Today’s economies are increasingly interconnected through global value chains (GVCs) and their relative position in this network has changed significantly over the past decades. In particular, most East Asian countries have achieved rapid economic growth while becoming increasingly interconnected through regional value chains. This has been the case especially for China, which has both increased participation in GVCs and become a central hub through increasing breadth and depth of interconnections with foreign customers and suppliers. Japan, however, is an exception in this trend.
Although Japan remains an important player in GVCs, it has become less ‘central’ within the network as suggested by recent studies such as Criscuolo and Timmis (2018a). Japan has been increasingly participating in GVCs as a supplier of inputs used in other countries for further exports – increasing their vertical specialisation. However, Japanese firms/industries have lost their relative influence within the regional GVC network. This period of falling centrality for Japan coincides with low productivity growth and sluggish innovative activity of Japanese firms.
How do these changes in the relative position and degree of participation in GVCs affect firms’ innovation? How does being more ‘central’ in a GVC network or being more vertically ‘specialised’ affect firms’ innovation performance?
In Ito et al. (2019), we calculate several network centrality measures to identify sectors that are central hubs and those that are peripheral in the GVC network, using OECD Inter-Country Input-Output (ICIO) Tables over the period 1995 to 2011. Central sectors are typically those that are highly connected (both directly and indirectly) and influential within global production networks. Conversely, peripheral sectors exhibit weak linkages to other sectors and are less influential. First, we investigate whether firms in sectors that are increasingly central in the GVC network become more innovative (in terms of patent applications). For instance, they may benefit from more knowledge spillovers from other foreign sectors in the network. Yet, many patenting firms are multinationals, with affiliates spread over many different industries and countries and thus potentially benefitting from additional knowledge spillovers beyond the parent’s sector. We therefore also examine whether multinationals that have foreign affiliates in increasingly central country-sectors become more innovative.
GVC embeddedness and patenting by Japanese firms
We construct measures of network centrality at the country-industry level, using the Bonacich-Katz eigenvector centrality metric which has recently been implemented in several studies to identify key players in a network (See Criscuolo and Timmis 2018a for details on the calculations of the centrality measures). More specifically, we construct backward and forward centrality measures: The former reflects central customers in the network while the latter reflects central suppliers. We also construct measures for the degree of participation in GVCs which reflects the extent countries/industries/firms are involved in vertically fragmented production chains. Following convention, we construct both the backward GVC participation measure (i.e. the value of imported inputs in the overall exports of a country) and the forward GVC participation measure (i.e. use of foreign inputs in exports (backward participation) and the use of domestic intermediates in third country exports (forward participation)).
Figure 1 shows the trends of aggregate GVC centrality and participation over time. We clearly show that while Japan has been increasingly participating in GVCs (left panel), its aggregate centrality has been declining (right panel).
Figure 1 Japan’s GVC participation and centrality, 1995-2011
Turning now to innovation measures, the number of patent applications at the Japan Patent Office gradually increased in the late 1990s but has been declining since the mid-2000s. Moreover, the average citation-weighted number of patent applications per firm also shows a declining trend (Figure 2). At an industry-level, we observe a weak positive relationship between changes in the GVC centrality and changes in the average citation-weighted number of patent applications per firm from 1995 to 2011. This correlation suggests that the declining GVC centrality may explain at least part of the decline in quality-adjusted patent applications by Japanese firms.
Figure 2 Average citation-weighted number of patent applications per firm for major industries
Note: Only firms with at least one patent application
Knowledge spillovers from downstream foreign customers
The trends noted above are suggestive of a positive relationship between GVC centrality and innovation. To show this formally, we look at whether firms in industries that become more central within GVCs also observe an increase in their innovation output, proxied by patent applications. Given their role as key customers or suppliers, they may be more likely to have access to a greater variety of skills and technologies embodied in foreign outputs or inputs. At the same time, the increase in backward and forward GVC participation, i.e. vertical specialisation, may also affect firms’ innovation activities for instance through the growth of offshoring which may allow domestic resources to be reallocated towards more innovative activities (e.g. Bloom et al. 2013).
We therefore examine the relationship between innovation outcomes of Japanese firms and our GVC centrality and participation measures, by estimating a linear model using the matched patent-firm panel data for Japan. We use China’s WTO accession as an instrumental variable (IV) for changes in Japanese centrality.
We find that indeed increasing forward centrality (i.e. having access to a greater breadth of customers directly and indirectly) tends to be positively associated with firm innovation activities (measured as the number of patent applications) particularly in the case of exporters. Our results are robust to using the IV strategy, suggesting that the positive link between GVC centrality and innovation is not solely driven by a market size effect whereby more participation in international trade leads to more innovation as in Aghion et al. (2018). We also find strong evidence that multinational firms with foreign affiliates in countries or industries with higher network centrality are more likely to apply for higher quality patents.
Our results suggest that firms located in key hubs in GVCs appear to benefit from knowledge spillovers from downstream foreign markets, complementing the literature on FDI spillovers for firms with multinationals as customers (e.g. Javorcik 2004).
Indeed, we find that while Japanese firms/industries seem to have become increasingly embedded within GVCs, embeddedness alone does not clearly translate into more innovation. What seems more beneficial to the development of new technologies is being more ‘central’ in the network and having access to a greater breadth of customers.
Aghion, P, A Bergeaud, M Lequien and M J Melitz (2018), “The impact of exports on innovation: Theory and evidence”, NBER Working Paper 24600.
Bloom, N, M Draca and J Van Reenen (2016), “Trade induced technical change? The impact of Chinese imports and innovation, IT and productivity”, Review of Economic Studies 83: 87–117.
Criscuolo, C and J Timmis (2018a), “GVCs and centrality: Mapping key hubs, spokes and the periphery”, OECD Productivity Working Papers No. 12, February, OECD Publishing.
Criscuolo, C and J Timmis (2018b), “GVC centrality and productivity: Are hubs key to firm performance?” OECD Productivity Working Papers, No. 14, June, OECD Publishing.
Ito, K, K Ikeuchi, C Criscuolo, J Timmis and A Bergeaud (2019), “Global value chains and domestic innovation”, RIETI Discussion Paper Series 19-E-028, April, Research Institute of Economy, Trade and Industry.
Javorcik, B S (2004), “Does foreign direct investment increase the productivity of domestic firms? In search of spillovers through backward linkages”, American Economic Review 94(3): 605–627.
The main research on which this column is based first appeared as a RIETI Discussion Paper No. 19-E-028, available here.