Towards a Taxonomy of Firms Engaged in International R&d Networks: an Evaluation of the Spanish Participation in Eureka

Innovation is increasingly becoming an internationalized process and a strategy that has recently been playing a central role in this scenario is that of R&D collaboration. To assess the outcomes of this strategy we develop an evaluation of Eureka Programme's impact for the case of Spanish companies participating in this initiative and that had projects finished in the period 2000-2005. A total of 77 firms were assessed through statistical association methods and cluster analysis. Company size, Role in the Project, Firm Sector and R&D intensity are significantly associated with the projects' impacts on Spanish participants. A consistent taxonomy is offered in which three clusters are built: (1) Risky Innovators; (2) Inventors; and (3) Consistent Innovators. goal fostering innovation through cooperation between organizations from different nations: the Eureka Programme. The objective is to achieve an approximation of companies' profiles when joining such an initiative, generating workable indicators of output determinants in this framework in a context of impact measure, a largely unexplored area of The analysis here undertaken is based on a quantitative approach of Eureka's Final Reports of projects completed by Spanish companies during the period 2000-2005. Spain represents an interesting case of study for cooperative R&D for as much as it is one of the most dedicated participants in Eureka, still shows a low level of R&D collaboration between agents and is historically highly dependent on foreign sources of technology (Fernández, Junquera and Vázquez, 1996). Moreover, Spanish firms' characteristics indicate a low propensity of achieving innovations through collaborative settings in comparison with internal R&D efforts (Vega-Jurado, Gutiérrez-Gracia and Fernández –de-Lucio, 2008). These reports are structured in a way that allows the assessment of descriptive information (general features of the companies such as size and status of participation in the project) and general impact of the project (technological achievements, commercial impact, industrial exploitation). Data regarding companies' main characteristics (more detailed data of size and industrial sector) were also combined with the original database. The methodological approach is divided in two parts: analysis of associations (through cross-tabs and chi-square tests) and a proposal of taxonomy of participants. The paper firstly outlines the literature on cooperative R&D with special focus on international relationships. The main features of Eureka are presented, as well as previous results of evaluations undertaken. Subsequently, the methodology of the research is presented, introducing the main characteristics of the sample and the specific methods applied. Results are presented and …


Introduction
Innovation policies are a matter of great concern worldwide and in the European Union this situation is not different.Much has been said about the "European Paradox", i.e., the difference between scientific capabilities and actual innovation performance (Georghiou, 2001) and, therefore, several measures took place in order to modify this scenario.The current rationale is one of coordination and convergence between Regional, National, and International levels regarding RTD policy within the European Union (Manjón, 2010).
Broadly speaking, these programs that stimulate innovative activities take place to correct the market failures associated with R&D investments (Klette, Moen and Griliches, 2000).Nonetheless, unsatisfactory results in this area are mainly attributed to lack of R&D investment and to a low productivity of the resources invested (Benfratello and Sembenelli, 2002) showing a strong need for the analysis, evaluation and measurement of current innovation and technological policies.
But this cannot be regarded as a simple task depending solely on recognizing the underlying difficulties and designating funds for it.Despite important conceptual and methodological advances in the economics of science and innovation in recent years, there is still little agreement as to what 'good' science, technology and innovation (STI) policy should look like and which instruments should be used (Laranja, Uyarra and Flanagan, 2007), which gives an idea of the complexity involved not only in formulating innovation policies, but also in evaluating their impact.Bin and Salles-Filho (2012) suggest that RTD policies must cope with the evolving characteristics of markets, thus adapting constantly to a changing environment and promoting desirable behaviors in agents within economic systems.
Embedded in this scenario is the existence not only of firms' strategies to cooperate in R&D, but also its international tendency and a whole set of initiatives that promote this kind of activity.It is well known that not only for firms, but for innovation systems, this sort of integration can be very beneficial for technological growth and evolution, being a key determinant of competitiveness (Archibugi and Iammarino, 1999;Suurna and Katel, 2010).Nonetheless, approaches in this regard are somewhat controversial and there still is an important gap in terms of policymaking implications of R&D cooperation initiatives as well as a stronger framework to foster these activities (European Commission, 2011).
The scope of this paper lies in analyzing through technological and commercial impacts at national level (the case of Spanish firms) one of the most relevant technological programs that take place in the Europe and that has as its main instead of direct financial assistance policies (De Jong and Freel, 2010).This interest from governments in promoting international research collaboration comes primarily from expectations of cost savings and other related benefits (Katz and Martin, 1997).Cooperative R&D policies gain even more importance when one considers that the extent to which a country's businesses, institutions and industries are linked with resources and capabilities located outside the country is likely to positively impact the innovation performance of that country (European Commission, 2010), creating local externalities from global relationships.Also, the idea of international scientific and technological cooperation can be regarded as fundamental for the development of products that demand joint R&D due to specialization patterns in different economies or regions, i.e., the idea of complementarities between firms should also be considered as promoting integration between technically and economically heterogeneous territories.Thus, collaboration fosters knowledge transfer in a context of international economics.Narula and Santangelo (2009) hypothesize that R&D alliances might even act as a substitute for collocation, or as a complementary mechanism for it, embedding the idea of international R&D cooperation in the economic geography framework.
In Europe, the creation of the European Research Area stands for a coordination of closer R&D cooperation between organizations of EU's Member States (Georghiou, 2001).It is interesting to highlight the adaptive role of the policies in this field -R&D cooperation did not follow governmental initiatives but the other way around.Also, An evaluation undertook by the European Technology Assessment Network (ETAN, 1998) concludes that European firms not only have a internationalized S&T profile, but are also increasing its technological alliances and international generation of innovations within Europe and beyond, even though not in the same level as firms in the United States (Foray and Lhuillery, 2010).
However, this growing interest in technological cooperation analysis is followed by a high level of complexity involved in studying it.Some models were developed in the past decade trying to cope with non-linear and non-direct relationships between the variables used in the evaluation.Crépon, Duguet and Mairesse (1998) wrote the most influent article in this sense -they approach this idea of complex interrelations with a model of simultaneous equations that allow the analysis of indirect relationships.Their results show that technological cooperation agreements have a positive effect in the achievement of innovations which leads to better economic outcomes, suggesting an indirect relationship between cooperation and economic performance via innovations.
History shows that R&D partnerships have been growing since the 1960s with a noticeable acceleration in the 1980s.This is the result of the increasing level of complexity of R&D projects in recent years, higher uncertainty surrounding R&D, increasing costs of R&D projects, stronger competition and shortened innovation cycles that favor collaboration in face of an environment with more specialized organizations in terms of knowledge production (Pavitt, 2002;Hagedoorn, 2002;Narula, 2001;Zeng, Xie and Tam, 2010;Katz and Martin, 1997;Jonkers and Castro, 2010;Pellegrin et al, 2010).Other benefits of cooperative R&D come from the assumption that it increases the efficiency of R&D efforts, provides more flexibility to adapt to technological changes and eliminates wasteful duplication; also cooperative R&D agreement may serve as a mechanism that internalizes the externalities created by spillovers while continuing the efficient sharing of information (Katz, 1986).Moreover, the process of globalization itself has influenced firms' behavior and technological characteristics of innovations by increasing outsourcing and strategic alliances and also by promoting increasingly multitechnological products (Narula, 2004).
As a consequence of these trends there is an emergence of new forms of interaction between firms, fostering an environment of "open innovation", meaning that many companies across industries externalize several R&D activities, focusing on their core competences and absorbing third parties' capabilities (Wagner and Edelmann, 2002;Herstad et al, 2010;Savitskaya, Salmi and Torkkeli, 2010).This implies that firms use R&D partnerships to access knowledge, expertise or skills and build global R&D networks, being the choice of partners dictated by the complementary resources which the counterpart controls, allowing companies to improve their performance (Miotti and Sachwald, 2003;Georghiou, 1998).One significant outcome of this scenario is that especially large companies are likely to capture results more easily -because of an expected higher absorptive capacity in comparison to SMEs -and to become less self-sufficient in their processes, being able to incur in the division of innovative activities (Fritsch and Lukas, 2001;Veugelers, 1998;López, 2008;Bayona, García-Marco and Huerta, 2001) which according to economic theory should lead to scale economies.This does not mean at all that R&D cooperation has no effect on SMEs: the point to be noticed here is that smaller firms are not likely to proceed to internalization of processes in the first place, making them more prone to outsourcing by their own organizational definition.Edwards-Schachter, Castro-Martínez and Fernández-de-Lucio (2011) add to this framework the possibility of SMEs using international R&D cooperation as a strategy to achieve access to new markets.Efforts on R&D cooperation are especially relevant in OECD countries, where the increasing number of R&D strategic alliances stands for a new organization in industrial technological structure focused on network promotion policies Cooperative R&D structures can be seen as innovative per se as it creates a new institutional framework for companies to cooperate in the generation of technological change.Policies fostering cooperation also show adaptive characteristics since they cannot be regarded as linear: they promote a more complex and holistic approach to innovative processes in opposition of direct funding initiatives.But one has to be very careful when analyzing collaborative R&D and its related policies.For many sectors, cooperation regarding innovation may be too dangerous for companies' appropriability strategies -as it is the case of the pharmaceutical sector which relies deeply on the launching of new products and in the intellectual property rights of these new drugs -sharing valuable information with competitors or even with agents from industries not directly related to the pharmaceutical sector might be too big of a threat for this organizations (which explains why this market is controlled by huge corporations with high degrees of internalization).Also, cooperation may happen in different stages of R&D.Some projects are related to basic R&D, others to pre-competitive activities and lastly (as it is the case of the Eureka Initiative), close-to-market cooperation (the one which poses the biggest risks for companies).Conceptually, R&D alliances can be distinguished from production-based alliances in terms of its fixed-term horizon and the fact that it covers only a small part of the value-adding activities of companies (Narula, 1999).So as it can be noticed, collaboration in the area of innovation can not only take different shapes in the interorganizational relationship (contracts, research joint ventures, etc.) but can also apply to R&D activities with different purposes.When dealing with evaluation of technological policies one cannot neglect these aspects.

The Eureka Programme: an overview
The Eureka Programme emerged as part of a concerted effort to bridge the widening technological gap observed since the 1960s between Europe and its global competitors: notably the USA and Japan (Eureka, 2005).It was created in 1985 by a French initiative as a complementary structure for the Framework Programmes aiming at enhancing collaboration between companies in a market oriented, non-bureaucratic, bottom-up approach promoting cooperative projects for national funding (Stubbs, 2001;Georghiou, 2001;Marín and Siotis, 2008).
It became a Europe-wide network that aims at increasing its participants' competitiveness through the promotion of cross-border ``market-driven'' R&D projects in which firms may seek entry for any projects that meet the broad criterion of developing advanced technology with a market orientation (Georghiou and Roessner, 2000;Bayona-Sáez and Conceptually, cooperative R&D consists of an arrangement among firms aiming at sharing costs and results of an R&D project and can be achieved through R&D contracts, consortia or Research Joint Ventures -The kind of cooperative agreement in which firms engage is largely determined by technological characteristics and sectors of industry (Sakakibara, 1997;Hagedoorn and Narula, 1996).The idea of open innovation formalizes the importance of these networking initiatives and absorptive capacity while reducing the focus on internalization of R&D activities (De Jong and Freel, 2010).As a matter of fact, external sources of knowledge and skills play an increasingly important role in innovation and the capacity of accessing and exploring this knowledge is fundamental for companies' competitiveness in the described context (Cohen and Levinthal, 1990).Also, an important prerequisite to manage the permanently changing dynamic market requirements and to secure the competitiveness is the linking and cooperation of companies (Wagner and Edelmann, 2002).
In an environment of constant technological change and high levels of R&D complexity, the best way to minimize risks and achieve sustainable competitiveness seems to be through specialization.It is impossible to imagine that this trend leads to economic growth if firms and agents do not interact with themselves (since they are all deeply specialized) or do not even have the capacity to do so.R&D cooperation practices have a twofold impact in this arena: on the one hand they create the possibility of firms addressing complexity in a multi-capability and multidisciplinary manner, promoting valuable innovations; on the other hand, R&D cooperation increases absorptive capacity and learning capabilities in the company, generating better prospects for future collaboration.This latter aspect is also pointed out by Barañano (1995).Therefore, promoting the strengthening of companies' technological skills through collaboration and therefore providing them with absorptive capacities is a fundamental focus that technological policies must consider (Luukkonen, 1998;Silipo, 2008).
But it is important to highlight that despite the increasing relevance of R&D cooperation and the growing literature about it in both the fields of management and industrial economics, there is little evidence on the performance effect coming from R&D collaboration (Belderbos, Carree and Lokshin, 2004).However, available analyses at the firm level show positive results: Zeng, Xie and Tam (2010) report that interfirm cooperation shows a significant positive impact on the innovation performance of SMEs in the Chinese environment.International R&D collaboration also seems to be positively associated with higher innovation expenditures (De Jong and Freel, 2010) and to provide firms with strategic flexibility to undertake short-term innovation projects with a variety of partners (Hagedoorn, 2002).b) Barañano (1995) suggests that Spanish Eureka participants see the improvement of the organization's public image as one of the most important features of the program; c) Marín and Siotis (2008) result's tell that it seems that Eureka serves the purpose for which it was designed, namely to correct the market failures associated with the generation of economically valuable knowledge; d) Fölster (1995) hypothesizes that, given that Eureka projects require cooperation but do not require result-sharing agreements, the likelihood of cooperation is not increased while do promote incentives to conduct R&D to the same extent as subsidies that do not require cooperation; e) Georghiou (2001) points that Eureka started with major projects but a decline since then took part driven by its divergence with national innovation policies.
So as it can be noticed, Eureka is a relevant target of innovation policy evaluation.But it is important to take into account that even though the results presented are mainly positive, continuous assessments and even different research foci might not only identify weaknesses of the program, but also provide information necessary for adaptations and changes in the initiative's characteristics.

The Sample
The sample consists in a subset of Eureka's database of Spanish participants in the initiative for the period 2000-2005.However, some adjustments had to be made for this database (consisting originally of 330 observations).The first stage consisted in two steps: 1.
Eliminating participants that did not respond the Final Report since information regarding their participation in the Eureka project was not available.

2.
Selecting those participants which were either Large Companies or Small and Medium Size Enterprises (SMEs) given the scope of the analysis.Research Centers, Universities and other institutions were then dropped from the database as we expect that these participants will not have market-driven behaviors necessarily.
After these adjustments the 2000-2005 database was left with 77 firms.A last effort was made to categorize companies according to their sector (NACE 2 digit Rev. 2) using the Amadeus database and to identify actual number of employees: 2 companies from the 2000-2005 subset could not be classified in this regard.García-Marco, 2010;Trabada, 2000;Marín and Siotis, 2008).It is also important to highlight the relevance of the bottom-up approach of this initiative: unlike programs that have clearly defined areas of interest for R&D projects, in Eureka, the nature and scope of proposals is defined by the proponents themselves.
Eureka is present in 38 countries and acts not through financial support but providing projects with a seal of approval that facilitates access to governmental funds in the national level as well as support in finding funding opportunities which makes it a fairly decentralized program (Stubbs, 2001;Georghiou and Roessner, 2000).Even though Eureka does not entitle firms to EU subsidies (it should be noted that Eureka is not an EU program), obtaining the Eureka "seal of approval" enhances firms' ability to receive support from their respective national authorities (Marín and Siotis, 2008).By conferring an objective seal of quality on a project, EU-REKA labeling greatly aids the process of negotiation with public sources of finance (several authors analyze signaling strategies and adverse selection risks in the context of R&D and innovation funding.For examples see Beatty, Berger & Magliolo, 1995;Takalo & Tanayama, 2010;Plehn-Dujowich, 2009;Janney & Folta, 2003;Bagella & Becchetti, 1998).Many member countries accord preferential treatment to labeled proposals by giving access to specifically reserved funding (Eureka, 2005).
Eureka's focus is on improving European competitiveness and productivity through an enhanced cooperation between companies and research centers in high-tech areas.Under Eureka, cooperation often consists of occasional meetings between firms at which information is shared (Fölster, 1995), but more formal ways of cooperation also take place.
Eureka carries out its own evaluation system through periodic reviews.In its first decade of existence, evaluations of projects were responsibility of the Member State holding the Chair for that year and in 1992-1993 Eureka had its first major evaluation, involving teams from 14 countries working together and carrying out a survey with all of the participants (Georghiou and Roessner, 2000).
However, besides its internal evaluations, Eureka is the focus of several academic analyses.Some examples: a) Bayona-Sáez and García-Marco (2010) demonstrate that participation in a Eureka Programme has a positive effect on firm's performance both in manufacturing and non-manufacturing sectors with a 1 year lag between project completion and performance improvements (which is in accordance with Benfratello and Sembenelli, 2002 results -they also highlight an increase in labor productivity and price-cost margins for participants);

Identification of Associations
Our empirical analysis starts with the results presented in table 2, which bring a summary of the cross-tabs (chi-square) results for significant associations between descriptive (columns) and impact (rows) variables.First of all, descriptive variables that did not show any significant relationship with impact variables were omitted from this table: a) Total Cost of Project; b) Total Duration of Project; and c) Role in the Project as Main Player or Partner.
First results report the relationship between companies' size and the group of selected impact variables.Results show that the size of the companies (SMEs or Large Companies) has an association with Commercial Achievements.SMEs seem to show a greater commercial impact as a result of their participation in the project than Large Companies.This result is somewhat expected since the commercial impact of one single project should be perceived as having a larger importance in smaller firms than it would be the case in larger corporations.
The analysis of companies' Role in the Project (as Producer, End User, Supplier, Research, Other or Multiple Roles) suggests that Technological Achievements appear to be related to companies' characteristics -Excellent achievements are obtained by firms playing the role of Producer; Good achievements are related to both Producers and companies that have Multiple roles in the project; and the poorest results can be associated with those companies that report having Other roles in the project (which might be an indication of smaller participation in Eureka).Also, it was found a significant relationship for firms that participate as End Users associated to Industrial Exploitation by Another Company.
When analyzing the association of results of the participation in the initiative with companies' sectors, it can be no-A general description of the sample used is depicted in Table 1 where the most relevant features of Spanish companies participating in Eureka with projects finished in the period 2000-2005 are compared in relative terms with the global average of Eureka's participants for the same period.

Towards a Taxonomy: Methodological Approach
Given the central purpose of this evaluation, the applied methodology consists basically in quantitative techniques that allow the construction of relatively homogeneous groups out of a sample and based on a set of predefined variables.Hence, the approach of this study consists in evaluating through statistical methods how variables are associated with themselves and how companies behave according to their characteristics and outcomes from their participation in the project.In a first moment, cross-tabs (chi-square) analyses are performed in an attempt to identify how descriptive variables of firms relate to their projects' outcomes.The second step undertaken is a cluster analysis that aims at verifying latent groups of companies with similar profiles either regarding their structure (size for example) or the impact of their participation in Eureka.This approach aims at generating in-depth knowledge on aspects that might contribute for the policy-making process at the Eureka (and maybe other similar initiatives) level.
The cross-tabs (chi-square) method represents a step ahead in the identification of associations, allowing for some inferential propositions.The approach described in this section is developed according to the following structure: descriptive variables are analyzed according to impact variables.The objective of this approach is to generate some knowledge on how variables such as company's size and its role in the project interact with the results achieved.It is worth noticing that these statistical interactions obey theoretical propositions.When analyzing impact variables, it is relevant for the study of innovation aspects to relate it to variables representing companies' size and their role in the technological project, as well as how technological achievements may influence commercial results, for example.Working with this set of Spanish companies we can, through this specific methodology be able to identify some valuable trends in the sample.
The cluster analysis developed in this paper has a rather exploratory character -instead of a confirmatory one.The objective is to provide some insights on a preliminary typology of Spanish participants in the Eureka Initiative based on a set of descriptive and impact variables.For this approach, the TwoStep Cluster (SPSS) method was used -this method is an exploratory tool designed to reveal natural clusters in the dataset according to the parameters indicated.As auxiliary tests showed, the TwoStep Cluster characteristics of the clusters are Companies' Size (Large company or SME), Role (as Main player or Partner), Role in the Project (Producer, End User, Supplier, Research, Other or Multiple Roles), Overall Technological Achievements, Functioning of the Project, Industrial Exploitation by the Company, Product Already on the Market and Commercial Achievements.
Table 4 brings a summary of the structure of the clusters built based on a TwoStep Cluster approach.One first aspect that has to be commented is that the outcome of the analysis suggested the division of cases in 3 clusters with rather similar sizes.Nonetheless, it is evident that some of the variables used in the classification do not necessarily perform a considerable separation between clusters as it can be seen in the composition of clusters and also through chi-square results for the variables.Results were kept in the original structure since this assessment has exploratory interests (and the cluster analysis itself is not an exact science).
As results show, the size of companies does not correspond to a good separation variable between clusters -Cluster 1 and 3 both have a similar structure and no particular cluster ticed that only commercial achievements show a statistically significant relationship.Regarding this result, Manufacturing and Services firms achieve better performances in comparison to firms from the Primary and Construction sectors.For this case, a more disaggregated level of sectoral analysis would be ideal, but the number of observations does not allow us to capture that picture.
R&D intensity is a variable that shows significant correspondence only with the launching of a new product on the market by the end of the project.This would be a hint that firms with higher levels of relative investment on R&D have a smaller time-to-market period, which can be useful information for Eureka when analyzing projects to be accepted and the specific goals of the initiative.

Taxonomy of Participants
In this part of this empirical assessment of Spanish companies' participation in the Eureka initiative for projects completed in the period 2000-2005, an attempt of developing an exploratory typology of firms included in the sample is performed.As it has been already mentioned in the methodological section, the set of variables used to define the Table 2. Summary of significant associations between descriptive and impact variables.
J. Technol.Manag.Innov.2012, Volume 7, Issue 3 correspond to the set of Large Companies -which are divided in small groups within clusters.A very comparable situation is provided by the Role as Main player or Partner -in this case, both clusters 1 and 3 are predominantly composed by Main players, while cluster 2 shows no defined characteristic in this aspect.These observations are supported by chi-square tests that do not provide either variable with a significant classification power.
The cluster analysis starts taking shape when considering Role in the Project as a separation variable.In this case each cluster has a clear predominance of each one of the three most common roles played by Spanish companies participating in Eureka for the period analyzed.Cluster 1 is mainly composed by Producers; Cluster 2 by End Users; and Cluster 3 by companies playing multiple roles.Nonetheless, chisquare results do not allow for an inferential confirmation of these patterns so Role in the Project performs as a rather suggestive variable instead of a confirmatory one.
Following this variable, Technological Achievements seem to provide some interesting level of discrimination between clusters: while Cluster 1 is mainly made of companies with excellent results, both Clusters 2 and 3 show companies with good technological results -this should be no surprise since 92,2% of the sample classified their technological achievements as either excellent (24,7%) or good (67,5%), but cluster 2 also shows the presence of weak technological results, which does not happen for either of the two other clusters.In this regard, the chi-square coefficient indicates that this variable represents a good classification aspect between groups.Functioning of the project, a variable that deals with internal aspects of management of the project, does well in separating cluster 1 from 2 and 3 in a similar manner to that generated by Technological Achievements (even though chi-square results show a good fit for this variable only for clusters 1 and 3).
Regarding Industrial Exploitation of results, Clusters 1 and 3 represent groups of companies that do have some level of exploitation, and Cluster 2 seems to be composed by both companies that exploit their project outcomes and those firms that do not (chi-square tests show a significance only for the latter case).A clearer division is provided by the variable Product Already on the Market: both Clusters 1 and 3 have the characteristic of having commercial activities already by the end of the project which does not happen with Cluster 2 (chi-square significant for groups 1 and 2).Lastly, the variable Commercial Achievements shows that Cluster 1 represents companies with a myriad of different results: while it is the only group containing firms with excellent results, it also comprehends companies with good commercial results, weak commercial results and even nil commercial outcomes.This structure is rather complicated to analyze as there is no clearly defined pattern (Excellent and Good results only account for 50% of cases).Cluster 2 is composed mainly by those firms with weak and nil commercial outcomes and Cluster 3 is related to those with good commercial achievements.
Focusing in those aspects that successfully divide clusters, the results indicate a general structure according to the following cluster profile: 1.
Risky Innovators -SMEs which participate in the project as Main Players, playing the role of Producers or End Users, that achieve excellent technological results through an excellent functioning of the project, exploit their results in the industry, have products being commercialized by the end of the project and this generates excellent commercial achievements for a group of companies comprehended in this cluster.The name of this cluster makes reference to the fact that companies comprehended in it have the best technical outcomes out of the three clusters, but only partially they can obtain satisfactory market results.

2.
Inventors -Large Companies and SMEs that play Multiple roles or the role of End Users in the project, that achieve good technological results through a good or weak functioning of the project, that do not necessarily perform industrial exploitation of results, that are not commercializing the outcomes of the project by the time of its completion, thus having nil and weak commercial achievements.These companies are classified as inventors for showing fair technical results without taking advantage of it in the market -which does not allow us to define them as innovators per se -at least by the time the Eureka project is completed.

3.
Consistent Innovators -SMEs which participate in the project as Main Players, playing Multiple roles or the role of producer in the project, that achieve good technological results through a good functioning of the project, exploit their results in the industry, have products being commercialized by the end of the project and this generates good commercial achievements.These companies have poorer technical results than the risky innovators, but truth of the matter is that they consistently achieve good commercial results.
One last aspect of this analysis concerns a quite obvious result according to theory, but that deserves some attention.Spanish companies participating in Eureka for the period 2000-2005 are mostly well satisfied with their technological attainments, which is an important aspect of the evaluation of any technological initiative.However, this is only part of the story: the companies' capacity of introducing their results in the market and exploiting the technical outcomes of the project clearly influence the point of view towards Table 3. Results of the TwoStep Cluster analysis commercial achievements -and when dealing with an innovation-driven approach (and not invention-driven), this latter part of the analysis is the one that matters the most.

Concluding Remarks
Technological policy evaluation is a process of utmost importance in any economic context that aims at fostering economic growth through technological progress and innovation.This is an exercise of constructive criticism with the ultimate goal of providing information and feedback that allow the continuous improvement and adaption of any kind of initiative -private, governmental or even supranational.
The work developed and presented in this paper represents an effort in this sense.A quantitative appreciation of a data-base composed by Spanish companies participating in the Eureka Initiative with projects finished in the period 2000-2005 made possible some interesting exploratory insights.The methodology used in our analysis had a quantitative character aiming at taking the step beyond purely descriptive assessments -even though we recognize the risks of it.We have seen that the overall rate of technological achievements is impressively high and even the commercial achievements can be considered outstanding in a context of innovation where R&D outcomes can be considered as uncertain by its own nature (Silipo, 2008).While this might indicate that Eureka is doing a really good job in selecting potentially successful projects, it might also suggest that companies may not be taking the level of risk necessary for introducing major relevant innovations in the market, which gather information of firms' outcomes from the Eureka project they have undertaken.These results basically support the outcomes of the clustering process while also providing valuable and workable information in the shape of indicators for the management of Eureka for both the selection of applicants and for the monitoring of ongoing projects.
However, our suggested taxonomy can become a bit blurry when thinking it through.What we name as risky and -especially -consistent innovators may be actually seen as mere improvers.Since we are not here able to assess the actual market relevance of their outcomes from projects, they might actually see their results as more relevant than a policymaker would in terms of significance of Eureka's contribution to a more competitive Europe.Also, the strategy of developing international R&D networks may foster long-run absorptive capacity improvements that may generate a critical mass of knowledge and capabilities for firms.Thus, one should not be surprised if inventors outperform their peers which we would in our taxonomy classify as innovators.
Main policy implications of our research point towards the relevance of Spanish firms' characteristics on the resulting impacts from their participation in a Eureka project.Clearly, the analysis of a given project per se does not provide nearly enough information for the decision makers to decide if whether this project should receive a seal of approval or not: characteristics such as the sector, the size, the R&D intensity and the specific role a company will be carrying out in the project are significant in defining to which group (according to our proposal of clusters) a company will belong to when the project ends.If the goal is to achieve ever increasing results, the outcomes of our research suggest rich information on what to consider in a firm before it participates in a European project, but one cannot forget that if the main goal is to sustainably improve the competitiveness of European markets through a more integrated R&D context, a short-term vision can be counterproductive.
Also, it is important to point out that Eureka seems to generate a fair amount of technological improvement for Spanish companies through fostering international R&D collaboration.Unfortunately, we could not assess the complexity of firms' networks outside Spain, which could give us valuable information and indicators for further analyses.Notwithstanding, Eureka's strategic position deserves revision.If it acts through labeling, i.e., information asymmetry reduction for funding markets (public and/or private), the promotion of activities solely focused on SMEs should be considered.Large companies already usually send enough signals to financial markets and also can create international R&D collaboration networks through FDI, internalization or other initiatives.
corresponds to Georghiou's (2001) criticism that the quality of Eureka's innovation projects seem to be diminishing over time.Or it could also mean that the questionnaires are failing in capturing the real complexity involved in the process (Georghiou, 1997) or are simply influenced by too optimistic respondents (Huggins, 2001), biasing the analysis using its data.Nonetheless, Eureka has been analyzed with companies' relatively objective data (as financial performance) -see Bayona-Sáez and García-Marco (2010) for an example -and this raises some serious questions on to what extent can Eureka be actually influent on companies' market performance as a whole.And even when one considers this as plausible, does it refer to the proper causality direction or companies do engage in Eureka projects when they are already more prone to achieve better performances?One has to be careful when assessing and interpreting such kind of "self-selection" issue.
Thus, the main contribution of our analysis produces a fairly robust cluster structure, dividing participants in 3 groups.Such procedure allows a segmented examination of firms' characteristics and achieved results.RTD policy evaluation can benefit from this methodology by providing markets' with solutions that are more likely to suit potential participants and that share different characteristics.Edwards-Schachter, Castro-Martínez and Fernández-de-Lucio (2011) make a contribution in this regard by suggesting that cooperative innovation policy should consider firm specific characteristics.
This step also allowed the confirmation of the idea that commercial achievements are strongly affected by the insertion of results in the market before or by the end of the project.
Cluster 1 was classified as risky innovators.One interesting aspect of this group in particular is that it seems to perform better than the other clusters except for the case of commercial results, which shows a very heterogeneous pattern.
Cluster 2 represents companies with poor market performance by the end of the project but with satisfactory technical results, therefore Inventors, and Cluster 3 is composed by moderately successful companies or consistent innovators.Cluster results also showed that both technological (marginally) and commercial (significantly) achievements are quite strong separation variables for groups of firms within the sample.Crossing this analysis with other Eureka samples (from different periods and territories) can be an interesting exercise for future validation of a Eureka-wide typology of participants, since we dealt with a relatively small sample of companies for a limited timeframe.
The results of the cross-tabs (chi-square) approach aimed at generating some insights on the relationship of descriptive variables of companies, i.e., those that are not related to their participation in the project, and impact variables that Thus, practical and useful results of our research suggest that the use of operational indicators should be taken into account together with the existent evaluation methodologies -which in the case of Eureka mainly derive from case studies and descriptive tables.Also, developing and updating taxonomies built on these indicators, for as imperfect as they can be, shall contribute to a better management of processes within initiatives, at least suggesting relevant information for interventions and control.
Hence, efforts in the sense of continuously evaluating the Spanish participation in Eureka have to be performed in order to complement and even provide a different perspective than the one presented in this paper.Nonetheless, achieved results are quite insightful and do well in offering an assessment of Spain's participation in Eureka.Future research should aim at combining data contained in both Eureka's reports and objective economic data available at the micro level.Also, comparing innovation impacts between different technological initiatives would result in even more relevant knowledge regarding policy evaluation.

Table 1 .
Comparison between Spanish Firms and Total of Participants in Eureka J. Technol.Manag.Innov.2012,Volume 7, Issue 3 method performs better than the K-means method -the Hierarchical method was also tested but its results did not seem to be analyzable.The Ratio of Schwarz's Bayesian Criterion (BIC) Changes was the test used for establishing the optimal number of clusters for the sample.Chi-square tests for the classification relevance of variables were also performed.The specific variables included in the settings of the cluster are: Companies' Size, Role in the Project (as Main player or Partner and as Producer, End user, Supplier, Research, Other or Multiple), Functioning of the Project, Overall Technological Achievements, Industrial Exploitation by the Respondent's Company, Product Already on the Market and Commercial Achievements.