اثرات ویژگی های شبکه بر عملکرد خوشه های نوآوری

Industry clusters provide not only economic benefits but also technological innovation through networking within a cluster. In this study, we analyze network-specific structural and behavioral characteristics of innovation clusters with the intention of delving into differences in learning performance in clusters. Based on three representative networks of real world, scale-free, broad-scale, and single-scale networks, the learning performance of entire organizations in a cluster is examined by the simulation method. We find out that the network structure of clusters is important for the learning performance of clusters. Among the three networks, the scale-free network having the most hub organizations shows the best learning performance. In addition, the appropriate level of openness that maintains long-lasting diversity leads to the highest organizational learning performance. This study confirms the roles of innovation clusters and implies how each organization as a member of a cluster should run their organization.

“Industry cluster,” also known as “business cluster,” or “innovation cluster,” refers to geographically close groups of interconnected companies and associated institutions in a particular field, linked by common technologies and skills across horizontal or vertical supply networks (Porter, 2003 and Wixted, 2009). Increased collaboration between organizations in a cluster leads to an economy in which networking becomes the most characteristic feature of a business organization (Deman, 2008). In particular, as the world is moving toward a knowledge-based economy and levels of information are advancing, industry clusters provide not only economic benefits but also individual firms’ technological innovation through networking within the various organizations in a cluster. It also enhances the firms’ overall competitiveness in related industry. Numerous studies on the network effects between diverse firms have been conducted, focusing on the way in which innovation clusters are continuously achieving competitiveness. (Asheim and Isaksen, 2002, Boschma, 1999, Feldman et al., 2005, Giuliania and Bella, 2005, Gilbert et al., 2007 and Oakey, 2007). Geographical advantages of being located in a certain region vary depending on how to capitalize on local resources including knowledge. From the perspective of innovation, sources of effects within a cluster must be clearly analyzed. However, dynamic mechanism in business habitats like innovation clusters is not easy to identify. According to a study on social networks by Granovetter (1973), case analysis cannot fully clarify how certain characteristics of knowledge networks specifically contribute to the performance of the networks. Nor can they identify the dynamic processes. Also, the difficulties of getting information on business network between organizations prevent researchers from field research. Therefore, the aim of this research is to analyze the interconnections between the factors taking effect in the innovation cluster through simulation. We analyze network structures and topological characteristics of innovation clusters with the intention of delving into differences in organizational performance resulting from such network-specific structural and behavioral characteristics. For the purpose of our study, we used the network types suggested by Amaral, Scala, Barthelemy, and Stanely (2000); scale free, broad scale, and single scale networks. These three networks are considered to best represent the Small World (Watts & Strogatz, 1998), the equivalence to the real world. Amaral’s three types of network are adopted to represent the diverse inter-organizational learning relationships within an innovation cluster. Based on those three innovation cluster types, structural and behavioral factors influencing performance of innovation clusters are systematically analyzed, focusing on the research questions below: • How does learning performance of innovation clusters differ depending on network structures of clusters? • How does learning performance of innovation clusters vary depending on the learning rate of organizations within clusters? • What factor leads to organizational diversity that affects the learning performance of organizations? In order to address these questions, this paper is organized as follows. In Section 2, we review the previous studies on innovation cluster, network structure, and learning performance. In Section 3, we propose our research method including the simulation algorithm. In Section 4, we compare and analyze the learning performance of an innovation cluster in the diverse situations derived from network structures and openness. Finally, the implications and further research issues are discussed in Section 5.

5.1. Discussion of Results This study analyzes the network structures and relationship characteristics of major innovation clusters, and how the structural aspects and learning openness of these networks influence performance of organizations in clusters. For three representative networks, scale-free, broad-scale, and single-scale networks, the learning performance of entire organizations in the cluster is examined by the simulation method. Furthermore we identified the specific factors influencing learning performance. In short, among the three networks, the scale-free network having the most number of hub organizations shows the best learning performance. Also, an appropriate level of openness which maintains long-lasting diversity leads to the highest organizational learning performance. However, openness beyond a certain level takes away diversity of competences from individual organizations within a cluster, and valuable knowledge from innovation clusters, which implies that distinctive competences must be kept in the process of inter-organizational learning. That is, unconditionally higher openness will not necessarily be beneficial to overall learning performances in clusters. However, at early stages of learning within clusters, higher openness will enable superior knowledge in cluster to spread seamlessly, leading to quick improvement of overall performance in short-term periods. This result is consistent with the findings of some empirical studies (Lauren & Salter, 2006; Leiponen & Helfat, 2010) which suggest firms that are more open to external sources or channels are more likely to have a higher level of innovative performance but there are moments or tipping points after which openness can negatively affect innovative performance, so external sources need to be managed carefully so that search efforts are not dissipated across too many search channels. Therefore this study confirms the roles of innovation clusters and suggests how each firm as a member of a cluster should run their organization. The results from our research have the following implications. Firstly, an organization in a cluster should keep an appropriate level of openness to contribute to better overall performance of the corresponding cluster. For this, each organization needs to pay constant attention to factors influencing openness such as corporate absorptive capacity, cultural factors, geographical factors and management directions. Simply by aggregating and locating organizations physically into an innovation cluster together, high synergetic effects cannot be guaranteed. On top of everything else, it is important for each organization to share knowledge and have an adequate level of willingness to cooperate with one another in an innovation cluster from the perspective of long-term periods. At the same time, it is necessary that organizations should make more efforts to boost mutual exchange of staffs and information to share tacit knowledge as well as explicit knowledge. Secondly, policy makers in charge of general operation of innovation clusters should endeavor to form hub organizations which put organizations together and support those organizations continuously. It must not be overlooked that organizations serving as leading hubs, regardless of actual tangible performance, can make optimal catalysts in knowledge sharing within a cluster. Influential measures worth considering for network centralization include provision of networking opportunity for firms, interest in networking activities, geographic and cultural characteristics, and existence of leading organizations. Dual appointment of outside directors, provision of networking opportunities for start-ups and networking opportunities among employees in different organizations may work. Lastly, long-term development of an innovation cluster eventually comes down to diversity. When the entire environment calls for diverse knowledge, a cluster limited to a certain industry or localization will end up with unfavorable performance. The diverse knowledge base will promote a new combination, through which the market can provide proper values. For that purpose, central hubs need to be created and voluntary exchange of knowledge should be fostered. Furthermore, in earlier stages of cluster formation, every organization needs to spread superior knowledge faster and supply it seamlessly and strategically to improve overall performance throughout the cluster. Along with these efforts, after the cluster formation, organizations should endeavor to maintain an appropriate level of openness for long-lasting diversity. 5.2. Limitations and future study directions Our study has the following limitations, which require further research efforts. The first issue is about the simplicity of payoff function used for performance evaluation of innovation clusters. The basic payoff function used in our study is the one that March (1991) used. This is most ideal when there is no restriction on learning at all. In reality, however, the appropriate level and difficulties of knowledge sharing differ across clusters. There may exist other restriction factors curbing learning performance. Accordingly, to get more realistic and practically meaningful results, the payoff function calls for generalization. Another challenge is the cluster evolution process. The present study examines the openness of individual organizations to external knowledge and the influence of network structural centrality on overall cluster performance without demonstrating the actual evolution process and path-dependency of cluster networks. To put it another way, further study should address dynamic evolutionary processes taking place in the course of forming, discarding, and merging networks as new organizations emerge in clusters. The final issue is the limitations of the simulation-based analysis. Our study generates and analyzes data through simulation, which must be supported by objective verification. For this, an in-depth case study and questionnaire survey through the major innovation clusters could be analyzed and compared with the simulation results. However, the simulation approach has shown more generic principles of collective innovation in the regional or industrial cluster. We can have a clearer theory than in the practice in a macro level of innovation where it is very difficult to collect an overall dataset in a longitudinal setting. So this can be an exploratory attempt to understand the industrial clusters at its own level of analysis.