We evaluate the "Industrial Cluster Project" in Japan initiated by the Ministry of Economy, Trade and Industry (METI) in 2001 in terms of industry-university partnership (IUP), using original questionnaire data of small and medium enterprises (SMEs). In this paper, we use the number of patent applications as the measure of both the performance of the cluster project and the industry-university partnership. Specifically, we test the following hypotheses: 1) The SMEs that participate in the cluster project apply for more patents than those that do not. 2) The effect of participation in the cluster project on R&D productivity is enhanced by collaboration with national universities within the same cluster area. We collected the data of 229 R&D intensive SMEs with up to 300 employees through a survey conducted in 2005. We employ negative binomial regression to test how participation in the cluster project affects R&D productivity, controlling for firm characteristics such as the number of employees, R&D intensity, the number of IUP projects, the dummy variable for collaboration with national universities, the dummy variable for joint R&D, the dummy variable for collaboration within cluster regions, and industry dummies. Moreover, we estimate the treatment effect model and the instrumental variables (IV) regression, considering the possibility that participation in a cluster project is endogenous. We use firm age as an instrumental variable because the cluster project aims at attracting start-ups and young firms. The estimation results can be summarized as follows. First, participation in the cluster projects alone does not affect patent application. Rather, local firms collaborating with partners outside the cluster show higher R&D productivity in general. Second, the cluster participants apply for more patents when they collaborate with national universities in the same cluster region. Further results reveal that, in this case, the quality of applied patents measured by the average number of claims does not significantly decrease, which is not in line with the argument that cluster firms are subject to administrative pressures to show off the performance of the cluster projects.
