International regional patterns of R&D networks involving low tech SMEs

A large number of studies have emphasized the spatial proximity of economic activity and its relation to the spatiality of knowledge creation in various types of connections. Far less attention has been paid to the understanding of the determinants of ‘cultural’ and geographical proximity in international R&D cooperation projects involving SMEs and the role of the quality of the regional innovation system (RIS). Using a database of completed Cooperative Research projects, we conclude that: 1) technologically more complex projects are more likely to involve ‘culturally’ and geographically distant partners; 2) RIS related variables determine ‘cultural’ proximity but not geographical proximity; 3) at first sight surprisingly, international cooperation projects involving the 1 st promoters of innovation-led regions (high patent propensity and high human capital levels) are culturally more distant.


Introduction
Several authors (e.g., Porter, 1990;Krugman, 1994) have empirically and theoretically made the case for the benefits of geographical proximity, or the location of firms in regional agglomerations (Storper, 1995). Proximity is said to enhance the performance of the firms through knowledge spillovers, informal information exchange, and the easier availability of inputs, skills, and other shared resources (Scott, 1992;Harrison et al., 1996;Rosenkopf et al., 2001). Concurrently, the trend toward formalized cooperation, namely through inter-firm alliances, joint ventures, and R&D and other agreements, has been intensifying (Powell and Brantley, 1992). Both these phenomena reflect the accelerating pace of innovation, and the related requirements of staying abreast of technological and market trends, integrating relevant knowledge, and developing new products and processes (Lundvall, 1992).
Recent studies (Doloreux, 2004;Kingsley and Malecki, 2004;Fristch, 2003) gather evidence that localized external networking is less prevalent than might have been expected. In fact, in many instances, regional networks for technological development are not the most predominant ones (Doloreux, 2004). Moreover, proximity does not seem to be an important factor in shaping the structure of informal networks or the use of information (Kingsley and Malecki, 2004). This evidence concludes that the importance of proximity has in fact been overestimated. It further demonstrates that SMEs make use of a mixture of local/regional, national and even international knowledge sources, and that their ability to sustain networks at different regional scales is a key factor in competitiveness and innovativeness.
There is thus a pressing need to develop a more refined line or argument that breaks with the current view on the localized character of innovation in order to provide a deeper understanding of other similar forms of linkages which are more dispersed in space. In Bunnell and Coe's (2001: 570) words, there is a '… need for a qualitative shift away from work which focuses on particular scales as the locus for understanding innovation, towards that which gives more credence to relationships operating between and across different scales.' Such re-focusing would lead to both theoretical enrichment and new practical applications in public policy with regard to innovation networks.
It is therefore apparent that while a large number of studies in economic geography has emphasized the spatial proximity and clustering of economic activity and its relation to the spatiality of knowledge creation in various types of connections, far less attention has been paid to the understanding of the determinants of 'cultural' and geographical proximity in international R&D cooperation projects. Furthermore, as Fritsch (2003) recognises, not much is known about the significance of interregional differences in the propensity of enterprises to cooperate, and the relationship between cooperation behaviour and the quality of the regional innovation system (RIS).
The present study seeks to address some of these gaps by providing some results on formal international R&D networks comprising SMEs and R&D institutes. In particular, it tries to assess the effect of technological complexity, project size, and the strength of the regional system of innovation on the cultural and geographic proximity of network participants. The paper proceeds as follows. The next section of the paper develops the theory and presents empirical evidence on Regional Innovation Systems (RIS), cooperation and networking. Then (Section 3) details the empirical analysis of successful R&D alliances by outlining some essential features of CRAFT projects, setting the methodology and presenting results. Finally, we present the main conclusions of the study.

Regional Innovation Systems (RIS), cooperation and networking
Although the 1980s witnessed the emergence of several complementary notions connecting technology and regional development policies, 1 the concept of RIS only became official at the beginning of the 1990s with Cooke's work (Cooke, 1992) and gained widespread dissemination with the book edited by Braczyk et al. (1998) entitled Regional Innovation Systems.
The RIS was inspired in the National Innovation Systems (NSI) (Lundvall, 1992;Nelson, 1993) and was similarly guided by the rationale of 'spatially based innovation systems' (Asheim and Gertler, 2004). Previous authors (e.g., Scott, 1995) emphasized the importance of the region, seeing it as an institutional depot of a given agreed-on, evolving, collective social order. Ohmae (1995) identified the regional, rather than the national level, as the key economic scale at which leading-edge business competitiveness is being organized in practical terms. Moreover, it became a matter of substantial scientific interest as to whether a distinctive kind of innovation phenomenon might exist at regional level (Cooke, 1998).
In the course of seeking for high value, better quality products, competitiveness and innovativeness have become inextricably linked to governments trying to restore their technological policies and create 'good business climates' in order to encourage their industries and firms to compete (Bell, 2005). Firms, in turn, responded to increasing demand and globalized competition by forming value added partnerships and strategic alliances, namely in R&D (Becker and Dietz, 2004). At the same time, they draw in those public or quasi-public agencies with complementary assets in terms of vocational training, enterprise support and technology transfer (Busom and Fernández-Ribas, 2004).
Cooperative agreements to perform R&D activities have been increasing over the last twenty years in OECD countries: the share of patent co-applications in triad patent families has almost doubled since 1980, and the number of strategic technology alliances has almost tripled on average (OECD, 2002). This trend is also documented by Hagedoorn et al. (2000).
The Community Innovation Survey (CIS), conducted in EU member countries, provides additional evidence on the importance of R&D partnerships and its variation across firms, industries and countries (Busom and Fernández-Ribas, 2004). According to CIS3, on average 17% of manufacturing firms with innovative activities indicate they had cooperation 5 agreements in 1998-2000; the share is significantly higher for large firms (61%). Partnerships with suppliers or customers are as frequent as partnerships with universities. Cross-country differences are significant: in Finland, 22% of SMEs in the manufacturing industries declared being involved in cooperative agreements in 2000, while in Spain or Italy barely 3% of SMEs did.
Several authors argue that the importance of R&D cooperation has risen steadily as a result of growing complexity, risks and costs of innovation (Coombs et al., 1996;Dogson, 1993;Hagedoorn and Schakenraad, 1992). Inter-firm collaborations occur especially within technology based industries. For instance, Arora and Gambardella (1994) demonstrate the high importance of R&D collaborations for large US chemical and pharmaceutical companies in the biotechnology sector. Colombo (1995) provides empirical evidence of complementary relationships between inter-firm cooperative arrangements and R&D intensity for a representative sample of international firms in the information technology industries (semiconductor, data processing and telecommunications). Veugelers (1997) identifies significant positive effects of R&D cooperation in the Flemish manufacturing industry on the level of R&D investments but only if firms have established absorptive capacities as a full-time staffed R&D department. For this reason, R&D-active Flemish firms are found to be more frequently engaged in technological cooperation, the more they spend on in-house R&D.
Accompanying firms' increasing cooperation efforts, public support programs to promote R&D cooperation have been implemented since the nineteen eighties in the US, Japan and the EU. The European Union's successive European Framework Programs are a significant example. Participation rates in these programs vary widely across EU member countries. In per capita terms, Denmark, Finland and Sweden had the highest participation rates in Framework Program 5, while Spain and Italy had among the lowest rates (Busom and Fernández-Ribas, 2004).
At the European level, significant efforts particularly targeting SMEs have been made to stimulate these firms' cooperative efforts as a way to overcome their research handicaps and scantiness of innovative inputs and as a booster of regional development. Cooperative research (CRAFT) is one of the main supporting measures within the EU Framework Programs targeting the R&D needs of SMEs and aiming to facilitate trans-national cooperation between SMEs and Europe's research community.
supply small and specialized markets, to rely on local resources and induce growth at local and regional level, the literature suggests however that SMEs often face a limited spatial horizon, with few opportunities to innovate, expand and be active in international markets (OECD, 2004). In this vein, the main factors that explain why SMEs would perform poorly are: lower production levels, limited access to scale economies; lack of appropriate organizational, administrative and managerial skills; and financial restrictions. These last two factors act as a barrier to access (use) new technologies or to engage human resources with the appropriate skills (Albu, 1997). As Benavente and Crespi (2003: 4) emphasize '… the main problem of SMEs is not their size but rather their complete isolation'.
Notwithstanding, there is clear evidence that SMEs actively network (Bryson et al., 1993;Johannisson, 1986;Pache, 1990;Szarka, 1990). Research has shown how SMEs network in a general context (MacMillan and Farmer, 1979;Pitt et al., 1990), or in an entrepreneurial context (Aldrich and Zimmer, 1986;Birley, 1985;Johannisson and Nilsson, 1989). As far as innovation activities are concerned, networks spread risks and costs amongst members (Nesta and Mangematin, 2004) and represent an important organizational form for SMEs which often lack the adequate financial and human resources.
The importance of interaction in innovation processes also makes it clear that networking is an essential means of knowledge exchange and learning (Doloreux, 2004). Contemporary innovation theories emphasize the interactive practice of innovation and the relation between firms and their environment. According to such innovation processes, they are not isolated in their origins, but rely on a variety of factors, internal and external to the firm (Edquist and Hommen, 1999;Kline and Rosenberg, 1986). External relations formed with other producers, suppliers, universities, research institutes and local support organizations can be the source of new ideas for innovation (Costa and Teixeira, 2005). As a result, innovative networks have become a persistent organizational phenomenon in industrial organization processes (Doloreux, 2004).
Several hypotheses have been proposed in the literature to explain the incentives that firms have to cooperate with other firms or with public research institutions (Caloghirou et al., 2003): 1) to develop innovation, firms may need complementary intangible assets, basically information, tacit knowledge and know-how, which cannot be easily contracted. In that case, monitoring and related problems can be better solved through a partnership; 2) cooperation allows partners to share risks and costs of innovating under growing technological complexity, as well as to exploit economies of scale and scope; 3) when out-coming R&D spillovers are a serious concern, R&D partnerships provide a mechanism to internalize them; 4) R&D cooperation may allow partners to increase market power in the product market.
Most of the literature on the networking and innovation performance of SMEs tends to highlight the role of physical (geographical) and cultural proximity and to stress the quality of the territory as a determinant in the performance of firms and the local/regional economy (Lundvall, 1992;Storper, 1995;Lorenzen, 1998). However, regional innovation networks require conditions beyond mere physical proximity and concentration to thrive (Gregersen and Johnson, 1997). As RIS literature argues, the spatial proximity of innovation networks is also tied to a common social and cultural understanding, without which the relationship between close agents can be hindered (Braczyk et al., 1998). So, proximity is not just an issue of geography, but equally of the degree to which economic, organizational, relational, social and cultural realities are shared. In this sense, 'cultural' proximity might be as (or more) relevant than physical proximity.
As Doloreux (2004) correctly argues, although proximity is an important determinant of interaction and cooperative behaviour implicit in networking, firms can benefit from participating in sector or knowledge specific networks linking interregional or international partners. In a complementary line of research, Fritsch (2003), studying the effects of R&D cooperation on innovation activity of manufacturing establishments in some European regions, argues that the hypotheses that cooperation boosts innovation is too simple as regions with low/high propensity to cooperate in R&D have high/low efficiency of innovation activities.
It is thus clear that the existing literature on innovation, networking and SMEs usually sees geographical and 'cultural' proximity as explanatory variables for cooperation intensity and propensity. To our knowledge, no study has so far empirically assessed the determinants of proximity, namely to what extent innovation related variables determine the intensity of cooperation between geographically or culturally close partners. Moreover, as pointed out by Becker and Dietz (2004), there remains an unexplored gap in the literature, which deserves further research: the need for the innovation effects of joint R&D to be analyzed under dynamic aspects with special attention paid to regional innovation systems (RIS).
Using a database of 118 international cooperation projects -'success stories' of completed CRAFT projects -, we aim at providing a contribution which may complement the existing literature on the relation between cooperation technology traits, RIS features and (cultural and geographical) proximity between cooperative participants. Thus, this study seeks to answer the following main questions: 1) What is the effect of technological complexity on the geographical and cultural distance of network participants?; 2) What is the effect of project size on the geographical and cultural distance of network participants?; and 3) What is the effect of the strength of the regional system of innovation on the cultural and geographical distance of network participants?
The following section describes CRAFT projects in detail and the database constructed for the empirical work.
3. The empirics of successful R&D alliances: the case of CRAFT projects

Some basics on CRAFT projects
In order to enhance the performance of member countries, their organizations and citizens,

The data -'successful' CRAFTs
In this study we analyze a sub-group of CRAFTs, those from the 3 rd to 5 th FP considered 'successful'. Information on these projects is available in http://sme.cordis.lu/craft/home.cfm.
Based on this information, we constructed a database which contains general information on each project, mostly gathered from pdf files that describe the projects, 3 and RIS-related variables (human capital, patent propensity and employment rate) taken from the main regional indicators published by the EC (2004).
Although success criterion is not explicit, some contacts with CRAFT national managers convey the information that 'successful' projects were those which the main performer reckons as well-succeeded, which are in general associated to a new product/service being commercialised or in a nearby market stage. 4 Although the inclusion of non-successful projects could have generated interesting insights, no public information on the former is available. Given that our main aim here is to assess the relation between Regional Innovation System quality and geographical and cultural proximity in international R&D cooperation projects involving low tech SMEs, restricting our selection to successful projects does not represent a significant hindrance to the analysis. Earlier works (e.g., Cooke and Wills, 1999) also used successful CRAFTs projects as their empirical basis for testing SMEs preferences for both domestic and foreign networking. Table 2 shows that there has been a trend toward a decrease in the success rate of CRAFTs. Table 2 about here> The 118 successful CRAFTs involved 791 SME from 21 countries (18 from the European Union plus Switzerland, Norway and Brazil). In global terms, the CRAFTs under analysis were allocated around 118 million euros, which gives an average of 1 million euros per project. From the total, 52% were financed by the UE, that is, around 61 million euros.

< Insert
When considering the successful CRAFTs, the main industrial areas are: Machinery and Equipment, Agriculture, Building and Metallic products. High technology activities, such as Telecommunications, Computing and R&D have a smaller contribution. This last characteristic may result from the fact that the program aims to enhance cooperative research by groups of SMEs with low or medium technological capabilities, with a restricted capacity for proper research, and encourages SMEs consortia to entrust research activities to a third party (e.g. University or R&D institutes). Table 3 about here > According to the Pavitt taxonomy (Pavitt, 1984), around 64.4% of projects belong to supplier dominated or scale intensive sectors. Science based sectors encompass approximately 20% of total projects. If we analyze the country of origin of the SMEs participating in successful CRAFTs, we conclude that 73% of participating SMEs belong to Germany, France Spain, the United Kingdom, The Netherlands and Italy.

< Insert
< Insert Figure 1 about here> Spanish companies (103) are concentrated in 36 projects. That is, when a Spanish SME participates in a project it pushes two additional Spanish companies. Although on a smaller scale than Spain, there is also a relative intensity of internal cooperation between Portuguese partners, as each Portuguese partner seems to engage two other partners of the same nationality. The two countries with a higher number of participating companies, Germany and France, are also those that participate in a larger number of projects: 64 and 53, respectively.
Generally, cooperation between research institutions of the same country is weaker than between same country SMEs. Austria is an exception, with 3 Austrian institutions per project.

< Insert Figure 2 about here>
The 'successful' CRAFT database identifies the main region where each Project is developed.
With few exceptions it is the region where the prime SMEs contractor is located. Although it was not possible to identify the region of origin for all the participating SMEs, we analyzed the regional origin of the first promoter in order to identify the regions with a higher 'promoting' ability. The regions with a higher promoting capacity, involved in successful CRAFTs (Table 4) are Noroeste (Spain) and Sud-Ouest (France) both with 7 projects each, Vlaams Gewest (Belgium), Baden-Württemberg (Germany) and West-Nederland (the Netherlands), with 6 projects each. Although country sizes are very different, the United Kingdom and France are the countries with more promoting regions, 8 and 7, respectively. Table 4 about here> A successful CRAFT involves, on average, about 10 participating entities, generally from Germany and France, in supplier dominated and low technology sectors. In successful projects funding allocated to SMEs attains 49.8%. The average cost per SME of a standard project is 241 thousand euros, and the average yield of participating research institutions is 400 thousand euros.

Model specification and variable description
The aim here is to assess whether the intensity of cultural proximity or geographical distance in successful CRAFTs is determined by factors associated to RIS strength -human capital intensity, innovation related performance, labour market performance, etc. For this purpose, we propose two main model specifications. In both specifications, the dependent variable is explained by two sets of independent variables: variables related to project characteristics and RIS strength. The two model specifications proposed differ in the variables related to project characteristics. In particular, the restricted model considers only 'core' project characteristics, namely technological diversity, industry code (according to the Pavitt and OECD taxonomies), SMEs average project cost and RTDs average project yield. The alternative specification (enlarged model), besides 'core' project variables, also includes country specific variables. The latter reflect, for each project, the countries' weight in total SMEs and RTDs participation.
Where i: is a given CRAFT project; y: is the dependent variable -promoter identity index (proxy for cultural proximity, y 1 ) or average promoter-participant distance (proxy for geographical distance, y 2 ); CSW: country specific weight of SMEs or RTD project participants; α;β;λ and π π π π: are the independent variables coefficient estimators; e: is the random white noise error term.
In the construction of the proxy for 'cultural' proximity -promoter identity index -, we adopt the following methodology: considering the nationality of the first SME performer (1st contractor), we compute, for each project, the relative frequency of other SMEs and RTDs that belong to the same country of the 1 st performer. Thus, we assume that when there are no participants of the same nationality as the 1st performer in a project, 'cultural' proximity is minimum (0). Alternatively, the maximum value that this variable could take is 100%, which corresponds to the case where all project participants belong to the same country of origin as the 1 st performer.
With regard to the geographical distance proxy, we calculate, for each project, the average distance (in kilometers) between each SME participant and the 1st performer. As information on the location of each project participant is not available, distance was computed with reference to each country's capital.
The detailed description of each project supplied by CORDIS contains information on the project's number of support technologies (different technological areas associated with the project). Therefore, we take the natural log of the number of support technologies in a given project as a proxy for the project's technological diversity.
The project industry NACE code was transformed into a smaller number of industries, using the Pavitt (1984) and the OECD (1992) taxonomies. The first one aggregates industries into four main groups, by increasing order of technological complexity and nature of knowledge source -supplier dominated, scale intensive, specialized suppliers and science based. The OECD taxonomy groups industries into three main sets according to R&D intensity -low, medium and high technology industries. IC assumes the values 1 to 4 in the case of the Pavitt taxonomy and 1 to 3 in the other case.
The SMEC variable is defined as a ratio between the project's total cost, expressed in million euros, and the number of SME participants. Similarly, the RTDY represents the ratio between the project's total yield, expressed in million euros, and the number of RTD participants.
The country specific weight of each project's SME (RTD) participants is computed by taking the total number of SMEs (RTDs) belonging to a particular country as a percentage of the total SME (RTD) participants. For estimation purposes, and as depicted in Figure 1 Table 5 about here> Based on the data and the constructed proxies for the relevant variables, we aim at testing three main hypotheses.
Hypothesis 1: Technologically more complex projects are more likely to involve 'culturally' and geographically more distant partners.
Projects encompassing a wider range of support technologies and belonging to intensive high technology industries would tend to require partners with more diverse and complementary technological capabilities, that is, from different regional systems of innovation. Thus, we would expect that We might expect that the higher the (average) yield, that is, the size of the project, the more internationalized the project is, and therefore, dissimilarity among the nationalities involved would be more likely and the geographical distance would be greater. That is, we would expect that

Estimation results
The model was estimated using OLS. The number of projects included as valid observation is 114, as no regional data was available for one of the countries (Norway), and for the remaining three projects some variables are missing.
Estimation results for the 'cultural' proximity proxy are presented in Table 6 and for the geographical distance proxy in Table 7.
In general, all the estimated models present a reasonable fit. In fact, when considering jointly the adjusted R square, F-statistics, and the Durbin Watson, we have sufficient statistical evidence that the models reflect reality quite appropriately.
With regard to the factors determining 'cultural' proximity, it is clear cut that regardless of the model specification (I-VI), the technological diversity variable is always highly statistically significant and negative. Controlling for all the other variables likely to influence the project's 'cultural' proximity, the lower the number of support technologies associated with the project, the higher the percentage of entities (SMEs and RTDs) with the same nationality as the 1 st performer. This seems to convey the idea that technologically more homogenous (and specialized) projects tend to be less demanding in terms of international cooperation.
The above results are reinforced by the industry code variables. In both the Pavitt and OECD taxonomies, the coefficient estimates are negatively and highly significant. Given that high values for these variables translate into higher technology intensive projects, a negative coefficient indicates that these types of projects tend to attract entities belonging to a wider range of countries.
When one does not consider country specific weights, higher SME average costs seem to induce lower 'cultural' proximity as the coefficient estimates become negative and significant. This highlights the fact that more costly projects, from the perspective of SMEs, tend to attract a broader range of international partners. However, the RTD average yield does not seem to matter when explaining 'cultural' proximity.
The introduction of country specific weight does not impact severely on 'cultural' proximityexcluding the cases of Spain and France (for SMEs), the estimated coefficients are not significant. For these two countries, although coefficient estimates are small, they indicate that the higher the percentage of French or Spanish entities, the higher the promoter identity index. This reflects that fact that French and Spanish SMEs are more prone to cooperate with SMEs of the same nationality.
RIS strength variables seem to stress that projects involving a lower 'cultural' proximity are more demanding in terms of regional innovation supporting environments characterized by a high innovation capacity (patents applications and human capital intensity). Finally, higher employment rates seem to determine higher 'cultural' proximity.
The results regarding the factors explaining average promoter-participant distance (proxy for geographical distance) are, in general, in line with those obtained for the promoter identity index.
CRAFTs characterized by a higher degree of technological diversity determine, all other factors constant, a higher average distance between promoter SMEs and other participants. This reinforces the idea stressed above that in projects with a low the number of support technologies, the percentage of entities (SMEs and RTDs) with the same nationality as the 1 st performer tends to be high, explaining in part the small average geographical distance between promoters and participants.
In contrast to the previous case, industry codes (both Pavitt and OECD) present no significant statistical coefficients, conveying the idea that project industry technological features are not relevant for explaining project average promoter-participant distance.
When we control for the intensity of SMEs and RTDs belonging to countries which have a greater presence in successful CRAFTs, the larger the size of the project, in terms of average RTD yield, the higher the average distance. From a cost perspective, however, high cost per SME projects include more geographically nearby partners.
It is interesting to note that the higher the weight of participants (SMEs and RTDs) from France, Germany (only in the case of SMEs), the Netherlands and the United Kingdom, the lower the projects' average promoter-participant distance. This might be explained by the fact that projects involving SMEs from these countries have a higher propensity to include RTDs of the same nationality than other countries. This decreases the average distance between the promoter and other project participants.
Excluding the employment rate, other RSI related variables do not seem to explain average distance among the different international cooperation projects. In fact, estimates are statistically insignificant for both patent and human capital variables. In comparison to the 'cultural' proximity case, where RIS strength seems to matter greatly, when it comes to geographical distance such a role is not underlined.

< Insert Figure 3 about here>
The above evidence suggests that our first hypothesis 'H 1 : Technologically more complex projects are more likely to involve 'culturally' and geographically more distant partners' is corroborated. This might be explained by the fact that, when controlling for indicators of RIS strength (patent propensity, human capital intensity and labour market dynamics), projects encompassing a wider range of support technologies and belonging to high technology intensive industries (more science based, according to the Pavitt taxonomy, or high tech, following OECD typology) tend to require partners with more diverse and complementary technological capabilities. Such complementary is more likely to be found in firms from different national systems of innovation, reflecting diverse, yet interrelated technological specialization fields.
As to our second hypothesis, 'H 2 : Larger international cooperative projects involve more culturally dissimilar and geographically distant partners', the evidence gathered only corroborates this when we introduce the SME or RTD country weights. In this case, and assessing project size by the yield accruing to participant RTDs, we conclude that the higher the (average) yield, the more internationalized the project is and, therefore, dissimilarity among the nationalities involved is more likely (lower 'cultural' proximity) and geographical distance is greater (average promoter-participant distance).
Given that RIS related variables, namely patent propensity and human capital, become significantly negative when related to 'cultural' proximity (although not statistically significant when it comes to geographical distance), we can point out that the data corroborates in part H 3 , that is, 'International cooperation projects involving 1 st promoters of innovation-led regions (high patent propensity and high human capital levels) are culturally more distant'. Corroborating such a hypothesis could be interpreted as meaning that promoters of international cooperative projects who come from more endowed and innovative-led regions tend to lead more complex projects from a managerial point of view.
In fact, to manage a wide range of firms of different nationalities implies considerable networking and relational competencies, which are more likely to be found in firms from well-developed regions. Firms that belong to weaker regions in terms of human capital and innovative performance may generally lack strong networking competencies, and they are thus induced to concentrate on more 'national', culturally nearby partners. 6 < Insert Table 6 about here> < Insert Table 7 about here>

Conclusions
In spite of increasing global flows of ideas, capital, goods and labour, recent research in urban economics and economic geography suggests that geographical proximity between innovators may be important to technological innovation (Won Sonn and Storper, 2003). Many authors also claim that the rise of a knowledge-based economy and changes in the organization of the innovation process have actually increased the value of such proximity to innovation. Yet there is little empirical research on whether this latter proposition is valid. Concurrently, the trend toward formalized cooperation, namely through inter-firm alliances, joint ventures, and R&D and other agreements, has been intensifying (Powell and Brantley, 1992).
Both these phenomena reflect the accelerating pace of innovation, and the related requirements of staying abreast of technological and market trends, integrating relevant knowledge, and developing new products and processes (Lundvall, 1992).
Recent evidence concerning the role of proximity is puzzling. Some studies (e.g., Doloreux, 2004;Kingsley and Malecki, 2004;Fristch, 2003) conclude that the importance of proximity was not substantiated, gathering evidence that, in many instances, regional networks for technological development are not the most predominant ones (Doloreux, 2004) and that proximity does not seem to be an important factor in shaping the structure of informal networks or the use of information (Kingsley and Malecki, 2004). Others (e.g., Won Sonn and Storper, 2003) find that proximity in the creation of economically-useful knowledge appears to be becoming even more important than was previously the case.
In the present study, with a view to complementing existing evidence in this field, using new data on successful international formal R&D alliances, we aimed to assess the determinants of 'cultural' proximity and geographical distance, highlighting the role of the quality of the regional innovation system (RIS). More specifically, we assessed the effect of project technological complexity, project size, and the strength of the regional system of innovation on the cultural and geographical proximity of network participants.
We conclude that technologically more complex projects are more likely to involve 'culturally' and geographically more distant partners and that the 1 st promoter's regional patent propensity and human capital endowments (proxies for RIS strength) determine 'cultural' proximity but not geographical proximity among network participants. Seemingly at odds with most geographically related literature, we find that international cooperation projects involving 1 st promoters of innovation-led regions (high patent propensity and high human capital levels) are culturally more distant. Such apparently disparate results might be interpreted as meaning that the promoters of successful international cooperative projects who come from more endowed and innovative-led regions tend to lead more complex projects. In other words, to manage a wide range of firms of different nationalities tends to imply considerable networking and relational competencies, which are more likely to be found in firms located in high quality (stronger) RIS. First SMEs promoters belonging to weaker regions in terms of human capital and innovative performance may, in general, lack strong networking competencies inducing them to concentrate on more 'national', culturally nearby partners. This paper's findings seem to imply that if RIS quality is a relevant condition for local SMEs to access a wider, more complex R&D network, then there is room for additional reflection about whether public policy should take this explicitly into account. Most technological policies are meant, by design, to be indifferent to location, precisely because they want to stimulate knowledge diffusion, competition, and avoid generating rents (Won Sonn and Storper, 2003). Along these lines, there appears to be some reason to think that local governments should privilege those activities that can endogenously create 'good business climates' in order to assist their industries and firms to compete (Bell, 2005). In particular, they should stimulate investment in human capital and easier procedures to facilitate and encourage patent applications, which would contribute to SMEs competencies accumulation, creating a basis for their involvement in increasingly technologically complex and more distant (successful) R&D networks. The evidence presented here emphasizes both the importance of localized technological interactions, namely the strength of RIS and the increasing relevance of long distant R&D cooperation networks. 1 Relevant literature in these lines includes 'regional innovation policies' (Antonelli and Momigliano, 1981), 'innovative milieux' (Aydalot, 1986;Maillat and Vasserot, 1986), 'regional technology policies' (Rothwell and Dodgson, 1991), 'regional innovation potential' (Meyer-Kramer, 1985), and 'innovation networks' (Camagni, 1991). 2 In http://europa.eu.int/comm/research/faq/index.cfm?lang=en&page=details&idfaq=3600, accessed on 22 April 2005. 3 Santos (2005) provides further details on the construction of this database. 4 The concept of 'successful' was provided by Margarida Garrido (Cordis). 5 We consider here NUTs level 1 as defined by EC prior to May 2003. 6 As prior literature has suggested cultural and geographical variables emerge from these results as connected factors. Recent thinking, however, suggests the linkage between these factors may be loosening as the mobility of cultures increases (Saxenian, 2005;Florida, 2004). In order to address (although partially) this point we further estimate the two set of models -one (Table A1 and A2 in Appendix) that includes the geographical (cultural) variable as an independent variable in the models that seek to explain the variability in the index of cultural proximity (geographical distance); another in which the dependent variable is a global index of cultural and geographical proximity (computed as a simple average of cultural proximity and 1-Index of geographical distance) ( Table A3). The cultural and geographic variables were thus segregate and tested both jointly and separately to discern differences in the effects. The results (significance and signals of the individual estimates) obtained do not significantly differ from the original set of results, and further evidence that cultural and geographical indexes are in this sample of projects statistically significantly related. We acknowledge one of the referees for such insightful point. To support SMEs in their needs for research and innovation.
To encourage international cooperation between SMEs and between these and research entities.

Number of participants
Consortia involving a minimum of three SMEs from two different countries.
Consortia must include at least two R&D partners from a minimum of two different members or associated countries.
Other companies or final users may participate in the consortia but they must share the costs and they are not allowed a dominant role in Project development. They must be independent from other project participants.

Owners of intellectual property rights Participating SMEs
Project value 0,5 -2 million €.