Localisation in knowledge-creating activities: Evidence from Japanese patent data

Hiroyasu Inoue, Kentaro Nakajima, Yukiko Saito

11 February 2015



Knowledge spillovers are a crucial ingredient for knowledge creation, which is one of the major drivers of the growth of modern economies. Since Marshall (1890), it has been well recognised that geographical proximity enhances knowledge spillovers and idea exchanges, which causes urban agglomeration (e.g. Davis and Dingel 2012). Japan and many European countries have been implementing an industrial cluster policy, a set of measures to promote the formation of industrial clusters, in an attempt to induce innovation through the closer geographical proximity of businesses.

It has been pointed out that collaborations enhance the knowledge spillovers between organisations with different knowledge stocks, which facilitates great innovations (e.g. Berliant and Fujita 2008). In Inoue et al. (2013), we analysed establishment-level collaboration relationships in patent inventions in Japan, and found that collaboration relationships are significantly localised. Furthermore, this tendency of geographic clustering hardly changed between 1985 and 2005 despite the vast development of information and communications technology (ICT) during this period. These imply that geographic distance significantly impedes collaboration and has not been overcome even with the development of ICT.

Agglomerations of knowledge-creating establishments

If geographical proximity is important for knowledge spillovers and idea exchanges, knowledge-demanding establishments should be more localised in certain regions to pursue more knowledge and ideas from other establishments. Carlino et al. (2012) investigated the pattern of localisation in research and development laboratories in the northeast corridor of the US, and found them to be more concentrated than manufacturing activities.

In our recent paper (Inoue et al. 2014), we analyse the localisation of knowledge creation activities in Japan. We use comprehensive data on overall patents published in Japan. Based on Japanese patents, we identify establishments that applied for patents and construct a unique establishment-level dataset. Then, we test the localisation of research establishments.

Figure 1 shows the overall distribution of establishments, including non-research establishments, and Figure 2 shows the distribution of research establishments. The colour in each grid represents the ratio of establishments. Red represents a higher ratio, and blue represents a lower ratio. From Figure 1, we confirm that establishments are localised in several regions, such as Tokyo, Osaka, and Nagoya, but broadly distributed over Japan. On the other hand, Figure 2 shows that research establishments are more concentrated in narrower regions.

Figure 1. Overall distribution of Japanese establishments

Figure 2. Distribution of research establishments

Difference in creativity

We then focus on the creativity of the establishment. Creative establishments would demand more knowledge spillovers from other establishments. If so, creative establishments would be more localised to pursue knowledge spillovers. To confirm this prediction, we focus on the number of patents created as a proxy for creativity. If our prediction were correct, the number of patents would be more localised. Figure 3 shows the distribution of patent creation. Again, the colour in each grid represents the ratio of patent creation within the grid. The distribution of patent creation is found to be more localised, which supports our prediction.

Figure 3. Distribution of patent creation

Formal analysis

We formally test the above findings by applying Duranton and Overman’s (2005) K-density method. We find that research establishments are significantly localised in the range of 80km at the 5% level.

Furthermore, we focus on technological differences and creativity differences. Demand for knowledge spillovers would depend on technology and establishment creativity. First, we focus on technological differences. We assume that inventing high-technology patents requires advanced knowledge and greater exploitation of the knowledge of other establishments compared to inventing low-technology ones. In this sense, establishments creating high-technology patents should be more localised. Table 1 shows the top 10 technology classes in terms of the degree of localisation. Most of them are in the high-technology classes. This result supports our theory that knowledge-demanding establishments are more localised.

Table 1. Top ten technologies (degree of localisation)

Next, we identify the creativity of establishments by the number of patents created and by the number of citations received. Then, we test the localisation weighting with the establishments’ creativities. We find that the degree of localisation is greater in the creativity-weighted analysis than in the non-weighted one. This implies that more creative establishments – in terms of the number of patents created and citations received – are more localised, as shown visually in Figure 3.

Concluding remarks

Localised knowledge spillovers have been the theoretical background for cluster policy. On the other hand, it is pointed out that distance does not matter anymore because of the vast development of ICT. We find that knowledge-demanding establishments tend to be clustered, and that geographic distance is more crucial for creative establishments. Our results suggest that geographically localised knowledge spillovers between establishments are still crucial for innovation.

Editor’s note: The main research on which this column is based (Inoue et al. 2014) first appeared as a Discussion Paper of the Research Institute of Economy, Trade and Industry (RIETI) of Japan.


Berliant, M and M Fujita (2008), “Knowledge Creation as a Square Dance on the Hilbert Cube”, International Economic Review 49(4): 1251–1295.

Davis, D and J Dingel (2012), “A Spatial Knowledge Economy”, NBER Working Paper 18188. 

Duranton, G and H G Overman (2005), “Testing for Localization Using Micro-Geographic Data”, Review of Economic Studies 72(4): 1077–1106.

Inoue, H, K Nakajima, and Y U Saito (2013) “Localization of Collaborations in Knowledge Creation”, RIETI Discussion Paper 13-E-070. 

Inoue, H, K Nakajima, and Y U Saito (2014), “Localization of Knowledge-creating Establishments”, RIETI Discussion Paper 14-E-053. 

Marshall (1920), Principles of Economics, London: Macmillan.



Topics:  Productivity and Innovation

Tags:  knowledge, spillovers, knowledge spillovers, innovation, patents, Japan, agglomeration, ICT

Associate Professor in the Department of Business Administration, Osaka Sangyo University

Associate Professor at the Graduate School of Economics, Tohoku University

Senior Fellow at the Research Institute of Economy, Trade and Industry, Japan