40 th ERSA European Congress, Barcelona European Monetary Union and Regional Policy

Transcription

1 40 th ERSA European Congress, Barcelona European Monetary Unon and Regonal Polcy Locaton and Network Effects on Innovaton Success: Evdence for UK, German and Irsh Manufacturng Plants James H Love* and Stephen Roper** * Aston Busness School, Aston Unversty, Brmngham B4 7ET, Unted Kngdom, Fax: +44 (0) E-mal: j.h.love@aston.ac.uk ** School of Management and Economcs and Northern Ireland Economc Research Centre, Queen s Unversty Belfast, Belfast BT7 1NN, Unted Kngdom. E-mal: s.roper@qub.ac.uk

2 Locaton and Network Effects on Innovaton Success: Evdence for UK, German and Irsh Manufacturng Frms Abstract Recent developments n the new economc geography and the lterature on regonal nnovaton systems have emphassed the potentally mportant role of networkng and the characterstcs of frms local operatng envronment n shapng ther nnovatve actvty. Modellng UK, German and Irsh plants nvestments n R&D, technology transfer and networkng, and ther effect on the extent and success of plants nnovaton actvtes, casts some doubt on the mportance of both of these relatonshps. In partcular, our analyss provdes no support for the contenton that frms or plants n the UK, Ireland or Germany wth more strongly developed external lnks (network or technology transfer) develop greater nnovaton ntensty. However, although nter-frm lnks also have no effect on the commercal success of plants nnovaton actvty, ntra-group lnks are mportant n terms of achevng commercal success. We also fnd evdence that R&D, technology transfer and networkng nputs are substtutes rather than complements n the nnovaton process, and that there are systematc sectoral and regonal nfluences n the effcency wth whch such nputs are translated nto nnovaton outputs. 1. Introducton Recent developments n endogenous growth theory have stressed the mportance of technologcal progress to economc advance (e.g. Romer, 1990). Lnkng techncal progress and economc growth, however, s the process of nnovaton, sometmes defned as the commercal applcaton of new or exstng knowledge. Tradtonally, nnovaton was understood as a lnear process startng wth new technologcal dscovery and endng wth new products. More recently, however, t has been recognsed that nnovaton s perhaps best understood as a contnuous and evolutonary process (Nelson and Wnter, 1982), shaped by nsttutonal routnes and socal conventons (Morgan, 1997) and the ntensty and extent of organsatonal and nter-personal nteractons (Grabher and Stark, 1997). The mportance of nnovaton, and the potental roles of nsttutons and polcy to mprove natons nnovaton record, has prompted consderable dscusson of ndvdual Natonal Innovaton Systems (most notably, Nelson, 1993) and dscusson of dfferent natons nnovaton polces (e.g. Nelson, 1993; Metcalfe, 1997; Roper, 1999). More recently, n the context of the EU debate surroundng regonal compettveness, these deas have also 1

3 been appled to the nnovaton systems of ndvdual regons (e.g. Braczyk et al., 1998; Frtsch and Lukas, 1999; Morgan, 1997). A key queston n ths strongly polcy orented debate has been on the approprateness of alternatve regonal nnovaton strateges to dfferng regonal contexts (see, for example, Hassnk, 1993; Ashem and Dunford, 1997). Referrng specfcally to Morgan (1997), Ashem and Dunford (1997) characterse supporters of a networkbased regonal development strategy as belevng that t offers the prospect of the development by devolved, ntermedate nsttutons lke regonal development agences of polces to enhance the socal captal and ncrease the nnovatve capacty of less developed regons (p. 451). In ths vew, local nter-frm networks and lnks between companes and local organsatons outsde the supply chan (e.g. hgher educaton nsttutons, development agences) provde a framework for ntraregonal learnng and knowledge transfer (Lazonck, 1993) 1. Also, and perhaps more mportant n a regonal context, are relatonshps wthn whch nter-regonal technology or knowledge transfers can take place 2. Wong (1992), for example, n an assessment of the role of multnatonal enterprses (MNEs) n Sngapore, argues that technology transfer through MNE plants may stmulate nnovatve actvty n locallyowned companes beyond that generated by ther own R&D and local networkng actvty 3. Other emprcal research suggests that such technology transfers may also explan the hgher nnovaton rates recorded for mult-plant corporatons n both the UK and Swtzerland compared wth ther ndependent ndgenous counterparts (Goddard et al., 1986; Brugger and Stuckey, 1987). Although the potental mportance for nnovaton of local networks and technology transfers have been wdely acknowledged, emprcal studes have tended to adopt a 1 In a prevous study we have shown that such network lnkages provde a drect boost to the number of new and mproved products ntroduced by UK manufacturng plants (Love and Roper, 1999). 2 One possble ndcator of the potental mportance of nward technology transfer to an area s the proporton of local patent applcatons made by non-resdents. Regonal evdence of ths sort s lmted, but n the Republc of Ireland n 1994, for example, OECD fgures suggest that only two per cent of patent applcatons were made by natonal resdents (Roper and Hewtt-Dundas, 1998, pp ). 3 For example n the context of qualty mprovement, Wong comments: the very process of the MNC mplementng ts strngent qualty/performance control system over the output suppled by the subcontractor may provde valuable feedback that wll greatly facltate technologcal learnng by the subcontractor. The more extensve the dagnostc efforts provded by the MNC buyer, the stronger the learnng facltaton wll be. 2

4 descrptve or dscursve rather than statstcal approach (see for example, Ashcroft et al., 1994; Roper and Thank, 1995 and the case-studes n Braczyk et al., 1998). Where nternatonal comparsons of regonal nnovaton performance or capablty have been made they have tended to compare specfc regons wth smlar ndustral or structural characterstcs (e.g. Jowtt, 1991; Hassnk, 1993; Cooke and Morgan, 1994). Although these regonal case-studes and comparsons have dentfed a number of factors whch dfferentate successful and unsuccessful regonal nnovaton systems (Cooke, 1992) ther prncpal weakness s ther nablty to dentfy the relatve mportance of each of these factors 4. Our object here s therefore to address four emprcal questons: Queston 1: How mportant are frm-specfc, regonal and (natonal) ndustry factors n determnng the ntensty of frms R&D, networkng and technology transfer actvtes? Queston 2: Are local networkng, technology transfer and R&D substtutes or complementary nputs to the nnovaton process? Queston 3: How mportant are frms R&D, networkng and technology transfer actvtes n determnng the level and success of frms nnovatve actvty? Queston 4: Do regonal and ndustry factors nfluence the effcency wth whch R&D, networkng and technology transfers are translated nto nnovaton outputs? The motvaton for consderng these questons les n ther potental mportance for the content and emphass of regonal development polces. How much resource, for example, should be allocated to local network development as opposed to encouragng R&D or technology transfer? Roper (1998), for example, suggests that currently less than 2 per cent of the ndustral development budget n Ireland s targeted at buldng collaboraton or co-operaton between frms. Smlarly, mprovements n what aspects of frms local operatng envronment are lkely to lead 4 Hedenrech and Krauss (1998), for example, dentfy what they regard as the key strengths of the Baden-Wurttemberg nnovaton system (.e. a dense network of nnovaton related nsttutons, a closely lnked supply chan, the educaton system, a progressve atttude to ndustral relatons, and an effectve bankng system) but are able to say lttle n terms of the relatve mportance of these varous nfluences on nnovaton potental. 3

5 to the greatest payoff n terms of the level of nnovatve actvty? How do these payoffs dffer between dfferent types of regonal economes? The remander of the paper s organsed as follows. Secton 2 descrbes our conceptual framework that s based on the dea of the nnovaton producton functon (Acs and Audretsch, 1988; Gerosk, 1990; Harrs and Tranor, 1995). Ths provdes the lnk between frms nvestments n technology transfer, networkng and R&D and nnovaton outputs. Frms regonal operatng envronment can then nfluence nnovaton outputs through ts effect on frms nvestments n technology transfer, networkng and R&D, and also through ts effect on the effcency wth whch these nvestments are translated nto nnovaton outputs (Love and Roper, 1999) 5. Secton 3 descrbes the man data sources used n the analyss and Secton 4 outlnes the emprcal results. The emprcal analyss covers the UK, Germany and the Republc of Ireland to take account of dfferences n both natonal and regonal contexts. 2. Conceptual Framework Attempts to model frms nnovaton actvty typcally start from standard behavoural assumptons that suggest that nvestment n technologcal development actvtes take place when the results of these nvestments (.e. nnovatons) are expected to earn postve post-nnovaton prce-cost margns. Moreover, the scale or ntensty of these nvestments s usually asserted to vary postvely wth the expected returns (see Gerosk, 1990, for a dscusson). In the sprt of the lterature on networks and learnng regons we envsage here that there are three possble routes by whch frms may obtan the knowledge necessary to undertake nnovaton: R&D, technology transfer and networkng. R&D we take to mean research or developmental actvty carred out by the enterprse alone, and on ts own premses. Networkng we regard as a collaboratve or sub-contract relatonshp between plants unrelated by ownershp (Roper et al., 1996). Technology transfer, we defne as collaboraton between plants wthn a group of companes. In the man, ths latter type of relatonshp wll be nter- 5 In the context of an examnaton of nnovaton n Northern Ireland, for example, Harrs and Tranor (1995) found that an urban locaton was assocated wth ncreased R&D nvestment but had no effect on the effcency wth whch these R&D nputs were transformed nto nnovaton outputs. 4

6 regonal although there may, of course, be multple plants from one group n a specfc regon. In each case, standard behavoural assumptons suggest that frms wllngness to nvest n these actvtes wll depend both on the expected return, frms ablty to approprate the profts of any development and the ease wth whch the development can be made or exploted. Ths suggests that frm s nvestments n R&D (R ), networkng (N ) and technology transfer (T ) may be represented by factor ntensty equatons of the form: R T N = γ = γ 21 = γ T R R N 33 N T π π e π e e L j L L j j L 26 k L L k k Z 27 Z Z + ε 1 + ε + ε 2 3 (1) Where π e s the expected level of post nnovaton returns of frm and L j, L k and Z are vectors of regonal, ndustry and plant-specfc factors whch regonal case-studes suggest may nfluence frms ablty to approprate or explot the benefts of any nnovaton (e.g. Braczyk et al., 1998). Plant-specfc factors mght nclude, for example, skll levels, natonalty of ownershp, group membershp etc. Followng Gerosk (1990) and others, the vector of plant-specfc characterstcs wll also nclude ndcators of frms market power desgned to capture any Schumpeteran plant sze effect. The regonal (L j ) and ndustry (L k ) vectors wll nclude aggregate parallels of smlar varables as well as nsttutonal or structural ndcators reflectng, for example, regonal nvestments n hgher educaton provson or publc sector R&D spendng. Assumng expectatons are ratonal, the factor ntensty equatons (1) can be estmated drectly by replacng π e wth actual post-nnovaton proft margns. In the estmatng equatons the sze and sgnfcance of parameter vectors γ 5, γ 6 and γ 7 wll ndcate the relatve mportance of regonal, ndustral and plant-specfc factors (see Research Queston 1 above). Parameters γ 2 and γ 3 wll suggest the type of 5

7 relatonshp whch exsts between frms R&D, networkng and technology transfer actvtes (see Research Queston 2 above) 6. Regonal and ndustral factors may also be mportant, however, n determnng the effcency wth whch knowledge acqured through R&D, technology transfer and networkng s translated nto nnovaton outputs (I ). These effects may ether be addtve or may depend on nteractons between frms actvtes and ther operatng envronment. Audretsch and Feldman (1996), for example, provde evdence that R&D spllovers are more mportant n areas wth concentratons of knowledge ntensve ndustres. Smlarly, the mportance of mleu-type effects, n whch the locatonal proxmty of frms may lead to the development of postve networks or nter-dependences, may depend on frm sze or ownershp (Mallat, 1991). The potental for such effects suggests a general form of nnovaton producton functon where the level of nnovaton outputs (I ) depends on the ntensty of frms R&D, networkng and technology transfer actvtes and other plant specfc and locatonal nfluences. That s for frm, located n regon j: I µ = ϕ 0 + ϕ1r + ϕ 2T + ϕ 3N + χ1i j + χ 2 I k + χ 3Z + θ1r I j + θ 2T I j + θ 3N I j + (2) where, followng Audretsch and Feldman (1996), we also allow for nteractons between the level of nnovatve actvty n the regon (I j ) and frms technologcal development actvtes. A central test of the hypothess that technology transfer and networkng ntensty are mportant to nnovaton wll be the emprcal sgnfcance of the ϕ 2 and ϕ 3 parameters (Research Queston 3 above). Smlarly, the estmated χ 1, χ 2, and χ 3 parameters wll ndcate the mportance of regonal, ndustral and frmspecfc characterstcs on the effcency wth whch plants nvestments n R&D, networkng and technology transfer are translated nto nnovaton outputs. Fnally, the θ 1, θ 2, and θ 3 parameters wll capture any smlar effects that depend on nteractons between frms technologcal development actvty and the level of nnovatve actvty n ther host regon (Research Queston 4 above). 6 Dutch evdence, reported n Audretsch et al. (1996), for example, suggests that the relatonshp between dfferent types of R&D may dffer between ndustres; n low technology ndustres, nternal and external R&D were substtutes; n hgh-technology ndustres they were complements. 6

8 3. Defntons and Data The man plant-level data source used n the estmaton s the Product Development Survey (PDS), whch contans comparable nformaton provded by 1700 UK plants, 1300 German plants and 500 Republc of Ireland busnesses (see Roper et al., 1996). In the UK, samplng for the PDS was structured by plant szeband, ndustry and super-regon (Fgure 1). In Germany, the orgnal samplng for the PDS was structured by plant szeband and ndustry for the former West and East Germany. Response rates were suffcent, however, to allow representatve results to be obtaned for each of the sx German super-regons dentfed by Lammers (1994). Plants respondng to the PDS provded nformaton on a number of ndcators of nnovaton outputs and success ncludng the total number of product changes and the proporton of sales derved from new and/or mproved products. The proporton of plants ntroducng new or mproved products over the perod was hghest n Germany, and n partcular n the former East Germany. Other German regons, partcularly Northern Germany and Baden-Wurttemberg also had hgh proportons of nnovatng plants (Table 1). The average number of new or mproved products and nnovaton ntensty (.e. the number of product changes per employee) also vared sgnfcantly between super-regons, although there was less regonal varaton n the average proporton of plants sales derved from new or mproved products (Table 1). Innovatng plants that responded to the PDS also provded nformaton on ther R&D actvtes, and whether they had engaged n ether networkng or technology transfer actvtes durng seven ndvdual elements of the product nnovaton process. In the estmaton, we use the standard ndcator of the ntensty of plants R&D actvty,.e. R&D employment as a percentage of total employment n the plant. Our measures of technology transfer and networkng ntensty are more expermental, and are based on the extent of plants external lnkages durng the dfferent elements of the nnovaton process. More specfcally, we calculate an ntensty score for technology transfer and networkng, rangng from zero for a plant whch engaged n no technology transfer or 7

9 networkng, to 100% for a plant that was nvolved n technology transfer or networkng durng all seven elements of the product nnovaton process (Table 1) 7. Three other sets of varables play an mportant role n the estmaton. Frst, we nclude a range of other plant-specfc varables derved from the PDS database whch other studes have suggested are mportant n nfluencng plants nnovaton capablty (see, for example, Klenknecht, 1996). Large plants, for example, seem more lkely to nnovate than ther smaller counterparts (Oakey et al., 1988; Santarell and Sterlacchn, 1990) 8. In addton, external ownershp has been shown to be a potentally mportant nfluence although the emprcal evdence s somewhat ambguous about whether external-ownershp boosts or constrans nnovatve actvty 9. Organsatonal flexblty has also been strongly lnked to the ablty to nnovate (e.g. Fartlough, 1994), and n the estmaton we nclude ndcators of workforce qualty, the nature of plants producton process and the organsaton of R&D, ntended to reflect the degree of nternal flexblty wthn the plant. The second set of varables ncluded n the estmaton s ntended to normalse for the possble effects on nnovaton of technologcal opportunty wthn the plant s sector and any concentraton effects related to plant and ndustry sze. PDS data, aggregated to 2- dgt ndustry level wthn each country, s used to generate varables to represent technologcal opportunty (e.g. the proporton of nnovatng frms, the proporton of frms wth R&D departments, sectoral R&D ntensty etc.). Concentraton ndcators are ntended prmarly to capture the potental effects of barrers to entry on nnovaton although the drecton of ths type of effect s uncertan. On one hand, barrers to entry may encourage nnovaton by provdng a secure, low-rsk market envronment n whch frms can undertake and beneft from nnovaton (Love et al, 7 The seven elements of the product development process dstngushed n the PDS were: dentfcaton of new or mproved products; prototype development; fnal product development; product testng; producton engneerng; market research; and marketng strategy. 8 Related evdence comes from a seres of studes by Acs and Audretsch who examned the relatonshp between sze and frms nnovaton ntensty (.e. the number of nnovatons made). Typcally, they observe ether a quadratc or cubc relatonshp wth nnovaton ntensty peakng at some ntermedate plant sze. In smlar work on Scottsh data, Love et al (1996) found maxmum nnovaton ntensty (nnovatons per employee) occurred at 1245 employees. 9 Arguments that external ownershp restrcts the autonomy of local plants lmtng ther wllngness and ablty to nnovate (Maleck, 1980; Howells, 1984; and, Harrs, 1991) have been contradcted by two recent studes, based on data for Northern Ireland and Scotland, whch concluded that externally- 8

10 1996); on the other hand, barrers to entry may act as a sheld, allowng frms to mantan ther level of proftablty even wth low levels of nnovaton. Data for the concentraton measures was taken from the UNIDO nternatonal database at 4-dgt level (3-dgt level for Ireland), and combned wth PDS data to estmate natonal market shares for ndvdual plants. The fnal set of varables ncluded n the estmaton s that ntended to capture any mleu or spll-over type effects offered by plants operatng envronment 10. For example, Shefer and Frenkel (1998) on the bass of evdence from Northern Israel, suggest that agglomeraton advantages, represented by populaton densty, are postvely assocated wth frms nnovaton actvty. Smlarly, Audretsch and Feldman (1996) suggest that local R&D spllovers are more sgnfcant where there are concentratons of hgh-tech or R&D ntensve ndustres. Other regonal varables ncluded n the analyss, mrrorng the arguments made earler at the plant level, were the percentage of plants undertakng R&D and the avalablty of graduates wthn the regonal labour force. Unemployment rates and relatve ncome levels were also ncluded as general ndcators of regonal prosperty and local market sophstcaton (Gudgn, 1995). 4. Emprcal Analyss Model estmaton was undertaken n two stages wth the estmaton of equatons for technology transfer, networkng and R&D ntensty precedng that of the nnovaton producton functon. As the technology transfer, networkng and R&D ntensty varables are lmted rather than contnuous, the approprate estmaton technque s Tobt, wth lower and upper tal censorng at zero and 100. Tables 2, 3 and 4 gve the estmated Tobt equatons for technology transfer, networkng and R&D ntensty for nnovatng frms n the UK, Germany and the owned establshments were more lkely to nnovate than ther locally-owned counterparts (Harrs and Tranor, 1995; Love et al, 1996). 10 Feldman (1994, p. 51) summarses the potental mportance of ths type of factor as follows: 9

11 Republc of Ireland 11. The frst notable features of the ntensty equatons are the strong and predomnantly negatve sgns on the smultaneous terms relatng to R&D, technology transfer and networkng actvty. These suggest that R&D, networkng and technology transfer are generally substtutes although there are natonal dfferences n the strength of ths substtuton effect. In the UK and Ireland, for example, there s strong evdence of substtuton between networkng and technology transfer and, controllng for the effects of external-ownershp and group membershp, between technology transfer and R&D (see also Love and Roper, 1999). In Germany, a smlar although statstcally weaker, substtute relatonshp exsts between plants networkng and technology transfer actvtes although there s lttle evdence of any clear relatonshp between plants R&D and other technologcal development actvtes. The dfferences n these results suggest that UK and Irsh companes are more wllng ceters parbus to substtute external sources of technologcal actvty (.e. technology transfer and networkng) for nternal actvty (.e. R&D) than German frms. One possble explanaton s that ths reflects dfferent manageral practces or atttudes. Roper (1997), for example, suggests that the nnovaton strateges of German small frms are more rsk averse and less market responsve than those of UK and Irsh small busnesses. Ths s reflected n German small frms more managed and strategc approach to product development, n contrast to the more responsve and often tme-crtcal developments undertaken by UK and Irsh small frms. Technology transfer or networkng actvty may allow a frm to accelerate the product development process but s lkely to carry addtonal rsks n terms of ntellectual property rghts and/or contract complance - a rsk/reward balance whch seems less lkely to appeal to more rsk averse German busnesses. An alternatve, and perhaps more persuasve explanaton for the greater wllngness or necessty for UK and Irsh frms to substtute external for nternal technologcal development actvty relates to Taken together, these complementary nsttutons provde resources and knowledge nputs to the nnovaton process, generate postve externaltes and spll-overs whch lower the cost of developng new nnovatons, and reduce the rsk assocated wth nnovaton. 11 Lmtatons n the PDS mean that the ntensty equatons for networkng and technology transfer can only be estmated for those plants undertakng some product nnovaton. R&D ntensty equatons can, however, be estmated for all plants. Comparson between the R&D ntensty equatons estmated for nnovators only and for all plants suggested no sgnfcant dfferences n the estmated coeffcents. The proftablty proxy proved statstcally nsgnfcant n all estmatons, and was dropped from the analyss. 10

12 skll shortages. The emprcal results ndcate a strong and consstent relatonshp between the level of graduate employment n plants n each country and R&D ntensty (Tables 2, 3 and 4). Roper et al. (1996), on the bass of evdence from the PDS, also noted that the nnovaton actvty of 32 per cent of UK and Irsh manufacturng frms was constraned by skll shortages compared to only 22 per cent of German frms 12. In ths stuaton, UK and Irsh frms may fnd t more necessary to substtute external technologcal development actvty for nternally conducted R&D, partcularly f they are operatng on a tght tmescale 13. R&D ntensty s also postvely related to plant sze n Germany, although there was no clear relatonshp n the UK, and some suggeston of a negatve relatonshp n Ireland. The apparent effect of external ownershp on R&D ntensty also dffered between countres although ths effect s condtonal on the level of R&D ntensty among ndgenously-owned companes 14. In the UK, external ownershp s assocated wth reduced R&D ntensty, wth no sgnfcant effect n Ireland 15. More surprsng, gven the hgh level of R&D ntensty among ndgenously-owned German frms, s that external ownershp n Germany s postvely assocated wth ncreased R&D ntensty. One possble explanaton, however, s that the strong German skll base has attracted the more research-ntensve actvtes of mult-natonal enterprses. External ownershp was also postvely assocated wth hgher levels of technology transfer actvty and, n the UK, wth hgher network ntensty. As ntuton suggests, beng a part of a group s generally postvely assocated wth technology transfer. More ntrgungly, n all three countres group membershp s also postvely assocated wth networkng, an actvty whch at frst sght mght be expected to be prncpally the preserve of SMEs. The wllngness of group member plants to network wth other companes may suggest that such plants have more 12 Dfferences n the proporton of plants reportng skll shortages were greatest for larger busnesses: 40 per cent of UK plants wth more than 500 employees reported sgnfcant skll shortages compared to only 14.0 per cent of smlar German plants. Source: Roper et al. (1996), Table 7.3, pp Further support for ths argument s provded by the reasons why frms undertake networkng and technology transfer. Roper et al. (1996) found that faster development was a much more mportant consderaton for UK and Irsh plants than ther German counterparts. 14 External-ownershp s defned as ownershp of the plant from outsde the respectve country. 15 Ths latter result suggests that the relatvely hgh levels of R&D spendng among externally-owned plants n Ireland, noted for example by Qunlan (1995), can be explaned largely or wholly by plants characterstcs regardless of ownershp. 11

13 experence and/or more developed systems for managng external relatons, and may have a broader range of functons over whch such networkng could occur. It may also suggest that plants whch are members of groups often owned from outsde the regon may be more embedded n ther host regon than would be suggested merely by examnng supply chan lnkages. Plants market share and average plant sze n each sector were also ncluded n the ntensty equatons to capture any scale or market concentraton effects. Both varables proved unmportant n the estmaton suggestng that n each country barrers to entry provde lttle effectve ncentve for plants to ncrease ther technology nvestments. Technologcal opportunty n the sector (represented by the sectoral (.e. 3-dgt) average level of the dependent varable) proved a much more mportant determnant of plants R&D, networkng and technology transfer ntenstes respectvely. The consstency and strength of ths effect for all three countres may suggest that ths s not merely the standard technologcal opportunty effect to be expected among sectors of dfferng technologcal ntensty, but may ndcate the presence of postve sectoral spllovers occurrng not merely n R&D (Audretsch and Feldman, 1996), but also n technology transfer and networkng. Incluson of regonal varables n the ntensty equatons was lmted to the UK and Germany (although see Roper (1998a) for a sub-regonal analyss for Ireland) 16. The frst notable feature of the estmated coeffcents on the regonal terms s that unlke Shefer and Frenkel (1998) and Brouwer and Klenknecht (1996), we fnd no evdence of any postve agglomeraton effect on the ntensty of frms technologcal development actvty. Secondly, there was lttle evdence of any sgnfcant lnk between the ntensty of plants technologcal development actvty and ndcators of regonal prosperty (.e. relatve GDP per capta, unemployment rates), publc sector R&D nvestment, or the avalablty of graduates n a regon. In terms of the man research questons outlned earler these results suggest that - n Germany and the UK at least plants nvestments n R&D, technology transfer and networkng are only very weakly condtoned by ther regonal operatng envronment. Much more mportant are the nature of the ndustry n whch the plant s operatng and the 16 In the emprcal analyss the level of regonal analyss was not the 12 super regons but 28 ndvdual regons.e. 11 UK standard regons, 16 German Länder, and Ireland 12

14 specfc characterstcs of the plant tself. These results also suggest that the contrasts between regonal levels of R&D, technology transfer and networkng actvty, noted n Table 1, are prmarly the result of technologcal opportunty n the markets n whch plants located n the regon are operatng and the more specfc characterstcs of the plants themselves. Agglomeraton effects and the type of factor-related effects antcpated n the regonal nnovaton system lterature seem, at least from our data, relatvely unmportant n nfluencng plants technology nvestments 17. Even f regonal nfluences are havng a relatvely small effect on the level of R&D, networkng and technology transfer ntensty, however, t s stll possble that they may be contrbutng to the effcency wth whch the nputs to the nnovaton process are combned to produce nnovaton outputs. Table 5 therefore gves estmates for the nnovaton producton functon (.e. equaton 2) where the dependent varables are nnovaton ntensty (.e. the number of product nnovatons per employee) and nnovaton success (.e. the proporton of sales attrbutable to new and mproved products). These output-based dependent varables are selected as mportant ndcators not merely of the of the extent of plant-level nnovaton, but of the commercal success of such actvty on whch much of the dynamsm of regonal economes depends. The model outlned earler relates frms factor nvestment decsons to fnal nnovaton outcomes and suggests two possble behavoural scenaros. Frst, frms decsons about factor ntenstes may be made takng nto account ther lkely mpact on nnovaton outcomes,.e. factor ntensty decsons may mplctly take account of the structure of the nnovaton producton functon. Alternatvely, because of lmted nformaton or the uncertanty mplct n the nnovaton process, factor nvestment decsons may be ndependent of later stages of the nnovaton process. In ths case, the decson process s essentally sequental wth the factor ntenstes beng predetermned wth respect to the nnovaton producton functon. The dstncton between these two alternatves s mportant n that t suggests dfferent modellng approaches. If decsons about factor ntenstes do take nto account the structure of 17 Ths clearly depends to some extent on the spatal scale of the regons ncluded n the analyss. For example, Harrs and Tranor (1995) do fnd some effect of ths type when they splt ther sample nto urban and rural areas. 13

15 the nnovaton producton functon, a smultaneous equaton approach s relevant. If nstead, the factor ntensty decsons are essentally separate from subsequent developments (.e. the level of nnovaton outputs s caused by, but does not cause, the factor ntenstes) the factor ntensty decson and the nnovaton producton functon can be modelled separately. To test the hypothess that our nnovaton output ndcators are exogenous wth respect to the factor ntensty nvestment decson we follow a procedure suggested by Smth and Blundell (1986). Ths nvolves smultaneously estmatng a Tobt model for factor ntensty and a regresson model for the nnovaton output ndcator and ncludng the nnovaton output ndcator as an ndependent varable n the Tobt model. Exogenety of the nnovaton output ndcator s then ndcated by a t-test of the hypothess that the error covarance s zero (see Greene, 1995, p. 610 for operatonal detals). Conductng these tests for each country and factor ntensty suggests that the nnovaton output ndcators are consstently exogenous wth respect to the factor ntensty equatons 18. In statstcal terms ths suggests that the factor ntensty decsons are ndependent of the nnovaton producton functon and that the factor ntensty equatons and nnovaton producton functon can be modelled separately. In more behavoural terms, ths result suggests that due ether to uncertanty or lmted foresght the factor ntensty nvestment decsons are made wth lttle regard or knowledge of subsequent nnovaton outcomes. R&D, networkng and technology transfer ntensty are ncluded n the equatons as explanatory varables both n natural terms and, reflectng the substtutablty noted earler, n the form of a seres of nteracton terms 19. In the estmaton, R&D ntensty proves mportant n terms of nnovaton ntensty n the UK and Germany but s of more lmted mportance n determnng nnovaton success. Technology transfer ntensty, on the other hand, s less mportant as a determnant of the extent of plants nnovaton actvty but does contrbute postvely to the subsequent commercal 18 Values of the t-statstc on the error covarance term for R&D ntensty, technology transfer ntensty and network ntensty respectvely were as follows: UK , , ; Germany 0.034, , 0.042; Ireland , 0.071, Terms were also ntally ncluded to reflect any nteracton effects between the level of nnovatve actvty n the regon and R&D, technology transfer and networkng ntensty. These proved wholly nsgnfcant n the estmaton. 14

16 success of plants nnovaton actvty. One possblty s that technology transfer allows plants to make hgher qualty nnovatons that then enjoy more commercal success. Alternatvely, technology transfer may allow frms to make nnovatons more quckly and obtan frst mover advantages n terms of ncreased sales and potental proftablty. More surprsng, partcularly gven the extensve dscusson n the regonal nnovaton system lterature, s the lack of sgnfcance of the estmated coeffcents on nter-frm networkng. Indeed, our results suggest that n none of the three countres does the ntensty of plants networkng actvty have any sgnfcant effect on ether the extent or success of plants nnovaton actvty (Table 5). Ths casts some doubt on the potental value for promotng nnovaton of network-based polcy ntatves of the type advocated by Morgan (1997), and exemplfed n hs descrpton of recent polcy developments n Wales (see also Ashem and Dunford, 1997). It also reflects the fndngs of Brouwer and Klenknecht (1996) who, on the bass of a Dutch data, found aganst all expectatons very lttle evdence that frms whch collaborate on R&D or acqure external technologcal knowledge have a hgher nnovaton output (p. 118). Brouwer and Klenknecht nterpret ther result as castng doubt on the value of the EU polcy focus on the promoton of R&D networkng and collaboraton, a contenton strengthened by the fndngs of the current study. However, they also suggest an alternatve nterpretaton due to Teece (1988), who argued that networkng may be undertaken by weaker nnovators n response to nternal resource constrants. If ths were the case, the postve albet nsgnfcant - coeffcents on the networkng ntensty varables n the nnovaton success equatons (Table 5) would suggest that networkng s of some value n compensatng for nternal resource defcences (Brouwer and Klenknecht, 1996, p. 118). Apart from R&D, networkng and technology transfer ntensty the nnovaton producton functons also suggest that a number of plant specfc characterstcs are mportant determnants of the extent and success of plants nnovaton actvty. As n prevous studes, our estmaton suggests that nnovaton ntensty vares nversely wth plant sze (Acs and Audretsch, 1988; Love et al, 1996; Love and Ashcroft, 1999), although there s weaker evdence of any relatonshp between plant sze and nnovaton success. External ownershp and group membershp both have a negatve effect on nnovaton ntensty; n Germany and Ireland, however, external ownershp s postvely assocated wth nnovaton success. The mplcaton s that although 15

17 plants whch are externally-owned or group members tend to make fewer nnovatons than ndgenously-owned frms these nnovatons are typcally more successful. As suggested earler ths may be due ether to the hgher qualty of nnovatons made by externally-owned plants or to the speed wth whch these are ntroduced to the market place. The possblty of a postve relatonshp between the qualty of a plant s nnovatons and market success s also suggested by the postve relatonshp between the presence of n-house R&D and hgher levels of graduate employment. As n the R&D, technology transfer and networkng equatons, the technologcal characterstcs of the sector represented n the nnovaton producton functons by sectoral nnovaton ntensty and success were also mportant. The mplcaton of ths s that the effcency wth whch nnovaton nputs were converted nto nnovaton outputs and nnovaton successes was strongly sectorally dependent. Unlke the ntensty equatons, however, regonal nnovaton actvty also proved mportant n some of the nnovaton producton functons. In Germany, plants located n regons where the average nnovaton ntensty and nnovaton success was hgher tended to be more nnovaton ntensve and successful themselves. In the UK, there was lttle evdence of any regonal effect on nnovaton ntensty, but frms operatng n regons where nnovators tend to be more successful tended also to be more successful themselves. 5. Summary and Conclusons The emprcal results dscussed above ndcate that there are some natonal dfferences both n the determnants of factor ntensty nputs nto the nnovaton process, and n the nnovaton producton functons themselves. Nevertheless, t s possble to broadly summarse the key fndngs wth respect to the four research questons posed earler. Queston 1: Plant and sectoral level factors are mportant n determnng the ntensty of R&D, technology transfer and networkng actvty at the plant level, but there s lttle evdence of regonal effects. 16

18 Queston 2: Where there s any sgnfcant effect, the relatonshp between R&D, technology transfer and networkng s one of substtuton rather than complementarty n the nnovaton process. Queston 3: R&D ntensty s mportant n terms of nnovaton ntensty n the UK and Germany but s of more lmted mportance n determnng nnovaton success. Technology transfer has lttle role as a determnant of the extent of plants nnovaton actvty but does contrbute postvely to the subsequent commercal success of plants nnovaton actvty. Networkng appears to have no postve role to play. Queston 4: There s some evdence of both sectoral and regonal nfluences on the effcency wth whch R&D, technology transfer and networkng are translated nto nnovaton outputs, and these effects vary between countres. Recent developments n the new economc geography, and the lterature on regonal nnovaton systems (e.g. Braczyk et al., 1998), have emphassed the potentally mportant role of networkng and the characterstcs of frms local operatng envronment n shapng ther nnovatve actvty. Modellng UK, German and Irsh plants nvestments n R&D, technology transfer and networkng, and ther effect on the extent and success of plants nnovaton actvtes, casts some doubt on the mportance of both of these relatonshps. In partcular, our analyss provdes no support for the contenton that frms or plants n the UK, Ireland or Germany wth more strongly developed external lnks (network or technology transfer) develop greater nnovaton ntensty. In terms of the lnk between plants external collaboraton and the extent of nnovatve actvty these results reflect the earler fndngs of Brouwer and Klenknecht (1996) usng Dutch data. Our results go beyond ther analyss, however, by suggestng that although nter-frm lnks also have no effect on the commercal success of plants nnovaton actvty, ntra-group lnks are mportant n terms of achevng commercal success. 20 In terms of the EU debate about regonal nnovaton polces these results have two man mplcatons. Frst, as Brouwer and Klenknecht (1996) argue, these results cast 20 Note, however, that there s some evdence from the UK that networkng may ncrease the number of nnovatons at plant level, ceters parbus (Love and Roper, 1999). 17

19 doubt on the valdty of the current emphass of EU technology polcy on promotng nternatonal collaboraton n R&D and nnovaton projects. Secondly, the results suggest the lmtatons of the type of network-based regonal regeneraton strateges whch have been advocated by Morgan (1997) and others (although see Hudson, 1997). Indeed, our results suggest that, because of the tendency for plants to substtute networkng actvty for n-house R&D, measures that promote networkng may have the perverse effect of reducng the extent and sales success of plants nnovaton actvty. It s possble, however, that encouragng networkng actvty may nfluence other dmensons of frms nnovaton actvty not consdered here. It may be, for example, that networkng allows frms to accelerate ther product development actvty reducng tme to market and ncreasng the rate of return. Alternatvely, networkng may allow an element of rsk-sharng allowng frms to undertake developments that would be too costly or rsky to undertake alone. Our emprcal analyss also hghlghts the dependency of the level of nnovatve actvty n a regon on the sectoral composton of manufacturng n the area and the characterstcs of the regon s plants. Where a regon s manufacturng sector s domnated by small frms, or sectors where lttle technologcal opportunty exsts, a low level of nnovaton actvty may be antcpated. External ownershp s assocated n our analyss wth relatvely low levels of effcency n translatng technologcal developments resultng from R&D, networkng and technology transfer nto nnovaton outputs: however, nnovaton by externally-owned plants s typcally more commercally successful than that by locally-owned busnesses. Other aspects of the operatng envronment wthn a regon were substantally less mportant n determnng plants level of R&D, networkng and technology transfer actvty. In partcular, our analyss provdes lttle support for the emprcal mportance for frms level of nvestment n R&D, networkng and technology transfer of agglomeraton effects, regonal prosperty, government R&D spendng or the avalablty of graduate labour. Stronger regonal effects were evdent on the effcency wth whch these nputs to the nnovaton process were translated nto nnovaton outputs. In Germany, plants located n regons wth hgh levels of nnovaton actvty and success tended themselves to both nnovate more and to be more successful. A smlar postve spllover effect was also evdent n terms of nnovaton success n the UK although we 18

20 could fnd no evdence of any such effect n terms of the extent of plants nnovatve actvty. These results provde cold comfort for more dsadvantaged regons. Structural weaknesses n such regons, n partcular a preponderance of low value added ndustry and low productvty small frms (e.g. Roper, 1998b), are lkely to lead to a low level of R&D, networkng and technology transfer actvty. Ths, n turn, s lkely to lead to a low level of nnovatve actvty lmtng plants growth potental and the ablty of the regon to generate postve spllovers n the nnovaton process. Structurally stronger regons, however, n whch R&D, networkng and technology are more wdespread seem lkely to generate hgher levels of nnovatve actvty and more substantal local spllovers. Our evdence also suggests, however, that attempts to mprove frms regonal operatng envronment through measures such as tranng and government R&D nvestment are lkely to have lttle effect on frms nvestments n R&D, networkng and technology transfer. More postve n the long term seem ntatves desgned to address the underlyng structural weaknesses of the regon through targeted nward-nvestment and by measures desgned to upgrade the nnovaton capablty of ndvdual plants. 19

21 Fgure 1: UK, German and Irsh Super-regons Used n the Product Development Survey Regons Scotland 2 Northern Ireland 3 Northern England 4 Mdlands and Wales 5 Southern England 6 Republc of Ireland 7 Northern Germany 8 Northern Rhne 9 German Md-West 10 Baden-Wurttemberg 11 Bavara 12 Eastern Germany

22 Table 1: Innovaton, Networkng and Technology Transfer Indcators by Super-Regon Product N Innovators Number of Product Changes Sales of New/Improved Products R&D Intensty Innovators Only Technology Transfer Intensty Networkng Intensty Innovaton Intensty % frms No. % % % % No per 100 Scotland Northern Ireland Northern England Mdlands and Wales Southern England Republc of Ireland Northern Germany Northern Rhne German Md West Baden-Wurttemberg Bavara Eastern Germany Memo Items: UK Ireland Germany

23 Notes 1. A product nnovator was a plant ntroducng a new or mproved product over the 1991 to 1994 perod. Sales of new or mproved products relates to the proporton of plants sales n 1995 derved from products whch were ether new or mproved over the 1991 to 1994 perod. R&D ntensty s R&D employment as a percentage of total employment n the plant. Technology transfer and networkng ntensty are defned n the text. Innovaton ntensty s the number of product changes (.e. the number of new or mproved products) per 100 employees. 2. Fgures relate to manufacturng plants wth more than 20 employees n Survey responses are weghted to gve regonally representatve results Source: Product Development Survey 22

24 Table 2: Tobt Models of R&D, Networkng and Technology Transfer Intensty n the UK R&D Innovators Only Technology Transfer Networkng Constant (0.592) (-1.797) Factor Intenstes Technology Transfer Intensty (%) (-5.244) Networkng Intensty (%) (-0.378) (-2.924) R&D Intensty (%) (-2.018) Plant Specfc Factors Externally-Owned Plant (-2.159) (4.502) Graduate Share of the Workforce (%) 0.32 (12.255) Natonal Market Share (%) (-0.805) Plant Sze (1993 Employment) (0.827) Plant Sze Squared (1993 Empl Squared) (-0.690) Plant Undertakng In-House R&D (-2.884) Plant s Part of Larger Group (8.309) (-0.711) (-5.000) (-1.277) (1.994) (1.307) (-1.184) (0.707) (1.675) Plant Has Formal R&D Department (1.771) Real Sales Growth (% pa) (-0.764) (0.789) Industry Factors Tech Transfer Intensty (mean %) (4.977) Networkng Intensty (mean %) (5.275) R&D Intensty (mean %) (5.047) Average Plant Sze (1993 Employment) (1.642) Regonal Factors Populaton Densty (per km 2 ) (-1.853) Relatve GDP Per Capta (EU =100) (-0.661) Govt R&D Employment (%) (0.845) Graduate Share of Workforce ( mean %) (-1.229) R&D Intensty (mean %) (0.932) (0.691) (0.656) (0.251) (-1.706) (-0.540) (0.946) (1.087) (-0.905) N Log Lkelhood

25 Notes and Sources 1. T-statstcs are gven n parentheses. Industry varables are sectoral averages of the factor ntenstes derved from the PDS at the 2-dgt level wthn each country. 2. Regonal varables relatng to R&D ntensty, the graduate share of the workforce, and the proporton of plants undertakng R&D and wth an R&D department are regonal averages of PDS data for NUTS 2 regons. Relatve GDP per capta n 1991 compared to the EU average (%) also at the NUTS 2 level was taken from Regonal Trends, 1994, Table 2.1. Populaton denstes, also at the level of the NUTS 2 regons were constructed from data from the same source. Government R&D s the share of the labour force engaged n government R&D n 1993 and s taken from R&D Annual Statstcs (Eurostat) 1996, Table 15b. Fgures are on a head-count bass. 24

26 Table 3: Tobt Models of R&D, Networkng and Technology Transfer Intensty n Germany R&D Innovators Only Technology Transfer Networkng Constant (-1.240) (-1.913) Factor Intenstes Technology Transfer Intensty (%) (-0.699) Networkng Intensty (%) (0.830) (-1.544) R&D Intensty (%) (0.187) Plant Specfc Factors Plant Sze (1993 Employment) (2.222) (0.885) Plant Sze Squared (1993 Empl. Squared) (-2.242) (-1.066) Externally-Owned Plant (1.745) (0.422) Graduate Share of the Workforce (%) (9.096) (2.416) Natonal Market Share (%) (-0.115) Plant Undertakng In-House R&D (-2.325) Plant Has Formal R&D Department (1.314) Real Sales Growth (% pa) (-0.859) Plant s Part of Larger Group (4.519) Industry Factors R&D Intensty (mean %) (6.591) (-0.660) Technology Transfer Intensty (mean %) (-2.528) (2.526) Networkng Intensty (mean %) (-1.233) (0.303) Average Plant Sze (1993 Employment) (-1.102) Regonal Factors Populaton Densty (per km 2 ) (-1.376) (1.037) Relatve GDP Per Capta (EU=100) (-0.615) (0.504) Govt R&D Employment (%) (0.913) (-1.529) Plants Wth In-House R&D (%) (0.757) Plants wth Formal R&D Departments (%) (-0.609) (0.888) R&D Intensty (mean %) (1.626) (-0.623) Graduate Share of the Workforce (mean %) (-0.096) (-2.947) (-3.454) (0.691) (2.909) (-2.353) (1.784) (-2.160) (1.784) (-0.830) (1.090) (1.990) (-2.268) (9.412) (-1.549) (2.033) (-1.933) (1.126) (1.648) N Log Lkelhood Notes and Sources: See Table 2. 25