نظم افزایشی اول و دوم و پیامد یادگیری در برنامه های مشارکتی تحقیق و توسعه

In this paper, we distinguish between firm-level learning effects that result from ‘first-order’ and ‘second-order’ additionalities in innovation policy interventions. ‘First-order’ additionalities represent direct firm-level R&D subsidies, whereas ‘second-order’ additionalities result from knowledge spill-overs, horizontal knowledge exchanges between firms, and from other meso- or community-level effects. Analyzing data from collaborative R&D programs in Finland, we show that enhancing identification with a community of practice among R&D program participants (proxy for second-order additionality) enhances firm-level learning outcomes beyond those resulting from direct R&D subsidy (proxy for first-order additionality). Learning effects facilitated by second-order additionality are not confined to technological learning alone, encompassing also business and market learning. We also show that aspects of program implementation enhance identification with a community of practice, which then mediate the relationship between program implementation and firm-level learning outcomes.

During the past couple of decades, innovation policy interventions have progressed beyond promoting first-order additionality through R&D subsidies. Increasingly, policy initiatives seek to generate also second-order additionality1 and promote externalities that facilitate firm-level innovation and learning outcomes (Cantner and Pyka, 2001, Malerba, 1997 and Park, 1999). One rationale for such interventions draws on the recognition of the importance of knowledge spill-overs as a facilitating mechanism for innovation and learning within ‘communities of practice’ (Brown and Duguid, 1991 and Jaffe and Trajtenberg, 1993). When knowledge spill-overs are present, economic agents can learn even without necessarily undertaking internal R&D by themselves (Katz and Shapiro, 1986 and Oerlemans and Meeus, 2005). Therefore, by facilitating knowledge spill-overs within communities of practice, innovation policy interventions may create a learning-enhancing externality that promotes firm-level innovation and learning outcomes beyond the direct effect of firm-specific R&D subsidy (Mahmood and Rufin, 2005). When policy interventions address a collective of firms and research institutions simultaneously, the firms targeted by such interventions may start identifying themselves with a community of practitioners, the participants of which share an active interest in the development and use of a given technology (Kaufmann and Tödtling, 2001). By experimenting around a given technology and related business practice, a community of practitioners would operate as a locus for knowledge externalities that would help boost firm-level learning beyond the direct effect of firm-specific R&D efforts (Powell et al., 1996). Hands-on, externality-focused innovation policy interventions are well exemplified by the National Technology Programs of the Finnish National Technology Development Agency, Tekes. Typically, a national technology program would seek to enhance the technological capabilities of the Finnish industry by carefully building a core group of firms and universities to jointly address a sector-specific innovation challenge. A national technology program could seek, for example, to enhance Finnish industrial capability in the manufacture of injection-molded plastic parts—an important industrial capability for the mobile telephony industry. The production of injection-molded plastic parts involves intense interactions among various players in the value chain, such as mold manufacturers, mobile phone designers, plastic part manufacturers, mobile phone manufacturers, as well as materials suppliers. To function effectively, these need to implement shared enabling technologies (such as shared computer-aided design systems) to facilitate coordination and the resolution of technical problems that may emerge when launching new mobile phone designs. The national technology program would assemble a core group of university departments and research institutes specializing in relevant technologies, as well as key firms undertaking the various value chain activities. Joint research projects would be set up, typically involving 1–2 universities and 1–3 industry participants. A program steering group would be formed to initiate, select, monitor, and prioritize projects, as well as to foster links and interactions among program participants. The program would be actively managed, and seminars and workshops would be arranged to foster exchange of experiences and knowledge, as well as promote collaboration among program participants. By actively fostering links among program participants, the intervention would attempt to support collective experimentation around given technologies and related organizational solutions, thereby enhancing the development and take-up of boundary-spanning technologies such as CAD–CAM and rapid prototyping applications. As numerous firms would jointly work on similar problems, they would learn from one another, the take-up of boundary-spanning technologies would be enhanced because of enhanced coordination among value chain participants, and the participating firms would discover not only new solutions to technical problems, but also, new ways to derive economic advantage from technological advances. Thus, the program would promote not only innovation around technical advances, but also, around related business practice. The activities of a hands-on, targeted technology policy intervention, such as the injection molded plastic parts program described above, are clearly more encompassing than the simple provision of subsidies for firm-level R&D. What is particularly intriguing in such policy interventions is the emphasis given to the promotion of links and knowledge exchanges among program participants. In addition to correcting market failure through R&D subsidy, the intervention also addresses specific sector-level goals, such as collective learning and experimentation around new technologies, or the take-up of enabling technologies and related coordination solutions. However, such interventions also present challenges for measurement and evaluation. Whereas measuring the firm-level learning impact of direct R&D subsidy (i.e., direct input additionality, or ‘first-order additionality’) is relatively straightforward, measuring learning effects from knowledge spill-overs (denominated here as ‘second-order additionality’) is more difficult (Georhiou and Roessner, 2000). It is particularly challenging to separate one from the other. How can we tell whether policy-measures designed to foster knowledge spill-overs within communities of practice actually promote firm-level learning that goes beyond the direct effect of firm-level R&D subsidy? Even though such effects are often cited as justification for sector-specific, hands-on policy interventions (Malerba, 2002), they have not been demonstrated empirically. We propose that the lack of studies attempting to quantify and measure second-order additionality hampers the development and targeting of effective policy interventions. There is little literature that empirically distinguishes between first-order and second-order additionalities in innovation policy interventions, and therefore, little evidence-based guidance for the design of diffusion-oriented innovation policy interventions. Our objective in this study to take a closer look at both direct and indirect mechanisms through which horizontal innovation policy interventions generate desired learning and innovation outcomes at the firm level. Drawing on primary empirical data, we build and test an empirical model that distinguishes between first-order additionality generated by direct R&D subsidies, as well as second-order additionality generated by community-level diffusion mechanisms. Focusing on selected national technology programs in Finland, we demonstrate how the generation of a second-order additionality boosts firm-level learning outcomes, and how this additionality mediates the effect of program organization on learning outcomes. By so doing, we seek to make several contributions to the literature. First, this is one of the few studies to compare the firm-level impact of direct (i.e. firm-level R&D subsidy) and indirect policy mechanisms (i.e., fostering knowledge spill-overs within a community of practice) using primary empirical data. Second, we develop an empirical approach to measuring the firm-level impact of second-order additionality, operationalized as the firm's identification with a given community of practice. Third, we do not limit our analysis to technological learning alone but consider an array of both technological, market, and business practice learning and show how second-order additionality helps generate broader firm-level learning effects than first-order additionality. Our paper is structured as follows. First, we briefly review theoretical rationales for innovation policy interventions, so as to clarify the distinction between first-order and second-order additionalities. Then, we develop a theoretical model on the impact of direct and indirect policy mechanisms on organization-level learning outcomes. This model is tested using primary survey data collected from among participants of three national technology programs in Finland. We conclude by discussing our empirical findings and the implications of our study for both the theory and practice of innovation policy interventions.

We set out in this study to examine firm-level determinants of first-order and second-order additionalities in collaborative R&D programs. Our study was motivated by the observation that whereas first-order additionality is relatively straightforward to measure, second-order additionality is more challenging to demonstrate because it is influenced by community-level processes. For this reason, for example, policy evaluations are easily truncated to mere input–output impact quantifications, and community-level processes are either ignored or cannot be linked to any meaningful goal functions. Quite often, community-level effects are assumed, but in the absence of tangible measures, the ability of innovation policy measures to generate such effects remains an article of faith. To address this gap, we developed and applied a proxy measure to approximate community-level effects at the level of the focal participating firm, namely, the impact of the R&D program on the focal firm's identification with a community of practitioners within its sector. The use of this proxy was justified with received studies and theory on innovation and communities of practice. Drawing on received theory, we proposed that a shared community identity among collaborative innovation program participants should be a reasonable proxy for the generation of community-level innovation effects, such as shared technological vision among community members, voluntary and involuntary knowledge spill-overs, improved coordination, and knowledge-based externalities for the diffusion of knowledge-based innovations. We proposed that firm's identification with a community of practice should be positively associated with organizational learning, an important precursor to innovative activity. In our theoretical model, we also proposed tangible and observable program-level configuration mechanisms that can be used to strengthen community identification, and therefore, indirectly, organizational learning in the participating firms. The theoretical model was tested with empirical data collected from among the participants of three collaborative technology programs in Finland. Our model received good support from empirical data. Perhaps the most interesting findings concern the differing roles of first- and second-order additionality in generating learning outcomes in collaborative R&D programs. Following expectations, we observed that first-order additionality was positively associated with direct technological learning. Similar effects could not be observed for change in technological distinctiveness or direct market learning, however. These results confirm the effectiveness of direct subsidies in generating first-order additionality, but they also suggest that the learning impact generated by direct subsidies, if not combined with active efforts to mobilize second-order additionality, will be limited to technological learning alone. When analyzing the effect of second-order additionality on learning outcomes, much broader effects were observed. We found that second-order additionality not only served to boost direct technological learning and technological distinctiveness, it also boosted market learning and internationalization learning. These latter two effects signal important horizontal learning effects in collaborative R&D programs, as the participating firms experiment with and discover new business practices, which are then copied by others. Importantly, our analysis shows the second-order additionality to provide a distinctive contribution to learning outcomes, the effect of which goes beyond the effect of first-order additionality. Knowledge spill-overs generated within communities of practice are important, not only because they serve to boost technological learning, but also because they strengthen the focal firm's ability to commercially exploit the technological advances developed. Community-level effects serve to boost learning outcomes in innovation policy interventions, thereby magnifying the effect of direct R&D subsidies. Combined, our findings send a fairly strong signal that the generation of community-level effects is an important aspect of firm-level innovation policy interventions, and it should be explicitly incorporated both in the design and monitoring of such interventions. In short, collaborative R&D programs should explicitly attempt to foster the formation of communities of practice among their participants.4 In order to provide for practice-relevant findings, we also analyzed influences of selected program-level configuration parameters on the strengthening of community identification. We anticipated that how collaborative R&D programs are organized and managed will affect how strong community identification they will be able to generate, and this, in turn, should contribute to learning outcomes. Indeed, we were able to observe full and partial mediation effects for the strengthening of interaction frequency and community identification on direct technological learning, increased technological distinctiveness, and internationalization learning. As for support from program management, a full mediating effect was observed for direct technological learning and partial mediation for internationalization learning. It testifies of the robustness of this effect that it could be observed for both direct and indirect measures of technological learning, as well as for business learning. Encouragingly for the policy-maker, thus, our findings suggest that the organization and management of collaborative R&D programs does matter for the generation of learning outputs, and this effect extends beyond the financial subsidy. This finding may also be challenging for policy-makers, as it highlights the importance of program implementation skill – e.g., hands-on management and interaction with participants – for the generation of desired outcomes. We believe that our theoretical model and its empirical validation have implications for both innovation policy research and practice. To our knowledge, this is one of the first studies to empirically demonstrate how community-level mechanisms generate distinctive firm-level additionality in innovation policy interventions. The effectiveness of firm's community identification as a predictor of both technological and business learning constitutes a highly relevant finding for policy-makers, since community identification can be actively monitored, and collaborative R&D programs can be organized so as to maximize this effect among program participants. The advantages of community building may well be most prominent when the knowledge involved is mostly tacit, since this is when the influence of shared norms and trust on informal knowledge exchanges is strongest. On the other hand, the policy-maker should be mindful that social capital and community building may not generate exclusively positive outcomes. There are studies to suggest that at very high levels of social capital may also reduce exploration, potentially contributing to premature selection of technological options. ‘Over-socialization’ may also result in the dominance of consensus-seeking decision-making mechanisms, therefore decreasing the diversity of technological alternatives (Yli-Renko et al., 2001). More research is required to determine whether community-building can also have undesired outcomes, what those outcomes might be, and under which conditions they would most likely occur. Our study suggests several avenues for further research. Above, we have noticed the dearth of empirical studies that address the organization-level impact of meso-level innovation mechanisms. The dearth of such studies is deplorable, because without empirical testing, the advance of theory is hampered. As Edquist and Hommen (1999) have noted, the emphasis of innovation system research has been more on description, rather than prediction. We believe that to advance the understanding of factors contributing to innovation policy effectiveness, it is important to develop testable hypotheses that predict organization-level innovation outcomes. In addition to advancing research, such research would make an important contribution to innovation policy practice. In this study, we have drawn on recent organization-level research and related theorizing from the field of strategic management, applying some of the measures and techniques developed in that stream to empirically test hypotheses concerning innovation policy measures. Because important advances have been made in strategic management research in understanding organization-level innovation phenomena, we believe that further cross-fertilization with such research streams will prove fruitful. Our findings also suggest pointers for further research on practitioner communities. Because it takes time to build up the shared resources that promote learning and advance knowledge-based innovation, the formation of an effective learning community may potentially take longer than the duration of a single program. If this turns out to be true, it is possible that innovation policy interventions could be sequenced to foster community building early on, followed by community-leveraging measures later. On the other hand, it has been shown that communities may also become a source of inertia, and lock-in to obsolete technologies may result if the community is unable to adjust. Established power relations have a tendency to grow stronger over time, with the result that established institutions may reduce interaction in organizational fields (DiMaggio and Powell, 1983). Thus, while our findings have highlighted the importance of community-building processes for organizational learning, we still know little about how long such communities are effective and when they become structural rigidities that inhibit, rather than support, innovation. This is an issue that merits further empirical research. Another interesting issue concerns technological lock-up to obsolete technological standard. What could and should be done to ensure the flexibility of the system to assume new directions when signs of such lock-up emerge? The mediation effects between program configuration, community identification, and the generation of learning outcomes also merit closer examination. In this study, we have used two fairly crude measures, namely, interaction frequency and support from program management. Clearly, more nuanced operationalizations of program organization would enable a much more detailed study of what factors contribute to the effectiveness of collaborative R&D programs. Richer measures of interaction could reveal more nuanced effects on learning outcomes. Our study has provided encouraging signs that community-level innovation effects can be measured, at least indirectly, at the level of the focal firm. The reader should be mindful of the limitations of our study, however. Most of these result from the study design. We have used cross-sectional data in an ex-post situation. This approach reveals little of the dynamics of the phenomena studied, and the possibility of recall bias cannot be fully discounted. In the future, more longitudinal studies, preferably using repeat measures from the same projects, should be undertaken to examine the temporal dynamic of the community-level mechanisms studied here. For example, while the effect of first-order additionality is likely to be short term and not extend much beyond the duration of the R&D project, the learning effect generated by second-order additionality may be more long-lasting, because the participating firms will continue to remain participants of their respective communities even after the program has ended. Also, because of the nature of knowledge spill-overs, first-order additionality may precede and give rise to second-order additionality in technology programs. Our cross-sectional data cannot tease out such temporal dynamics, and longitudinal data are therefore required to cast light on temporal effects. Second, we have used data from a single country only. Particularly, because we have used community identifcation as an important predictor variable, we cannot fully exclude the possibility that some cultural factors may have influenced our results. There is empirical evidence to suggest, for example, that Finland's business culture lends itself more readily for informal collaborations and knowledge exchanges than USA's (Taylor, 2006). Also, while our study has measured the effect of a hands-on policy approach on second-order externality, such effects may also arise without any active interaction, as, for example, the experience in Silicon Valley shows. Our focus on collaborative R&D programs therefore does not imply that similar second-order effects could not occur in any R&D support programs, hands-on or not. Future studies should determine the robustness of the findings reported here across different national cultures and different national innovation systems. Third, we have used aspects of organizational learning as our main dependent variable. While other empirical studies have linked organizational learning to innovative performance, this is a link that has only been assumed here. Fourth, we have used firm's identification with a community of practitioners as a proxy for community-level effects. Future research should examine alternative proxies of such effects. In conclusion, we have demonstrated that policy-makers can contribute to firm-level learning outcomes by actively facilitating second-order additionality. We hope that this study will inspire further explorations into this important area.