Co-operative R&D: why and with whom? An integrated framework of ...
19 Pages
English
Downloading requires you to have access to the YouScribe library
Learn all about the services we offer

Co-operative R&D: why and with whom? An integrated framework of ...

-

Downloading requires you to have access to the YouScribe library
Learn all about the services we offer
19 Pages
English

Description

Research Policy, 2003. Co-operative R&D: why and with whom? 3. An integrated framework of analysis. 4. Luis Miottia, Frédérique Sachwalda,b,*. 5 a University ...

Subjects

Informations

Published by
Reads 34
Language English

Exrait

3 4 5 6 7 8
Research Policy, 2003
Co-operative R&D: why and with whom? An integrated framework of analysis Luis Miotti a , Frédérique Sachwald a , b , a University Paris 13 (CEPN), Paris, France b Institut Français des Relations Internationales (IFRI), 27 Rue de la Procession, 75740 Paris 15, France Received 21 November 2001 ; received in revised form 20 November 2002 ; accepted 26 November 2002
9 10 Abstract 11 Firms use R&D partnerships to access knowledge and build global R&D networks. This article develops an integrated 12 framework to examine the determinants of the choice of partners with which rms co-operate on R&D. This resource-based 13 perspective underscores the interactions between three major questions: why co-operate, who does and with whom? It argues 14 in particular that the choice of partners is dictated by the complementary resources which the latter command. The framework 15 is then expanded to predict the relative efciency of R&D co-operation with different partners, including suppliers, clients, 16 rivals, academic institutions and foreign rms. The empirical analysis, which is based on responses to France’s version of the 17 second European community innovation survey (CIS-2), strongly supports the overall framework of analysis. 18 © 2002 Published by Elsevier Science B.V. 19 Keywords: Co-operation; R&D partnerships; Innovation; Resource-based theory of the rm; European R&D consortia 20
21 1. Introduction cal partnerships. Since transactions involving the ex-35 change of knowledge are notoriously imperfect, they 36 22 In the context of the emerging knowledge-based tend to be embedded in various types of alliances. 37 23 global economy, both supply and demand for technol- More frequent and diverse knowledge exchanges have 38 24 ogy have been increasing at the world level. Since the thus constituted one major driving force behind the 39 25 late 1980s, the role of innovation as a factor of com- growing number of domestic and international tech-40 26 petitiveness and the accelerating pace of technologi- nological alliances since the 1980s. 41 27 cal progress have combined to make rms deepen and This article focuses on inter-rm co-operative 42 28 broaden their innovative capabilities. Firms have al- agreements as one of the major modes rms use to 43 29 located increasing resources to R&D to speed up the access knowledge and build global R&D networks. 44 30 pace of innovation and diversify their technological The relative exibility of co-operative agreements 45 31 capabilities. Firms have also designed new R&D prac- has been underscored as one of the main reasons for 46 32 tices, including both internal organisational changes their remarkable development since the 1980s ( Kogut, 47 33 and the building up of complex networks to deal with 1988; Ciborra, 1991; Teece, 1992; Gomes-Casseres, 48 34 growing outsourcing and various types of technologi-1996; Sachwald, 1998 ).Strategic and organisational 49 perspectives have further shown that the choice of 50 Corresponding author. co-operative R&D, rather than internal R&D, eq-51 E-mail address: sachwald@ifri.org (F. Sachwald). uity relationships or outsourcing, depends on the 52 1 0048-7333/02/$ – see front matter © 2002 Published by Elsevier Science B.V. 2 doi:10.1016/S0048-7333(02)00159-2
53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96
2 L. Miotti, F. Sachwald / Research Policy 1607 (2002) 1–19 characteristics of the technologies involved, as well ‘literature-based alliance counting’ data 2 and tests the 97 as on the characteristics of rms’ competencies. 1 set of hypotheses on the reasons for co-operation and 98 This article does not focus on the choice of R&D the prole of R&D partners. Section 4 then exam-99 partnerships among the diverse organisational and ines the relative efciency of R&D co-operation with 100 inter-organisational arrangements that rms use to the different types of partners. The conclusion sum-101 acquire technology, but rather on the choice of part- marises the results and discusses theoretical and pol-102 ners. It shows that the choice of R&D partners with icy implications. 103 a specic prole depends on the type of complemen-tary R&D resources rms seek to access, which, in turn, depends on their own prole. Partnerships have 2. The why–who framework of R & D 104 developed in particular between high-tech (HT) start co-operation 105 ups and larger incumbents, but also between suppli-ers and clients in various sectors, and more rarely A large proportion of the literature on technological 106 between competitors. At the same time, rms tend alliances has focused on the issue of the motivation 107 to enter various co-operations with universities and for co-operation. In view of the increasing complex-108 public institutes, including as part of large research ity and multi-disciplinarity of research, rms seek to 109 consortia. Furthermore, since national innovation access complementary resources from beyond their 110 systems tend to nurture specic creative activities boundaries. In this context, R&D partnerships have 111 and more generally reect national specialisation been analysed as organisational answers to the require-112 patterns, R&D networks may have an international ments of innovation-based competition and rapid tech-113 dimension. nological change, the ‘why’ issue being related to the 114 The contribution of this article to the analysis of forms taken by co-operation. Interactions between the 115 R&D co-operation is threefold. First, it develops an motivation for co-operation and the prole of partners 116 integrated framework which relates the set of ex- have been less systematically explored. This section 117 ternal R&D resources rms target to the choice of reviews the literature with the view to establish a set of 118 partners. This resource-based perspective underscores hypotheses relating the motivation for co-operation to 119 the interactions between three major questions: why the prole of partners. Some of these hypotheses have 120 co-operate, who does and with whom? Second, this already been discussed in the literature and the pur-121 framework is used to predict the relative efciency pose here is to build an integrated framework, which 122 of co-operation with different types of partners to relates the R&D resources rms seek to access— 123 innovate. Third, the empirical analysis is based on a why co-operate?—to their own prole—who co-124 French survey of rms’ innovation practices, which operates ?—and to the prole of their partners—with 125 provides a large sample of observations. Co-operation whom? 126 on R&D is dened as a rm’s behaviour rather than measured by a count of technological partnerships as 2.1. Why co-operate on R & D? 127 in databases on alliances. The paper is organised as follows. Section 2 dis- The literature has extensively discussed the moti-128 cusses the results from the literature on the interactions vations for entering into co-operative agreements as 129 between the motivation of alliances and the prole of organisational forms. The transaction cost perspective 130 partners. This discussion is used to build the integrated studies the circumstances under which co-operative 131 framework that relates the prole of co-operating rms agreements are the most efcient form of organization 132 with the prole of their partners, including suppliers, ( Stukey, 1983; Hennart, 1988, 1991 ; Robertson and 133 clients, rivals, public institutions and foreign partners. Gatignon, 1998). In depth studies of the attributes of 134 Section 3 describes the data, explains differences with the knowledge involved and of the characteristics of 135 the innovation process itself have further contributed 136 1 For different perspectives and types of empirical support, see ( Kogut and Zander, 1993; Nagarajan and Mitchell, 1998; Narula, 2 This term is used by Hagedoorn (2002) in his presentation of 2001 ). the MERIT-CATI database.
137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
L. Miotti, F. Sachwald / Research Policy 1607 (2002) 1–19
to the analysis of the role of internalisation in the evolution of rms’ capabilities. 3 Besides, the logic of transaction cost minimisation does not capture many of the strategic advantages of alliances, and strategic management perspectives are complemen-tary ( Foss, 1994; Gulati, 1998; Tidd and Trewhella, 1997; Sachwald, 1998; Hagedoorn et al., 2000 ). In the resource-based perspective, partnerships are driven by a logic of strategic resource needs and, this ap-proach is well suited to studying simultaneously the motivations of alliances and the characteristics of partners. The resource-based perspective suggests that rms conducting expensive, risky or complex research projects will seek R&D co-operation. In turn, these rms tend to be concentrated in high-tech sectors. Sectoral studies broadly support the idea that R&D is a major area of co-operation in high-tech or emerging industries. Incumbents may use alliances to enter new product areas or technological elds, as they allow them to expand their knowledge sources with limited investment exposure. Incumbents can thus test the importance of the new market or technology as well as evaluate strategic solutions ( Mitchell and Singh, 1983 ). Such behaviour has been well documented in the pharmaceutical industry, where incumbents have extensively resorted to alliances in order to ex-pand their knowledge base in biotechnology ( Pisano et al., 1988; Arora and Gambardella, 1990; Powell and Brantley, 1992; Sharp et al., 1994 ). Conversely, entry by new biotechnology rms is eased by vertical alliances with pharmaceutical, chemical or marketing rms, which possess complementary assets ( Shan et al., 1994; Calabrese et al., 2000 ). In the emerg-ing multimedia elds, where speed to market and innovative product combinations constitute major competitive strengths, rms actively knit networks of complementary assets ( Gomes-Casseres, 1996; Quélin, 1996 ). In the MERIT-CATI database, the proportion of “R&D partnerships” 4 in pharmaceuticals and informa-3 Kogut and Zander (1993) studies the role of complexity and tacitness; Nagarajan and Mitchell (1998) and Narula (2001) studied the impact of the extent to which new knowledge is related to core R&D resources or to more peripheral assets. 4 Which includes joint ventures and other inter-rm agreements that “contain some arrangements for transferring technology or joint research” ( Hagedoorn, 2002 , p. 491).
3
tion technologies has increased from 40 to 80% of the 178 total between 1980 and 1998. 5 The positive inuence 179 of R&D intensity on the propensity to co-operate has 180 been recently conrmed for large cross-sectoral sam-181 ples of German and Spanish rms ( Fritsch and Lukas, 182 2001; Bayona et al., 2001 ). The rst hypothesis be-183 low is thus included to start building our integrated 184 framework, in which the quantity and quality of R&D 185 resources inuence both the propensity to co-operate 186 and the propensity to co-operate with specic partners. 187 Hypothesis 1. The propensity to co-operate on R&D 188 is higher for rms from sectors with relatively high 189 R&D intensity. 190 191 The strategic need for high R&D efforts may also 192 explain intra-sectoral co-operative patterns. During the 193 1980s, electronic products at the earliest stages of the 194 life cycle exhibited a higher number of R&D alliances 195 (Cairnaca et al., 1992). Similarly, among a sample of 196 new American semiconductor rms, the most inno-197 vative ones and those faced with the fastest pace of 198 technological change exhibit a higher propensity to 199 co-operate on product development ( Eisenhardt and 200 Schoonhoven, 1996 ). 201 Hypothesis 2. The propensity to co-operate on R&D 202 is higher for rms that draw the most on scientic 203 resources to innovate, as opposed to rms further away 204 from the technological frontier. 205 206 The literature on innovation and technology trans-207 fer has established that access is not sufcient to learn 208 from external knowledge sources, adequate absorp-209 tion capacity being a necessary complement ( Cohen 210 and Levinthal, 1989 ). Absorption capabilities depend 211 on specic investment, including in particular the ex-212 istence of an R&D department and enough qualied 213 personnel. Internal R&D capabilities have thus a com-214 plex inuence on the propensity to co-operate. On 215 the one hand, co-operation may become necessary be-216 cause internal resources are insufcient to meet the 217 rms’ strategic objectives. On the other hand, the ex-218 istence of adequate absorption capabilities increase 219 5 The share of R&D partnerships in aerospace and defense has rather decreased since the 1970s, which may be related to the shrinking of defense activities in the 1990s, while IT activities have expanded.
220 221 222 223 224 225 226 2 2 87 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257
4
L. Miotti, F. Sachwald / Research Policy 1607 (2002) 1–19
the returns rms can expect from access to external resources. This second effect has been found to be stronger in biotechnology ( Arora and Gambardella, 1990 ) and in a survey of UK rms ( Lowe and Taylor, 1998 ). Hypothesis 3. The propensity to co-operate on R&D is higher for rms with stronger absorptive capabi-lities. 2.2. And with whom? This section relates the above set of hypotheses on the prole of rms that choose to co-operate on R&D to complementary hypotheses on the prole of their partners. 2.2.1. Partners with the right complementary resources The resource-based perspective considers that the necessity for complementary resources is a key driver of inter-organisational co-operation. 6 It suggests that the adequate partners should possess the resources which the rm is seeking. The latter may be classied into two broad categories, depending on the needs of the partners. If partners aim at reducing costs and risks through economies of scale and rationalised in-novation processes, 7 they will pool similar resources to the alliance. If partners aim instead at managing technological convergence such as in the multimedia nexus or inter-dependence among innovation pro-cesses, they will combine complementary resources. This distinction is crucial for choosing the right partners. Suppliers and clients play an important part in the innovation process as they can contribute crucial in-formation on technologies, users’ needs and markets. Hence, innovation requires vertical interactions and communication ows. The latter may be more im-portant in some sectors and may be organised in dif-ferent ways, but the general need is quite pervasive. 6 This perspective is widely adopted, explicitly or implic-itly, by the management literature ( Roberts and Berry, 1985; Kogut and Chang, 1991; Gomes-Casseres, 1996; Eisenhardt and Schoonhoven, 1996; Doz and Hamel, 1998; Mowery et al., 1998 ). 7 This rationale for R&D co-operation has been explored from different perspectives, including industrial organization models ( Katz, 1986; Jacquemin et al., 1985 ).
Vertical R&D co-operation is thus hypothesised to 258 be an integral part of the innovation process, espe-259 cially so now that rms tend to focus on a smaller 260 set of businesses Bresnahan (1999) emphasises this 261 feature in the case of the computer industry by forg-262 ing the notion of ‘co-invention’ involving buyers and 263 sellers. 264 Rivals may nevertheless possess complementary 265 R&D resources. They may also be attractive partners 266 to team up with in order to reduce costs and risks 267 for large projects. They are however potentially dan-268 gerous because they sell on similar markets and may 269 access the rm’s own R&D resources through collab-270 oration. The industrial organization literature has de-271 veloped models to analyse both the incentives and the 272 risks of and R&D co-operation. They draw attention 273 to the risks involved in co-operation, related to in-274 voluntary ‘outgoing spillovers’ to partners ( Cassiman 275 and Veugelers, 1998 ). Such considerations suggest 276 that co-operation between competitors is particularly 277 risky and should be limited to two types of cases: 278 rst, when a particularly strong common interest has 279 been identied and, second, when the co-operation 280 concerns far-from-market research leading to generic 281 results. 282 The tension between the resource considerations, 283 which constitute an incentive to co-operate, and the 284 risks involved, which may inhibit co-operation, is 285 stronger in the case of alliances with rivals since risks 286 are lower with suppliers and clients. In high-tech 287 sectors, rms may nevertheless co-operate with ri-288 vals as they feel strong incentives to pool R&D 289 resources and/or integrate networks in order to estab-290 lish standards. The literature has amply documented 291 such cases ( Mariti and Smiley, 1983; Garrette and 292 Dussauge, 1995 ), which may give the impression that 293 co-operation with rivals is frequent. 294 Hypothesis 4a. Vertical R&D co-operation is more 295 frequent than horizontal co-operation with rival rms. 2 9 76 Hypothesis 4b. Horizontal co-operation with rival 298 rms is more frequent in high-tech sectors. 299 300 Co-operation with public partners does not involve 301 commercial risk. Public research institutions do not 302 seek commercial applications and tend to focus on the 303 most generic or basic end of the R&D complex. Con-304
305 306 307 308 309 310 311 312 313 3 1 54 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348
L. Miotti, F. Sachwald / Research Policy 1607 (2002) 1–19 5 sortia involving a large number of rms, including ri- medical equipment ( Hobday, 1994; Sharp et al., 1994; 349 vals, tend to focus on this type of research and have of-Veugelers, 1995; Mouline, 1999; Sachwald, 2000 ). 350 ten been supported by public funds ( Sachwald, 1990 ; Sakakibara, 1997, 2001a,b ; Branscomb and Keller, Hypothesis 6a. Co-operation with American partners 351 1999 ). More generally, when co-operative research is is more frequent in sectors where the US has a com-352 supported by public funding, it is designed in order to parative advantage, especially in high-tech. 3 5 34 maximize disclosure and spillovers. 8 Hypothesis 6b. Firms conducting research at the 355 Hypothesis 5. Co-operation with public research in- technological frontier co-operate more with American 356 stitutions is most attractive to rms that conduct R&D partners. 357 at the technological frontier. 358 National innovation systems and technological spe-359 2.2.2. Why choose foreign partners? cialisation are closer between European countries than 360 The dynamic global competitive environment and between European countries and the US, co-operation 361 efforts by rms to expand and reorganize their inno- R&D. As a result, intra-European R&D co-operation 362 vative capabilities have led to reconsidering a num- will typically not aim at pooling complementary re-363 ber of results from the economic literature on multi- sources. It may however be used to pool similar re-364 nationals and on national innovation systems. Evolu- sources in order to reduce costs. 365 tionary and resource-based perspectives emphasize the stickiness of innovative capabilities. The latter evolve Hypothesis 6c. R&D co-operation of French rms 366 along specic trajectories, which depend on both ge- with European partners aims at sharing the costs of 367 ography and history as rms’ capabilities are em- innovation. 36 89 bedded in national systems of innovation. Yet, in the context of globalization, as rms strive to access ex-ternal resources through webs of technology trans-fer and learning, they act upon national trajectories. Likewise, as internationalization provides access to foreign systems, innovative activities become some-what less dependent on the innovation system of home countries. International co-operative ventures can provide rms with access to country-specic advantages em-bedded in their partners and R&D co-operation can be viewed as a vehicle for tapping into the compara-tive advantages of foreign countries. The nationality of R&D partners should thus depend on the relative technological strength of their country in the relevant elds. From this perspective, there is a broad distinc-tion between European rms and American rms as the US tends to be closer to the technological fron-tier in a number of high-tech sectors. Studies based on diverse data sources indeed show that European rms tend to choose American partners in sectors where the US has developed the strongest technolog-ical advantages, such as biotechnology, electronics or 8 Which corresponds to the suggestions of theoretical models on the role of disclosure within alliances ( Katz, 1986 ).
3. Testing the determinants of choice 370 of partners 371 This section describes the data and presents the 372 empirical test of the why–who framework developed 373 above. 374 The empirical work is based on the French CIS-2 375 survey conducted in 1997 by the SESSI (Ministry of 376 Industry) and covering manufacturing rms located 377 in France. Questions related to innovation practices 378 over the period 1994–1996. Albeit non-compulsory, 379 the survey features an outstanding response rate of 380 85%. The sample of 4215 rms gives a reliable image 381 of the behaviour of the manufacturing rms with more 382 than 10 employees. 9 The survey includes a question 383 on whether rms have co-operated in order to inno-384 vate, meaning active participation in joint R&D and 385 projects (contracting out is thus excluded). Firms that 386 co-operate in R&D are innovative, i.e. they have stated 387 that they innovated over the 1994–1996 period (prod-388 9 The sample weighted with the expansion coefcient repre-sents 20,997 rms; rms with more than 500 employees are all included.
389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426
6
L. Miotti, F. Sachwald / Research Policy 1607 (2002) 1–19
uct, process, patent). The innovative sample includes 2378 rms (9832 when weighted). 3.1. Scope of R & D co-operation in France According to Table 1 , a third of the rms choose to co-operate in R&D. This is substantial but also means that only a minority of rms enter into R&D co-operation as part of their innovative process. The proportion is substantially higher for rms from high-tech sectors (53%) 10 and for the largest rms (67%). Firms from inter-related groups of rms also tend to have a relatively high propensity to co-operate (49%), but their major partners are other rms from the group. This means that rms consider that co-operation within groups involving subsidiaries does indeed constitute co-operation. From the point of view of inter-rm co-operation, however, intra-group co-operation has to be considered as a specic case. Competitive risks are a priori much lower, which may for example be an incentive to co-operate for relatively small rms from a group. The specic con-tribution of intra-group R&D co-operation may also be different in nature. The propensity to co-operate with competitors is particularly low, which conrms that rms tend to avoid R&D co-operation with rivals ( Hypothesis 4a ). 11 Conversely, Table 1 conrms the more im-portant role played by clients and suppliers in the innovative process. Co-operation with academic or-ganisations, which is substantial, is markedly more intense for the largest rms and for patenting rms. Table 2 indicates that rms co-operate rst with French partners. The domestic scope of the major-ity of technological partnerships has been under-scored mainly in the case of the US ( EU, 1997; Hagedoorn, 2002 ), but may actually be a quite general phenomenon. The reason for this different observa-tion may be due to the type of data used in differ-ent studies. Databases on technological partnerships 10 Kleinknecht and Reijnen (1992) reported a different result but used a broad classication to distinguish sectors and did not really isolate high-tech industries (their Table 3 includes such aggregates as chemicals and plastics and does not isolate electronics for example). 11 For similar observations see ( EU, 1997 ) and sectoral studies. According to data not restricted to R&D alliances, co-operation between rivals is relatively more important ( Veugelers, 1995 ).
largely rely on public sources to identify alliances 427 and thus tend to be more exhaustive on operations 428 from large rms, and probably too on rms from 429 the largest countries. National surveys, such as the 430 one used here provide a better coverage of smaller 431 rms, which tend to have a smaller geographical 432 reach. As indicated in Table 2 , large rms and rms 433 in high-tech sectors are more likely to choose foreign 434 partners. 435 The table also indicates that among foreign part-436 ners, French rms tend to rst choose EU partners. 437 As a result, intra-European alliances are substan-438 tially more frequent than transatlantic alliances— 439 more than twice as much for all rms. The share of 440 intra-European technological partnerships is on the 441 contrary lower than the share of transatlantic ones 442 in the MERIT-CATI database. 12 The source of this 443 difference may also be the more systematic coverage 444 of smaller rms in the CIS survey. Large rms and 445 high-tech rms exhibit a relatively higher propensity 446 to team with American partners. 447 Fig. 1 further describes the data set. It shows in 448 particular, that R&D co-operation is relatively intense 449 in mid-high-tech (MHT) sectors, such as chemicals 450 and automobiles. It means that R&D co-operation in 451 France is not strongly concentrated in high-tech sec-452 tors, which actually reects the structure of the French 453 industrial activities in France. 454 3.2. Test design 455 The different tests below are built on a similar de-456 sign in order to test the integrated framework devel-457 oped above. Dependent variables are dummy vari-458 ables, which are equal to 1 when a rm co-operates 459 on R&D with certain types of partners. The same set 460 of independent variables is used to successively test 461 the different hypotheses—with the exception of one 462 variable in the test of the geographical origin of part-463 ners. This design results in a set of logit specications, 464 which allows a clear interpretation of the inuence of 465 the different independent variables. 13 The regression 466 coefcients estimate the impact of the independent 467 12 In the 1980s and 1990s, but not in the 1960s and 1970s according to Hagedoorn (2002) . 13 A multi-nomial specication has been run, but interpretation was difcult and this design is more satisfactory to test successively the whole set of hypotheses and reect on the framework.
L. Miotti, F. Sachwald / Research Policy 1607 (2002) 1–19
7
Table 1 Propensity to co-operate on R&D among innovating rms by type of partner (%) Partners Firms’ characteristics All rms a Groups >500 In high-tech sector b Patent in 1994–1996 New product in 1994–1996 All types of partners 33.6 49.0 66.8 52.7 47.4 40.2 Within groups c 16.9 33.6 50.1 31.6 24.8 21.8 Clients 15.0 20.9 27.9 23.5 21.8 20.9 Suppliers of components 11.6 17.3 33.5 22.2 18.9 16.8 Suppliers of equipment 8.8 13.2 28.9 16.6 14.6 11.9 Competitors 4.3 7.3 15.4 12.8 7.1 5.5 Universities, institutions 13.3 19.8 37.6 22.1 25.8 16.2 Number of rms a 9832 4766 717 608 3044 4377 a Innovative rms of the survey with more than 10 employees; weighted numbers (unweighted total is 2378). The propensity to co-operate is the ratio of the number of rms that co-operate over the total number of rms. b OECD classication. c Inter-related groups of rms including subsidiaries (50% threshold).
468 variables on the probability that the rm will conduct way for other types of partners: clients or customers, 485 469 co-operative R&D, either in general or with specic public institutions. 486 470 partners. In the case of foreign partners, the sample is reduced 487 to rms co-operating with American and/or European 488 471 3.2.1. Dependent variables rms in order to specically identify the determinants 489 472 The dependent variable in the rst test is a dummy of the choice of an American or of a European partner. 490 473 variable which is equal to 1 when the rm co-operates In the test of the determinants of R&D co-operation 491 474 to innovate, whatever its partners. It thus provides with US partners for example, the dependent vari-492 475 a general perspective on the determinants of R&D able is a dummy which is equal to 1 when rms 493 476 co-operation and a sort of baseline for the other co-operate with American partners and 0 otherwise. 494 477 tests, which address the hypotheses on the choice The sample is further restricted to French rms, as for-495 478 of partners. They are conducted on the sample of eign subsidiaries tend to co-operate with their parent 496 479 rms which co-operate and each singles out one company. 497 480 type of co-operation. To test for the determinants of 481 co-operation with rivals for example, the dependent 3.2.2. Independent variables 498 482 variable is a dummy variable which is equal to 1 if Four sets of independent variables are included as 499 483 the rm co-operates with rival partners, and 0 other- determinants of the propensity to co-operate in R&D; 500 484 wise. The dependent variable is designed in the same they relate to sectoral characteristics, rms’ character-
Table 2 Geographical distribution of partners, in percentage of the rms that co-operate on R&D a Nationality of partner Firms’ characteristics All rms b >500 In high-tech sector c Patent in 1994–1996 New product in 1994–1996 French 83.3 88.4 87.0 85.6 83.0 European 44.3 69.4 57.4 51.8 50.4 American 19.9 45.3 36.4 29.3 25.3 Japanese 6.5 19.1 12.1 10.7 8.7 Others 9.2 14.5 16.5 10.5 11.4 a As indicated in Table 1 , these represent 33% of the total of innovative rms. b Innovative rms of the survey with more than 10 employees; weighted numbers (unweighted total is 2378). c OECD classication.
8
L. Miotti, F. Sachwald / Research Policy 1607 (2002) 1–19
Fig. 1. Propensity to co-operate on R&D among innovating rms, by sector. Note: as in Table 1 , the propensity to co-operate is the ratio of the number of co-operating rms to the total number of rms. The average propensity in the total sample is 33.6% ( Table 1 ).
501 istics, obstacles to innovation and public funding. The introduced in the equation of the propensity to co-519 502 focus of the discussion below is on R&D variables, operate with American rms, and a positive impact is 520 503 which are central to our hypotheses. expected. 521 504 522 505 3.2.2.1. Sectoral variables. In order to test 3.2.2.2. Firms’ characteristics. Most statistical stu-523 506 Hypothesis 1 , sectoral R&D intensity variables are dies show that the propensity to co-operate in R&D 524 507 introduced. Dummy variables are included to indi- is positively related to the size of the rm ( Bayona 525 508 cate whether the sector to which the rm belongs et al., 2001; Fritsch and Lukas, 2001 ). 14 Veugelers 526 509 is high-tech, mid-high-tech, mid-low-tech (MLT) or (1997) goes deeper into this issue by exploring the 527 510 low-tech (LT), using OECD classication. Fig. 1 interactions between internally nanced R&D ex-528 511 above suggests to expect a positive inuence of both penditure and co-operation. She nds that big R&D 529 512 HT and MHT on the propensity to co-operate. 513 The discussion above argues that the propensity to 14 Kleinknecht and Reijnen (1992) found no inuence of rms’ 514 co-operate with partners of a given country should de-size on the propensity of Dutch rms to co-operate in R&D (except e advanta . H for co-operation with research institutes). This surprising result 551165 ipsenbdasoendtohneirthceoimdpeaetitthivatrmscgoempe y ti p ti o v t e he a si d s va 6 n a -may be due to the fact that the estimate of the probability to 517 tage are related to national comparative advantage. A co-operate included other independent variables positively related to size, in particular the propensity to export and the existence of 518 measure of the US comparative advantage (USCA) is an R&D laboratory in the rm.
530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 56 65
L. Miotti, F. Sachwald / Research Policy 1607 (2002) 1–19
spenders have a signicantly higher probability of co-operating, but that smaller innovative rms are more likely to co-operate than larger ones. In other words, the most relevant variable would not be the size of the rm but the research orientation of rms 1 . 5 Furthermore, the existence of a permanent R&D struc-ture within rms positively inuences their propen-sity to co-operate ( Kleinknecht and Reijnen, 1992; Veugelers, 1997 ; Colombo and Garone, 1998; Bayona et al., 2001 ). Having no data on the amount of R&D spend-ing, we include both the size of rms (log of the number of employees) and an indicator of their ab-sorption capacity (dummy variable for internal R&D) as independent variables. They should both have a positive inuence on the propensity to co-operate ( Hypotheses 1 and 3 ). “Science”, is introduced to test Hypothesis 2 ; its value varies between 0 and 9 as the rm draws more heavily on external sources close to scientic research, including patents, uni-versities and research institutes. Science should pos-itively inuence the probability to co-operate on R&D. The market share of each rm is also included in log form as it inuences incentives to innovate and may thus increase the propensity to co-operate. 16 Fi-nally, since descriptive statistics in Table 1 under-score the extent of intra-group R&D co-operation, we incorporate group as a control variable. It is a dummy that is equal to 1 when the rm belongs to a group. Group should inuence positively the propen-sity to co-operate. Besides other reasons, both the size of the rm and its integration into a group may have a positive inuence on co-operation as they indicate access to a substantial pool of resources which are complementary to R&D. 17 15 Interestingly in this perspective, Arora and Gambardella (1990) found no statistically signicant inuence of the size of biotech-nology rms on their propensity to co-operate with rms or uni-versities. On the contrary, the number of patents had a positive inuence on the number of partnerships. 16 The role of R&D co-operation in oligopolistic sectors has been emphasised ( Delapierre and Mytelka, 1998 ; Sakakibara, 2001a,b ). 17 Lowe and Taylor (1998) suggest that the positive inuence of size on the inward licensing could be due to the fact that it is a proxy for complementary assets that are necessary to benet from licensing.
9
3.2.2.3. Obstacles to innovation. R&D co-operation 567 is motivated by not only the need to draw on comple-568 mentary external resources but also by the risks and 569 costs of innovation. It may help in overcoming a num-570 ber of specic obstacles to innovation. Co-operative 571 behaviour may thus be positively related to a number 572 of obstacles to innovation. The following ones are in-573 cluded with dummy variables: cost of innovation, high 574 risks, and lack of market information. Variables are 575 constructed with rms’ answers in the CIS survey on 576 questions about obstacles to innovation. 577 578 3.2.2.4. Public funding. Public funding tends to 579 have a positive inuence on rms’ R&D spend-580 ing. Veugelers (1997) considers that public funding 581 thus has an indirect inuence on the propensity to 582 co-operate in R&D. The European innovation pol-583 icy as well as national schemes sponsor co-operative 584 R&D projects, which may constitute a further in-585 centive to co-operate. The survey questionnaire does 586 not allow distinguishing this specic source of R&D 587 funding. As a consequence, equations include a pub-588 lic funding dummy, which is equal to 1 when the rm 589 benets from R&D subsidies. 590 3.3. Results 591 The equations have a good explanatory power and 592 strongly support the set of hypotheses which build our 593 integrated framework of interpretation of co-operative 594 R&D. 595 3.3.1. Propensity to co-operate on R & D 596 Table 3 presents the results of the rst test on the de-597 terminants of R&D co-operation, which may be con-598 sidered as a reference points for the other tests on 599 co-operation with specic partners. 600 The hypotheses founded on the resource-based the-601 ory of the rm and the need to access complemen-602 tary R&D resources are strongly supported. Firms 603 from high-tech, but also from mid-high-tech sectors 604 tend to co-operate more than rms in less R&D in-605 tensive sectors ( Hypothesis 1 ). Moreover, rms which 606 conduct R&D close to the technological frontier also 607 exhibit a higher propensity to co-operate on R&D 608 ( Hypothesis 2 ). The positive interactions between in-609 ternal R&D capabilities and co-operation with external 610 partners are also conrmed ( Hypothesis 3 ): The exis-611
10 L. Miotti, F. Sachwald / Research Policy 1607 (2002) 1–19 Table 3 rms which consider that the lack of market infor-636 Determinants of R&D co-operation mation constitutes an obstacle to innovation and they 637 Variable name Coefcient P > χ 2 might resort to co-operation with clients in particu-638 Constant 6.8836 0.0001 lar to alleviate these problems. Firms that co-operate 639 Size 0.8103 0.0001 with rivals do not face similar obstacles. Rather, 640 Group 0.5461 0.0001 they cite R&D costs as an obstacle to innovation. 641 PPueblicanFeuntndRin&gD00..2490251900..00006011 This tends to conrm that rivals team up in order 642 rm-tech0.27390 to exploit economies of scale and reduce individual 643 HMiigdh-high-tech0.15510..00736910 costs of innovation in high-tech sectors, as argued 644 Science 0.1526 0.0001 above. 645 Market share 0.1069 0.0174 According to Table 4 , permanent R&D does not 646 LHaicgkhofsttechnologicalinformation00..0133410500..17402781 signicantly inuence the relative propensity to 647 Highcriosk 0.0449 0.6146 co-operate with private partners. On the contrary, it 648 Lack of market information 0.0389 0.6424 strongly inuences the propensity to co-operate with 649 McFadden R 2 27.73 – public institutions. Firms which co-operate with pub-650 log Likelihood 2424.88 – lic institutions are not concentrated in R&D intensive 651 Probability (LR stat) 0.000 – sectors. They tend, however, to draw on close to 652 Sample: 2378 rms (weighted: 9832). science resources to innovate ( Hypothesis 5 ). Con-653 sequently, they exhibit different features from rms 654 612 tence of an internal laboratory in a rm substantially that co-operate with rivals, which are concentrated 655 613 increases its probability to co-operate on R&D. Over- in high-tech sectors but do not focus specically on 656 661154 ealnlt,atthioenseotfofrvmarsiasublbesstainntdiiacllatiinngcraesatsreonthgeriersperaorpcehnosriit-yfrontierR&Dintheirrespectivesectors.Thismay 657 616 toco-operateInterestingly,vayriablesrelatedtothevar-berelatedtoanothercontrastingfeaturebetween 658 the determinants of the two types of partnerships: 659 617 ious obstacles to innovate, including costs and risks, rms co-operating with public institutions do not 660 618 do not inuence the propensity to co-operate. encounter cost obstacles to innovation (the cost vari-661 619 icalThweotrekstdiaslscouscsoendarbmovser,essuulcthsfarsotmheprpeovsiitoiuvseeinmpir-ablehasasignicantlynegativeimpact),whileitis 662 662201 encfizeandpublicfunding.Ahighmarketsharue-thecasewithrmswhichpartnerwithrivals.On 663 622 alsoesotimsulatesco-operationtopofthesizeeffect.thecontrary,rmsthatco-operatewithpublicinsti-664 n o tutions consider that insufcient market information 665 623 So does belonging to a group, as suggested by Table 1 . constitutes an obstacle to innovation. This may be 666 because their R&D activities aim at more radical in-667 624 3.3.2. Choice of partners novation for which markets are still uncertain. These 668 625 In order to examine the more specic determinants rms may not devote much resources to marketing 669 626 of co-operation with each type of partner, the sam- either. 670 662287 cplleeairslynsohworwessttrhiacttecdo-tooperratmiosnthwaitthcod-ioffperatte.t Tab e l s e o 4 fOverall,theseempiricaltestsstronglysupportour 671 eren yp framework. In particular, they show that more re-672 629 partners is driven by quite different factors. search oriented rms are more likely to co-operate 673 663310 rarCeo-(o H p y e p ra o t t i h o e n sis w 4 it a h),irsivaslus,bstawnhtiiaclhlyismorreelaltiikveellyyonR&D.Ourresultsfurthersuggestthatconducting 674 research close to science and exhibiting high R&D 675 632 in high-tech sectors and, to a lesser extent, in intensity are two different ways to be research ori-676 633 mid-high-tech sectors ( Hypothesis 4b ). Conversely, ented: each requires specic types of resources, which 677 6634 lvoewrt-itceaclhcos-eocptoerrsa.t 1 i 8 onViesrtrieclaaltivceol-yopmoarteiofnreqinuveonltveinsexplainsco-operationpatterns.Usingtheabovedis-678 35 er tinction ( Section 2.2.1 ), we may say that co-operation 679 18 The coefcient of id-high-tech is negative and that of with rivals aims at pooling similar resources to face 680 inicmhmeansthatthesec high R&D costs, while co-operation with universities 681 hviegrthi-ctaelchcios-onpoenratsiognisarnetl,atiwvheilcymorefrequentaretloorsw-wahnerde targets complementary resources to work at the tech-682 mid-low-tech. nological frontier. 683