton Panel conducted n Also Harab (1995, p. 70) shows for the Swss ndustry 7 In the OSLO-Manual, expendtures for patentng and lcensng as well as staff tranng are also part of the nnovaton expenses (OECD 1994). However, we only taken nto account actvtes whch refer to the course of the nnovaton process drectly (Clark and Fujmoto 1991).

11 10 that the contrbuton of ndustral sources to techncal progress - especally for frms wthn the same lne of busness - appears to be most mportant, whereas non-ndustral sources (unversty research, other government research nsttutons, state companes and agences, professonal assocatons, etc.) were rated as relatvely unmportant. Another possblty for automoble supplers to expand ther technologcal opportuntes s the establshment of R&D co-operatons (Sterlacchn 1994), n partcular wth academc nsttutons. R&D co-operatons between nnovatve supplers and academc nsttutons can be set up more or less systematcally. The spectrum of formal and nformal lnkages ranges from rregular collaboraton, such as sporadc consultng servces (e.g. expertse, evaluaton reports, etc.), and regular exchange of nformaton and experence (workshops, symposa, etc.), to systematc contractual agreements wthn mutually fnanced R&D projects. In the survey, the automoble supplers were asked whether they had formed contractual R&D co-operatons wth other nsttutons or frms. Generally, 62.8 per cent of the nnovatve frms had sgned up R&D contracts wth other frms or unverstes to jontly develop new products or processes (Table 2). Whereas the contrbuton of technologcal nformaton from unverstes to the nnovaton process s generally consdered as less mportant, the most nnovatve automoble supplers n 1993 and 1994 had formed formal R&D co-operatons wth unverstes rather than wth ndustral partners. Not suprsngly, supplers wth academc R&D-co-operatons rated the mportance of unverstes as external sources of technologcal nformaton sgnfcantly hgher than other frms (mean of 4.15 to 2.67). But no nterndustral dfferences could be found by frms responses to the queston of R&D co-operaton wth unverstes. INSERT TABLE 2 HERE Table 2 also shows the rankngs related to the mportance of dfferent co-operaton partners for frms n the German manufacturng sector on the whole. The ntensty of co-operaton s much hgher n the automoble supply ndustry than n the average of the whole manufacturng sector. In the manufacturng sector only 37 per cent had developed new products or processes jontly wth other frms or publc organzatons. Even f the frms from the Mannhem Innovaton Panel were asked for formal and nformal R&D co-operatons, the level of jont development s much hgher n the automoble supply ndustry than n the whole manufacturng sector. At least n the second survey, conducted n 1996, the nnovatve automoble supplers were asked to evaluate the mportance of dfferent felds of academc research for ther nnovaton actvtes wthn the last fve to ten years. We used felds of academc research, specfed by Harab (1995) for the Swss manufacturng sector, n whch research seems to be foccused on facltatng technologcal advance of varous knds (see also Klevorck et al. 1995). Table 3 lsts the relevance of dfferent felds regardng basc scence, appled scence, and engneerng scence for suppler frms n the German automoble ndustry. New fndngs n theoretcal

12 11 basc research, appled mathematcs and operaton research were ranked as less mportant than research results n materal scence, mechancal engneerng, and nformatcs and computer scence. As Klevorck et al. (1995, p. 190) stress, work n felds lke metallurgy, materal scence, computer scence, electrcal engneerng, and pathology, all of whch are strongly represented n academa as well as n ndustry, drectly facltates technologcal advance and enhances the problem-solvng capacty of those who endeavor to make such progress. INSERT TABLE 3 HERE But the hgh standard devatons suggest that the relevance of academc research heavly depends on the technology produced. Thus, we examned the relevance of felds of academc research for effects n lnes of busness. Academc research n electronc scence or electrcal engneerng was relevant only for supplers of electronc components, whereas the results of appled chemstry were mportant solely for supplers of chemcal materals and plastc parts. In contrast, automoble supplers n seven out of the nne nvestgated lnes of economcs rated new fndngs of the academc research n mechancal engneerng and materals scence as mportant. The emprcal results of the descrptve analyss of the mportance of academc research as an external knowledge resource to enhance the technologcal opportuntes of frms n the German automoble supply ndustry can summarzed as follows: (1) The contrbuton of academc research on supplers nnovaton actvtes s generally cted as less mportant than that of nter-ndustral or ntra-ndustral organzatons. (2) Despte ths lower mportance, unverstes are the most lkely partners of formal R&D arrangements for jontly developng new products or processes. (3) New fndngs n materal scence, nformatcs and computer scence, and n mechancal engneerng are the most mportant academc research felds for the German automoble supplers. IV. Measurement Issues and Data Descrpton In a next step, we want to estmate the effects of academc reserach and other factors on the nnovaton nput and output of frms n the German automoble supply ndustry. The basc specfcaton for explanng supplers nnovaton actvtes x s as follows: x = a + b ATO +cito +dto+etc +fmc +gap + ha +ε, (4) where ATO and ITO represent proxes of supplers technologcal opportuntes stemmng from academc research and ndustral nformaton sources, TO reflects the level of technologcal opportuntes on the lnes of economcs. TC defnes a parameter of supplers technologcal capablty, MC represents condtons of supplers as well as the buyers market structures, AP stands for the degree of approprablty, and A represents condtons regardng the sze and the degree of nterndustral dversfcaton of frms. Table 4 shows the dfferent exogenous varables used n the estmatons for the year 1994.

13 12 We use three dfferent ndcators for measurng the level of technologcal opportuntes related to the relevance of academc research or other sources of new knoweldge: (1) the ratng of the mportance of unverstes and other frms or nsttutons as sources of technologcal nformaton for frms nnovaton actvtes, (2) dummes for supplers wth formal R&D-cooperatons wth unverstes and customers, and (3) the relevance of dfferent felds of academc research. ad (1): Followng Arvants and Hollensten (1994), and Felder et al. (1996), we reflect the technologcal opportuntes of supplers by the contrbuton of three dfferent sources of nformaton to ther nnovaton actvtes, employed by the scores of factor analyss (see Appendx Table A1): scentfc nsttutons lke unverstes and professonal assocatons (TEC_SCIE), customers and supplers (TEC_CUSU), and compettors/reverse engneerng (TEC_COMP). 8 We take supplers ratngs of the mportance of unversty nformaton (TEC_UNI) separately n the regresson to stress the relevance of acadamc research on ther nnovaton actvtes. ad (2): We use a dummy varable COOP_UNI whch takes a value of 1, f supplers have formed contractual R&D co-operatons wth unverstes. Further, the dummy varable COOP_CUST s ncluded to check the nfluence of co-operatve R&D arrangements among supplers and ther customers. As Veugelers (1997) and Colombo and Garrone (1996) stress, a problem of smultanety can arse f supplers probablty of co-operatng depends on ther nhouse R&D actvtes and vce versa. But we consder that supplers regularty to nvest n R&D rather than the level of nhouse R&D determne supplers wllngness to engage n cooperaton, n partcular wth unverstes or customers. Therefore, we assume no causalty relatonshp between supplers co-operatve agreements wth unverstes or customers and ther ntenstes of R&D (nnovaton) actvtes. ad (3): To nvestgate the effects of dfferent felds of academc research on the nnovaton actvtes of frms n the German automoble supply ndustry we used the results of a factor analyss of the data from the second sample to obtan four factors descrbng specal felds of academc research (Appendx, Table A2): theoretcal and appled mathematcs, physcs, computer scence (SCI_MATH), materals, metallurgy, mechancal engneerng (SCI_MATL), electroncs and electro-techncs (SCI_ELEC), and theoretcal and appled chemstry (SCI_CHEM). In ths context, we delete ndustres wth less than three observatons. We then computed the mean values of the four factors for all lnes of economcs (see Fgure 1) and lnked them wth the data from the frst sample. Industral effects of the level of technologcal opportuntes were also measured by the factorspecfc contrbuton of technologcal know-how from ndustral research (R&D_CAP). Usng data from Meyer-Krahmer and Wessels (1989), a dummy varable reflects the lnes of 8 Reverse engneerng s seen as one mportant possblty for frms to learn about the technologcal know-how of compettors.

14 13 economcs wth above average stocks of R&D-captal (chemcals, automobles and electronc parts). Herewth, we correct for fxed ndustry effects. 47 per cent of the supplers belongng to these lnes of economcs have above average stocks of R&D captal. To capture supplers nternal technologcal capactes (TEC_CUST), we used the responses to questons concernng the share of sales of automoble parts developed to the customers specfcatons (detal-controlled parts). Accordng to Cusumano (1989), automoble supplers wth low technologcal capactes make fewer R&D efforts wth parts controlled n all detals by ther customers than frms wth hgh technologcal capactes whch manly produce own developed and black-box parts. But the varous specfcatons of automoble parts also show aspects of demand structures and the adaptaton of external nformaton (von Hppel 1988; Peters 1998). So, automoble parts produced from own suppler drawngs may be less customer-specfc than detal-controlled or black-box products. Consequently, automoble parts developed to customers drawngs, can be offered to a smaller number of customers, whch promotes the market power of potental buyers. On average, 39.4 per cent of supplers automoble sales had been done wth detal-controlled parts. In our estmatons we also ntroduced varables reflectng further characterstcs of supplers and specfc market condtons. As a proxy varable for busness unt sze the sales of automoble parts (n logs) s chosen (SIZE). Ths catches the effect of frm szes specfc to the automoble sector. As the lterature does not offer a standard economc nterpretaton of sze effects (Arvants and Hollensten 1994), busness unt sze can be used as a proxy for varous economc effects, and therefore can have ambguous effects on the nnovaton actvtes of supplers. In the samples used, busness unt sze s strongly correlated wth the whole sze of frms measured by the log of the number of employees. In order to control nter-sectoral synergy effects by the development of new products and processes, a dummy varable (DIVERS) was ncluded, reflectng the 18 per cent of supplers havng made less than 40 per cent of ther sales wth automoble parts n We assume that supplers actng n more than only n the automoble supply ndustry have more opportuntes to adapt external knowledge generated n other sectors. INSERT TABLE 4 HERE Data on market concentratons were drawn from our survey and from the German Monopoly Commsson (1994). To avod the problem of smultanety wth nnovaton determnng market concentraton, we use lagged concentraton data referrng to The concentraton n the domestc suppler markets s measured by the Herfndahl ndex, computed for the lnes of busness unts (SCR). Where such a classfcaton was not feasble, the SCR of the respectve two-dgt ndustry s used. For descrbng the suppler concentraton n domestc and foregn markets, we use a dummy varable (COMP_H), whch characterzes olgopolstcally structured markets wth 1 to 10 compettors. About 74.4 per cent of the sample frms had up to ten compettors.

15 14 Followng Peters (1998), who found sgnfcant effects of buyer concentraton on the nnovaton nput and output of automoble supplers, we use a sales weghted CR3- concentraton rato of domestc buyer markets (BCR). The weghts of these buyer concentraton ndex relate to the shares of supplers sales made drectly wth automakers, commercal vehcles or wth frst- and second-ter supplers n The nteracton of domestc buyer and suppler market structure s measured for each frm by the product of the suppler and buyer market concetraton rato SCR*BCR. In addton, a dummy was ntroduced to defne all 21.6 per cent of supplers wth more than 10 buyers (customers) wthn the automoble ndustry at home and abroad (CUST_H). Because the desgn of suppler contracts s most mportant for capturng nnovaton returns n the automoble ndustry (Peters 1998) three varables related to frms approprablty condtons are ncluded n the estmatons. The relevance of contractual appropraton condtons s measured by a dummy varable (APPR_CO), characterzng the 17.3 per cent of supplers who use long-term contracts (model-lfe contracts wth a tme of delvery of 3 to 5 years) to a hgher extent than short-term contracts (contracts of one year or less). The extent to whch supplers were confronted wth declnng revenues n 1993 and 1994, due to a prce pressure of the buyers, s also used as an ndcator for frms approprablty condtons (APPR_PR) wth a mean of Fnally, we use the ratngs of a queston n our survey concernng the effcency of patents to approprate new technologcal fndngs as an exogenous varable (APPR_PA). In the sample, about 17 per cent of the supplers reported that they can protect ther R&D fndngs wth patents very well. Dependng on the knd of nnovaton ndcators, dfferent estmaton methods are used to analyse the effects of academc research on the nnovaton actvtes of frms n the German automoble supply ndustry. To estmate the dfferent effects on the nnovaton nput (see Table 5), such as the nnovaton ntensty (INNO_INT), the R&D employment ntensty (R&D_INT), the nnovaton employment ntensty (INEMPL_INT), and the nnovaton R&D employment ntensty (INR&D_INT), the standard Tobt model s employed because some nnovatve frms have no R&D employment or have not nvested n ther nnovaton process. But ths method does not allow the dentfcaton of parameters affectng supplers decson to partcpate n R&D and the extent of ntensty lke the two-step verson of the Heckman model used by Becker and Peters (1997). 9 An ordered probt model wth sample selecton s employed for estmatng the extent of R&D to realze an nnovaton (from 0 = actually none to 6 = to a hgh extent). As some nnovatve frms have not performed any R&D, the probablty for supplers havng to engage n R&D n order to be successful n realzng an nnovaton 9 We have tred to use the two-step verson of the Heckman model to estmate INNO_INT and R&D_INT. But only less than than 2.5 per cent of the nnovatve supplers reported no nnovaton expendtures or R&D employment. Ths small number has rejected the applcaton of the Probt estmatons n the frst step.

16 15 (R&D_NEC), and the extent of R&D for realzng an nnovaton (R&D_EXT) has been estmated. 10 INSERT TABLE 5 HERE We preferred a double censored Tobt model for analysng the effects of academc research on the nnovaton output. The mpacts were checked accordng to the development of really new (INNO_NEW) and mproved (INNO_IMP) products or processes. These ndcators were computed by factor analyss on supplers ratngs regardng the mportance of the development of dfferent types of nnovatons (Appendx, Table A3). When the problem of (multplcatve) heteroscedastcty was present, we corrected the standard devatons of the estmated parameters. V. Regresson Results We nvestgated the mpacts of academc research knowledge on the nnovaton actvtes of frms n the German automoble supply ndustry under dfferent aspects. In Model 1 we tested the effects of the engagement of supplers n R&D co-operatons wth unverstes to enlargen ther technologcal opportuntes (COOP_UNI). In the basc specfcaton (4), technologcal opportuntes stemmng from academc research were modfed to b ATO = b 1 COOP_UNI + b 2 SCI_MATH j + b 3 SCI_MATL j +b 4 SCI_ELEC j, (5) whereas j ndcates the lne of busness of frm. As the factor Chemstry s hghly negatve, correlated wth the factor Electroncs, we excluded the frst factor (or otherwse). In Model 2 we checked the nfluence of unverstes as an external nformaton source (TEC_UNI). Because of the problem of multcollnearty we have had to exclude COOP_UNI from Model 2. The specfcaton (5) was modfed as b ATO = b 1 TEC_UNI + b 2 SCI_MATH j + b 3 SCI_MATL j +b 4 SCI_ELEC j.(6) In both models we took nto consderaton the varables representng technologcal opportuntes stemmng from ndustral sources: c ITO = c 1 TEC_COMP + c 2 TEC_CUSU+ c 3 COOP_CUST. (7) Snce the estmatons testng the mpact of unverstes as an external source for technologcal nformaton on the level of factor scores (TEC_SCIE) n general yeld the same sgn of coeffcents (but wth lower statstcal sgnfcance) as wth TEC_UNI, the fndngs related to these aspects are not mentoned. 10 The ordered probt model wth sample selecton can be descrbed as follows (Greene 1990): In the frst, we estmate z*=γ w+u, wth u [0,1], whereas z* s unobserved. But we observe z=1 f z*>0 and z=0 f z 0. In the second, y satsfes the ordered probt specfcatons, y*=β x+ε only f z=1, wth (ε,u) N2[0,0,1,1,ρ].

17 16 Estmaton of the Impact of Technologcal Opportuntes on Supplers Innovaton Input As seen n Table 6, technologcal opportuntes stemmng from academc research seem to substtute supplers nvestments n ther R&D and nnovaton actvtes. The effects of COOP_UNI and TEC_UNI on the R&D employment ntensty of frms n the German automoble supply ndustry (R&D_INT) are negatve but wthout statstcal sgnfcance. The estmatons for the nnovaton employment ntensty (INEMPL_INT) and the nnovaton R&D employment ntensty (INR&D_INT) are also negatve but statstcally sgnfcant. Thus, frms wth close relatonshps to unverstes spend lower nnovaton expendtures per unt of (R&D) employment than other frms. The data only reveal a small postve but nsgnfcant mpact of R&D co-operatons wth unverstes on supplers nnovaton expendtures per unt of sales (INNO_INT). In general, the level of technologcal opportuntes measured by the mportance of dfferent nformaton sources has sgnfcant (hgh) effects, n partcular for the R&D employment ntensty (R&D_INT), as the values of the Wald Test and of the F-Test show. In ths context, technologcal nformaton stemmng from compettors (TEC_COMP) or nter-ndustral sources (TEC_CUSU) seem to be more mportant than nformaton from unverstes (TEC_UNI). The adaptaton of ndustral know-how stmulates the R&D ntenstes, whereas academc nformaton reduces the R&D ntenstes of automoble supplers. In the other estmaton models, the coeffcent of TEC_CUSU has the expected postve sgn but wthout any statstcal sgnfcance. 11 At a frst glance, the fndngs for the nput effects of academc research are surprsng. Arvants and Hollensten (1995) for the Swss ndustry, and Becker and Peters (1997) for the German manufacturng ndustry, observe that academc research s postvely related to frms nnovaton nput. Perhaps, n the German automoble supply ndustry frms seem to substtute ther own (generc) nnovaton expendtures by usng externally generated knowledge from academc research. Nevertheless, we found stmulatng effects of COOP_UNI and TEC_UNI to the extent to whch nhouse R&D s necessary for supplers to realze an nnovaton. As shown n Table 7, R&D_EXT depends postvely on the mportance of academc nformaton and on the wllngness of supplers to co-operate n R&D wth unverstes. Supplers wth hgh nhouse R&D actvtes focus on academc research n order to be successful n ther nnovaton actvtes. As expected, the necessty of R&D for beng nnovatve (R&D_NEC) does not depend on the usage of the results of academc research. Rather, supplers decson to co-operate wth unverstes had been nfluenced by (the necessty of) nhouse R&D actvtes (Veugelers 1997). But we have to remark that the estmated correlaton between the error terms n the two step model s small and far from sgnfcant. Only mnor threshold effects wth respect to the necessty and extent of nhouse R&D actvtes can be observed. 11 Wth one excepton by the estmaton of the necessty of R&D to realze an nnovaton (R&D_NEC).

18 17 The values of the Wald Test show jont sgnfcant effects of the felds of academc research on the nnovaton expendtures per unt of sales, on the R&D employment ntensty as well as on the probablty and extent of R&D necessary for supplers to realze an nnovaton. But only the fndngs for the research felds electroncs and electro-techncs (factor scores) have an ndvdual nfluence on supplers R&D employment ntensty and on ther nnovaton expendtures per unt of sales. The negatve sgns of the coeffcents suggest that the adaptaton of fndngs n the academc research of electronc/electro-techncs reduces the supplers nvestment n own nnovaton actvtes. Ths can explan why academc nformaton substtutes supplers spendngs n own (generc) R&D. INSERT TABLE 6 AND 7 HERE Further, n all models, we have examned ndustral effects of technologcal opportuntes (R&D_CAP). Automoble supplers actng on markets wth hgh R&D captal stocks nvest more money n ther nnovaton actvtes per unt of employees and per unt of R&D employees than other frms. They also undertake relatvely more R&D to realze an nnovaton than frms n ndustres wth lower R&D captal stocks. But surprsngly, R&D_CAP shows a negatve sgn n the regresson of the necessty to nvest n R&D. One explanaton s that frms n these ndustres are, on the one hand, less engaged n R&D but, on the other hand, - f they partcpate - nvest more n R&D than n other ndustres. We have also studed the relevance of ndustral sources of knowledge, n partcular of nformaton generated by customers. It s nterestng that supplers wllngness to co-operate formally wth ther customers (COOP_CUST) affects ther nnovaton nput postvely. By contractual R&D co-operatons supplers have the possblty to nternalze ther customers knowledge, whch ncrease the productvty of ther own nnovatve efforts. Thus, the customers nformaton seems to be more relevant for supplers wthn formal R&D agreements than vce versa. In general, ths can explan the low nfluence of TEC_CUSU and the sgnfcant nfluence of COOP_CUST. However, n all cases we have to keep n mnd that the relevance of customers n the nnovaton process of supplers depends on the fact that customers can have nformaton about demand and approprablty condtons rather than about aspects to enlargen the technologcal opportuntes (Peters 1998, von Hppel 1988). The nfluence of customers can also be seen n the knd of automoble parts produced by supplers. Thus, the sgns of the estmated parameters of TEC_CUST - except n the estmatons for INEMPL_INT and R&D_NEC - reveal that an ncreasng share of black-box or own developed products (ncreasng technologcal capactes) had forced supplers to be actve n the R&D and nnovaton process. Strong postve effects of TEC_CUST are only observable n the estmaton of the nnovaton expendture per unt of R&D employment (INR&D_INT). As Peters (1998) shows, automoble supplers wth hgh shares of parts, developed to customers specfcatons, have a hgher probablty of realzng a process nnovaton than wth lower shares. Further, as shown n Table 6, they nvest less n ther